Folia Forestalia Polonica 2013, Vol. 55 (2)

PL ISSN 0071-6677

© 2013 by Forest Research Institute © 2013 by Polish Academy of Sciences

Folia Forestalia Polonica, series A, 2013, Vol. 55 (2), 49–57

ORIGINAL ARTICLE

DOI: 10.2478/ffp-2013-0006

Genetic overload of silver fir (Abies alba Mill.) from five populations from central Bosnia and Herzegovina

Dalibor BallianUniversity of Sarajevo, Faculty of Forestry, Zagrebacka 20, 71000 Sarajevo, Bosnia and Herzegovina, e-mail:[email protected]

AbstrAct

The research was conducted with a view to establishing physiological parameters of the silver fir (Abies alba Mill.) with regard to germination energy, absolute germination and genetic overload produced by silver fir populations. The intention was to identify significant variability within populations and subpopulations, or rather within the two combined, and to give recommendations on the usage and usability of seeds from specific populations. The results form a basis for establishing silver fir genetic overload within five fir populations of central Bosnia and Herzegovina.

The results of research on the characteristics such as absolute seed weight, germination energy and absolute germination are within a scope of previous studies.

The studied physiological characteristics of silver fir seeds in the central Bosnia area showed distinct individual and inter-population variability when compared to variability of the subpopulations analyzed.

The inclusion of non-physiological data (height, DBH, age and the presence of mistletoe) concerning sampled trees proved to be effective new quality of research. No correlation was found between DBH and seed physiological characteristics but increased tree height had a positive effect on seed quality in terms of absolute germination and growth, proportional to tree height. This in turn shows that seeds should be collected only from trees of above aver-age height, because they are less genetically overloaded.

The characteristics of healthy but ungerminated seeds from less tall trees displayed lower parameter values, also indicating that seeds should be collected from taller than average trees only.

The presence of mistletoe had a very interesting effect of increasing rotten seeds percentage, and therefore reducing seed quality. Accordingly, seed collection should be limited to trees not affected by mistletoe, however further research on this aspect is required.

Key words

seeds, germination, ungerminated seeds, rotten seeds, empty seeds, populations

Received 13 December 2012 / Accepted 2 June 2013


Dalibor Ballian50

IntroductIon

Silver fir (Abies alba Mill.) is Bosnia and Herzegovina’s most important coniferous species from the perspective of both the environment and husbandry. In Bosnia and Herzegovina it is widespread along the Dinaric Alps in a few remote areas, of which some are fairly large (Fu-karek 1970). According to the old forest inventory, tall forests with silver fir account for 562 237 ha or about 50% of all tall forests in Bosnia and Herzegovina (Matić et al. 1971). The total reserve mass of silver fir timber in forests is 69 923 000 m3 (1990 data) that is 23% of to-tal tree mass in all tall forests which is equivalent to the share of silver fir in the timber industry. Unfortunately, the general state of health and well being of these for-ests is not good. Nothing like the same type of forest in Croatia, where the silver fir is slowly dwindling due to adverse effects of atmospheric and soil pollution (Tikvić et al. 1995), in Bosnia, the deterioration is due to poor forest management, which is undermining the stability of fir forests and causing increased occurrence of mistletoe (Viscum album var. abietis Beck.) (Uščuplić 1992). Mis-tletoe is a constant companion as mismanagement and unskilled felling of stands favour its spread followed by disintegration of many fir forests by allowing intrusion of aggressive beech into fir stands. Slight damage from atmospheric pollution was registered just before the war, and namely in 1991/1992, near the border with Croatia, in the Bosanska Krajina region and in small areas nearby heavy pollutant producers (thermal power plants, chemi-cal industry, metallurgy, etc.), however, it was largely contained. Not like in Bosnia and Herzegovina, several studies on silver fir germination and absolute weight, as well as its physiological features, have been conducted in Europe (Gagov 1973; Lafférs 1970, 1979; Popnikola 1979). In Bosnia and Herzegovina, however, some plan-ning was carried out along these lines, in particular by Čabrajić (1960) and Đikić et al. (1965).

The literature on seed germination provides exten-sive data, and especially those obtained by: Vanin (1960) who observed germination of silver fir seeds from 40% to 70%, Gajić (1962) – from 14.6 to 49.0%, Machaniček (1967) and Kantor (1967) – from 5 to 89%, Čabrajić (1960) – from 31 to 40%, and Gagov (1973) who found that thepurplish form of seed indicated a germination level from 26% to 37%, whereas the brownish form ranged from 18.3% to 54.9% and the pale yellow form – from 25.9% to

50.1%. Popnikola (1979) showed the values from 27% for the grey form to 46.2% for the intermediate form. Roh-meder (1960) stated that seed germination was 28.2%. According to Gagov (1973), germination values ranged from 22.6% to 50.8%, depending on seed origin. Đikić et al. (1965) claimed that the germination of seeds in Bosnia was 8– 10% after mechanical processing. Finally, Gradi (1973) stated that after storing at temperature from 80C to 100C and humidity from 7% to 8% for three years, seed germination decreased to 10– 15%.

The present study was conducted with a view to establishing silver fir physiological parameters with re-gard to germination energy, absolute germination and genetic overload produced by silver fir populations. The intention was to identify significant variability within the populations and subpopulations or rather within the two combined, and to give recommendations on the us-age and usability of seeds from specific populations. The results obtained form a basis for establishing the genetic overload of the silver fir within the five popula-tions of central Bosnia and Herzegovina.

MAterIAl And Methods

Selection and description of field sites

Silver fir trees from central Bosnia were selected in the autumn when the seeds ripen, based on the full produc-tion of seeds in five populations (fig. 1):1. Igman population consisting of:

a. Ravne subpopulation, limestone, altitude1200 m, northern exposure (Longitude 43°47’16’’– Latitude 18°16’29’’)

b. Lasički Stan subpopulation, limestone, alti-tude 1350 m, southern exposure (Longitude43°45’20’’ – Latitude 18°16’48’’)

2. Fojnica population, consisting of:a. Kozigrad subpopulation, phyllite, alti-

tude 1050 m, northern exposure (Longitude43°56’22’’ – Latitude 17°54’16’’)

b. Bistrica subpopulation, phyllite, altitude 1320 m,eastern exposure (Longitude 43°54’50’’ – Lati-tude 17°52’53’’)

3. Olovo population, consisting of:a. Jelovac subpopulation, limestone, altitude 950 m,

north-western exposure (Longitude 44°03’32’’– Latitude 18°33’49’’)

Genetic overload of silver fir (Abies alba Mill.)… 51

b. Mali Jelovac subpopulation, limestone, altitude870 m, northern exposure (Longitude 44°04’05’’– Latitude 18°36’41’’)

4. Kakanj population, consisting of:a. Mehurić subpopulation, peridotite, sandstone,

altitude 1000 m, eastern exposure (Longitude44°13’13’’ – Latitude 18°13’10’’)

b. Podborovica subpopulation, Werfen, altitude 1100m, north-eastern exposure Longitude 44°12’50’’– Latitude 18°15’09’’)

5. Crepoljsko population, consisting of:a. Vukinjača subpopulation, limestone, altitude

1200 m, western exposure (Longitude 43°58’48’’– Latitude 18°27’35’’)

b. Žorci subpopulation, limestone, altitude 1330 m,south-western exposure (Longitude 43°57’52’’ – Latitude 18°30’16’’).

Fig. 1. Silver fir distribution in Bosnia and Herzegovina (Fukarek 1970) and localisation of the populations studied

All the populations were situated at least 30 km away from each other. Two groups were selected within every population, at different altitudes and on differ-ent exposures, and were examined separately to ensure that the population in question was represented most effectively. Each subpopulation was represented by 10 selected trees growing at least 50– 100 m apart (a given population was represented by 20 trees). Dendrometric

parameters of each tree were recorded, and the presence or absence of mistletoe was noted, and then the data ob-tained were used in correlation analysis. Care was also taken to ensure that the populations differed in their geological – or more accurately- pedelogical features and phytocenosis.

Collection of samples and measuring of the characteristics studied

There were collected 50 cones from each of the se-lected trees. Seed germination energy was studied along with absolute germination, and the numbers of: healthy and ungerminated seeds, rotten seeds and empty seeds were recorded. Seeds collected from the 50 cones were homogenized manually. For germina-tion observations, seed wings were carefully removed in order not to damage the turpentine bubble which is essential for germination. The seed would not germi-nate if there were any turpentine touch to the embryo. Before beginning of the experiment we also investi-gated the pre-treatment of the seeds (wet-cold meth-od), but the results suggested that there was no benefit to be obtained from this procedure.

The seeds were germinated in Petri dishes on filter paper at room temperature (18–22°C) in normal day-light. Four repeat sessions (4 × 100 seeds) were con-ducted, with collections of seed portions for observa-tions on germination energy (after 14 days) and absolute germination (after 35 days). The seeds were treated with Radomil fungicide during germination.

Data processing

The parameters analyzed were processed statistically using SPSS 15.0 package for Windows. The analyses included the following: – individual intrapopulational variability of quantita-

tive characteristics through descriptive indicators: average value, standard deviation, significance of differentiations and variability coefficient (sd/mean × 100),

– interpopulational variability through standard sta-tistical indicators such as average, standard devia-tion, the minimum and maximum values and vari-ability coefficient of the quantities,

– analysis of variance (ANOVA), – canonical discriminant analysis, – interdependencies through correlational indicators.



Dalibor Ballian52

The data on descriptive indicators were analyzed within the populations to establish whether there were any statistically significant differences between the ana-lyzed populations. Correlation analysis was conducted to determine whether there were any co-dependencies between the characteristics – for this we introduced some of the dendrometrical parameters, such as the height and diameter of the trees from which the seeds were collected, as well as tree age and mistletoe damage.

results

Descriptive analysis

As seed weights depend on the population which from the seeds originate, the data from this research (tab. 1) correspond with those on the weight of 1000 seeds ob-tained by Lafférs (1979) from 32.69 g to 76.42 g as well as Gagov (1973) and Popnikola (1979) – from 34 g to 82 g.

Tab. 1. Descriptive indicators

Traits N Min. Max. Mean Std. Deviation

Weight of 1000 seeds (g) 100 32.70 76.35 55.33 8.85

Germination energy (%) 100 2.00 30.50 5.32 5.67

Absolute germination (%) 100 4.00 60.25 24.82 12.68

Healthy ungerminated seeds (%)

100 11.25 71.00 34.94 11.46

Rotten seeds (%) 100 0.00 16.00 3.65 3.16Empty seeds (%) 100 15.00 71.00 36.72 10.11

We were unable to make any comparisons for ger-mination energy, as there are no data available in the literature. The results of this research range from 2% to 30.5% (tab. 1).

As far as absolute germination is concerned, we have access to some information from Čabrajić (1960), Vanin (1960), Gajić (1962), Đikić et al. (1965), Machaniček (1967), Gagov (1973) and Popnikola (1979) and others, with figures from 5 to 70%, which is in agreement with the results presented in tab. 1. Even

though they can play a large part in the assessment of the genetic overload of the silver fir, there is little in-formation available on the number of healthy, ungermi-nated, rotten or empty seeds,. On this note, data is avail-able from Gradečki-Poštenjak (2002), pointing out that poor quality ungerminated seeds account for between 29.20% and 73.77%, without differentiating between ungerminated, rotten and empty seeds, but distinguish-ing between limestone and silicate bases.

Analysis of variance at a subpopulations level

We conclude from the results of the analysis of subpopu-lation variability (tab. 2) that 1000 – seed weights indi-cated no statistically significant differences, even though this was expected on the basis of unprocessed data.

Tab. 2. Analysis of variance at a subpopulations level

Traits Sum ofSquares df Mean

Square F

Weight of 1000 seeds

Between Groups 13.24 1 13.24 0.16Within Groups 7,747.80 98 79.05Total 7,761.04 99

Germina-tion en-ergy


Absolute germina-tion


Healthy unger-minated seeds

Between Groups 6.25 1 6.25 0.04Within Groups 13,016.89 98 132.82

Total 13,023.14 99

Rotten seeds

Between Groups 16.20 1 16.20 1.63Within Groups 972.63 98 9.92Total 988.83 99

Empty seeds


Analysis of variance for germination energy at a subpopulation level showed no significant differences, but when the data analyzed were tested separately for each population by a paired difference test (tab. 2), 1% or 5% statistically significant differences were found (Ballian 2000).

SzmitP

Rectangle


Dalibor Ballian54

separations into two different groups for the function 1: Kakanj and Olovo populations fell into one group and other populations into another, but this was not suffi-cient to obtain significant differences. The reason should be sought in the great variability obtained during the study, by comparison with the research in which mor-phological characteristics of cones, seeds and skin were analyzed (Ballian and Čabravdić 2005), which yielded three completely different groups differing from those in the current study.

Cluster analysis

The results of discriminant analysis, were assumed as potential differentiation by the function 1 and suggest-ed that a cluster analysis can be performed. The results confirmed that there were indeed two main groups, as indicated by the results of discriminant analysis. One group consisted of the Kakanj and Olovo populations, and the other of the Igman, Fojnica and Crepoljsko pop-ulations, as shown in tab. 5 and 6.

Correlational connections of the researched traits

Correlation analysis was conducted for all the dendro-metrical findings and the seeds, germination and ab-solute germination characteristics of healthy seeds as well as those of rotten and barren seeds. Germination energy analysis yielded a striking correlational between

absolute weight and absolute germination of healthy, rotten and barren seeds at a level 1%. The presence of a negative correlation is of particular interest in relation to healthy ungerminated seeds and barren seeds (tab. 7).

The correlation between absolute germination and the parameters analyzed showed connections between the height, absolute weight, germination energy and healthy ungerminated seeds and barren seeds, where the correlation was negative. Here, positive correla-tion is of interest in regard to tree height, revealing that higher germination corresponds to a greater tree height. This is divergent in case of the position of cones on the tree (always on tree top), and relatively poor pol-len produced when compared with other conifers. Taller trees are more easily pollinated, and this reduces self-pollination or inbreeding, which have a negative impact on germination. In case of healthy ungerminated seeds, there was observed a negative striking correlation (the order 1%), for germination energy, absolute germina-tion and rotten seeds. There was also a negative cor-relation with regard to tree height and age, suggesting that as the height and age increase percentage of healthy ungerminated seeds declines.

Correlations between rotten seeds and dendromet-rical as well as physiological tree attributes analyzed were statistically interesting, being of the order 5% in relation to the presence of mistletoe on the trees. There was also a negative correlation observed with regard to

Tab. 5. Displaying frequency and grouping of descriptive indicators

Germination energy

Absolute germination

Healthy ungerminated seeds Rotten seeds Empty seeds

Mean Std. Dev. Mean Std.

Dev. Mean Std. Dev. Mean Std. Dev. Mean Std.

Dev.

Cluster1 3.22 1.90 18.32 7.78 38.68 9.76 2.91 2.58 40.31 9.302 8.46 7.71 34.56 12.43 29.32 11.65 4.76 3.62 31.33 8.90

Combined 5.32 5.67 24.82 12.68 34.94 11.46 3.65 3.16 36.72 10.11

Tab. 6. Displaying frequency and grouping of populations

PopulationIgman Fojnica Olovo Kakanj Crepoljsko

Freq. % Freq. % Freq. % Freq. % Freq. %

Cluster1 20 100.0 20 100.0 0 0.0 0 0.0 20 100.02 0 0.0 0 0.0 20 100.0 20 100.0 0 0.0

Combined 20 100.0 20 100.0 20 100.0 20 100.0 20 100.0


healthy ungerminated seeds. In case of barren seeds, a striking positive correlation of the order 1% for abso-lute seed weight, germination energy and absolute ger-mination as opposed to expected negative correlation.

dIscussIon

No notable results were obtained with regard to the ge-netic overload of the populations studied by means of descriptive statistics, analysis of variance and discrimi-nant or cluster analyses. The best results were shown by correlations between seed physiological characteristics and dendrometrical indicators.

Since silver fir grows in mixed and age-varied for-ests in Bosnia and Herzegovina, it is very difficult to come to a satisfactory decision, based on which sites should be selected for production of reproductive mate-rial. Another difficulty is that there are already existing solutions for central European forests which are very different ecologically from those of Bosnia, where silver fir is of superior quality. The present study attempts to resolve the above question by means of a wide-ranging physiological analysis of silver fir seeds from five popu-lations, each divided into two subpopulations. The posi-

tion of male and female flowers of silver fir is out of the ordinary, as they are located near the top of the tree. In addition, silver fir pollen is one of the largest, and con-sequently the heaviest, conifer pollens, and this places certain restrictions on pollination. As a rule, cross-pol-lination occurs between neighbouring trees, and these features are also associated with self-pollination and inbreeding, resulting in the production of sterile seeds and generating genetic drift or genetic isolation of the population (Hadžiselimović 2005), often purely repro-ductive in nature. The presence of healthy ungermi-nated seeds is associated with the specific physiological processes of maturing in favourable conditions, which can be overcome by simulation of germination by wet-cold pre-treatment.

We deduce from the results of correlation depend-encies that germination energy is markedly related to the absolute weight and germination of healthy un-germinated seeds and barren seeds, which is to be ex-pected, as healthy heavy seeds germinate most rapidly. The speed of germination diminishes with percentage of ungerminated seeds in the sample, either rotten or sterile. Absolute germination acts in the same man-ner, and it should be noted that this is often associated with mechanical damage to the seed. When turpentine

Tab. 7. Correlations between studied parameters

Correlation Height DBH Age Presence of mistletoe



Healthy unger-minated seeds Rotten seeds Empty

seeds

Germination energy 0.157 0.110 –0.013 –0.024 0.337** 0.703** –0.476** 0.163 –0.401**

Correlation Height BHD Age Presence of mistletoe


Germination energy

Healthy unger-minated seeds Rotten seeds Empty

seeds

Absolute germination 0.268** 0.113 0.164 0.145 0.360** 0.703** –0.667** 0.191 –0.575**



Germination energy

Absolute germination Rotten seeds Empty

seeds

Healthy unger-minated seed –0.318** –0.163 –0.213* –0.110 –0.174 –0.476** –0.667** –0.370** –0.179



Germination energy



Empty seeds

Rotten seed 0.175 0.101 0.060 0.214* 0.153 0.163 0.191 –0.370** –0.141



Germination energy



Rotten seeds

Empty seed –0.061 –0.002 –0.001 –0.135 –0.319** –0.401** –0.575** –0.179 –0.141

* Correlation is significant at 0.05 level.** Correlation is significant at 0.01 level.



Dalibor Ballian56

damage occurs, the seed is unable to germinate. It is also of interest that there is a strong positive correlation between tree height and absolute germination (germi-nation increases with tree height). This is at variance with the position of flowers and cones on trees. Taller trees are better able to exchange pollen with other trees and other populations, reducing self-pollination and in-breeding, which results in far greater genetic vitality, as evidenced by a higher proportion of germinating seeds. The analysis of healthy ungerminated seeds yielded a negative correlation of the order 1% for germination energy, absolute germination and rotten seeds. Of par-ticular interest is a negative correlation with tree age and height, suggesting that an increase in the height and age reduces percentage of healthy ungerminated seeds. The explanation as regards the height is that taller trees are better able to exchange genetic material, and the height is also related to age. It would be interesting to obtain further results relating to the aging process and how it affects the increase of the number of healthy un-germinated seeds, but we were unable to explain this within the scope of this study.

One of appealing results obtained concerned posi-tive correlation between rotten seeds and the presence of mistletoe at statistically significant level 5%. This can also be attributed to the loss of nutrients in seeds as a result of taking nutrients from the tree by mistletoe. This can prevent seeds from developing fully and they begin to rot during the germination stage. Mistletoe is one of the most significant parasites affecting the silver fir, (Uščuplić et al. 2007). It reduces seed quality, and thus should be strictly monitored in regard to silver fir natural regeneration. Further research should be under-taken on silver fir seeds and mistletoe impact. It is also interesting that there is a negative correlation between rotten and healthy ungerminated seeds. Therefore, ge-netic barriers which led to ungermination of healthy seeds should be more researched.

As the silver fir is of great importance and econom-ic value in Bosnia and Herzegovina, it should be con-sidered as subject of further long-term attention. The results of the present study can be of great value by add-ing an important part to silvicultural activities related to silver fir propagation and the selection of individuals or sites for the production of reproductive material. Fur-ther care should be also taken over actual extraction of seed characteristics, ensuring that above-average sam-

ples are selected from as many individual tall silver fir trees as possible and avoiding less tall, single-genera-tion, single-level stands so as to assure seed quality.

conclusIons

All the characteristics studied, such as absolute seed weight, germination energy and absolute germination are within the ranges of previous studies.

Physiological characteristics of silver fir seeds stud-ied in the populations of Bosnia and Herzegovina re-vealed significant interpopulation variability compared with the variability between subpopulations. The inclu-sion of tree characteristics (height, BHD, age and the presence of mistletoe) proved to be justified in achiev-ing new quality of research. Tree height had a positive effect on seed quality – or rather on absolute germina-tion, which increased with tree height. Accordingly, seeds should be collected only from the trees of above average height. The share of healthy ungerminated seeds decreases proportionally to tree height, so these should be henceforth collected from silver fir trees of above average height.

The effect of mistletoe is also of interest, as its pres-ence increases percentage of rotten seeds with conse-quential loss of quality. Further research should be con-ducted in this regard.

references

Ballian D. 2000. Intra-population and inter-population variability of some morphological and physiologi-cal characteristics of the silver fir (Abies alba Mill.) in one part of the natural range in Bosnia and Her-zegovina. Annales forestales, 24 (1), 1– 23.

Ballian D., Čabaravdić A. 2005. Međupopulacijska vari-jabilnost nekih morfoloških svojstava obične jele (Abies alba Mill.) iz središnje Bosne. Rad. Šumar. Inst. Jastrebarsko, 40 (1), 5– 18.

Čabrajić T. 1960. Analiza šišarica i sjemena jele i smrče. Šumarstvo, 3– 4.

Đikić S., Jovanćević M., Panov A. 1965. Principi i pers-pektive unapređenja proizvodnje šumskog sjemena u Bosni i Hercegovini. Šumarski fakultet Sarajevo, Posebna izdanja.


Fukarek P. 1970. Rasprostranjenje i raprostranjenosti bukve, jele i smrče na području Bosne i Hercegov-ine, Akad. nauka i umjetnosti BiH, 39 (11), 231– 256.

Gagov V. 1973. Izmenčivost pri semenata ot oblikno-venata ela ot različni populacii v NR Bulgarija. Naučni trudovi na VLTI, Sofija, Tom 19, 51– 56.

Gajić M. 1962. Još jedno saopštenje u vezi sa bojom semena jele (Abies alba Mill.). Šumarstvo, 1/ 2, 75– 76.

Gradečki–Poštenjak M. 2002. Varijabilnost nekih svo-jstava češera i sjemena obične jele (Abies alba Mill.) u dijelu prirodnog rasprostranjenja u Hrvatskoj, Šumarski fakultet Sveučilišta u Zagrebu, Magistar-ski rad, Zagreb.

Gradi A. 1973. New techniques and progress in pro-cesses of extraction of forest seeds. Internat. Symp. on Seed Processing. Bergen, 1 (7), 1– 14.

Hadžiselimović R. 2005. Bioantropologija – Biodiverz-itet recentnog čovjeka. INGEB, Sarajevo.

Kantor J. 1967. Pispevěk ke studiu některych dědčných vlastností jedle bílé (Abies alba Mill.). Lesnícky časopis, 13 (4), 309– 318.

Lafférs A. 1970. Hodnotenie semenášikov jedle pro-veniencií zo Slovenska. Vedecke prace VULH, 12, 25– 62.

Lafférs A. 1979. Evaluation of seed weight in fir of Czechoslovakian and foreign provenance in rela-

tion to modified geographical latitude, to geograph-ical longitude and to particular mountain ranges of Europe. Lesnícky časopis, 25, 111– 125.

Machaniček J. 1967. Stanoveni životnosti jedlovich se-men. Prace VULHM, 34.

Matić V., Drinić P., Stefanović V., Čirić M. 1971. Stanje šuma u SR Bosni i Hercegovini, prema inventuri na velikim površinama u 1964– 1968 godini. Šum. fak. i inst. za šum. posebna izdanja, 7.

Popnikola N. 1979. Morfološke karakteristike i vari-jabilnost sjemena jele (Abies alba Mill.) u prirod-nim populacijama SR Makedonije. Šumarstvo, 2/3, 39– 55.

Rohmeder E. 1960. Bastardierung der Gattung Abies. Silvae Genetica, 9 (5), 136– 137.

Tikvić I., Seletković Z., Anić I. 1995. Propadanje šuma kao pokazatelj promjene ekoloških uvjeta u atmos-feri. Šum. List, 11/12, 361– 371.

Uščuplić M. 1992. Uticaj sistema gazdovanja na po-javu imele (Viscum album L.). Glasnik šumarskog fakulteta, 7– 18.

Uščuplić M., Dautbašić M., Treštić T., Selman E., Mujezinović O., Nišić T., Jokanović B. 2007. Boles-ti i štetnici obične jele (Abies alba Mill.) u Bosni i Hercegovini. Društvo za zaštitu bilja u BiH, Sa-rajevo.

Vanin A.I. 1960. Dendrologija, Moskva.


© 2013 by Forest Research Institute © 2013 by Polish Academy of SciencesReceived 14 January 2013 / Accepted 5 May 2013


ORIGINAL ARTICLE

DOI: 10.2478/ffp-2013-0007

Improved methods of classification of multispectral aerial photographs: evaluation of floodplain forests in the inundation area of the Danube

Tomáš Bucha1 , Martin Slávik2

1 National Forest Centre, Forest Research Institute, T.G. Masaryka 22, 960 92 Zvolen, Slovakia, e-mail: [email protected] Czech University of Life Sciences, Faculty of Forestry and Wood Sciences, Kamýcká 129, CZ – 165 21 Praha 6 – Suchdol,

Czech Republic

AbstrAct

The Gabčíkovo hydroelectric power plant has significantly influenced Danube water regime, thus the condition of floodplain forests in the region. Forest condition has been regularly monitored since 1995 using aerial photos. The subject of this study was to improve the procedure of floodplain forest health evaluation based on digital multispec-tral aerial images. Firstly, the forest mask was created with overall accuracy 89%, and next, tree health was evaluated using defoliation as health indicator. We applied orthogonal transformation of 4 original bands of multispectral im-agery into two-dimensional space. Marginal values of digital numbers (DN) of the first component (New Synthetic Channel – NSC1) were defined by fully foliated willow and poplar. The second component (NSC2) was optimised for damage estimation. Calculated DN values of NSC2 represented a perpendicular distance from the line of DN values of the first component. The distance from the line was proportionate to tree damage extent in a given pixel. We generated linear regression model between pair values of NSC2 and defoliation evaluated for 38 trees in the field, respectively, from aerial photos. A decline prediction resulted in r-square equal 0.86. Finally, we used the model to predict defoliation for each picture element (pixel) of the component NSC2.

Key words

floodplain forest, the Gabčíkovo hydropower plant, defoliation, aerial photos, orthogonal transformation, New Syn-thetic Channel

IntroductIon

The Waterworks Gabčíkovo was put into operation in 1992. Its construction ensured energy exploitation of the Danube; waterways transport at the critical part of the ford and safety against floods. A crucial Danube part – from Hrušov to Sap – was re-routed into a 31 km deri-

vational canal. Negative impacts on floodplain forests are connected with this solution caused by groundwa-ter level lowering due to drainage effect of the Danube riverbed. Therefore, suitable environment for existence of floodplain forests towards preservation of their pro-ductive potential was created by implementing hydro-technical modifications. Forest health state is monitored

Improved methods of classification of multispectral aerial photographs… 59

in transboundary cooperation following the agreement between the governments of the Slovak Republic and Hungary that was concluded in 1995.

Aerial photos are commonly used for ecological monitoring (Morgan et al. 2010). For the purposes of evaluation of forest condition based on infrared pho-tographs, visual interpretation keys were developed (Gross 2000; Ciesla 2000). Damage evaluation is based on a combination of shape, texture, and colour char-acteristics. Automated procedures of the classification were developed for digital aerial multispectral images. There have been carried out assessments concerning forest damage classification based on the use of the maximum likelihood method (Meddens et al. 2011), re-gression analysis (Bucha et al. 2009), neural networks (Klobučar 2010), non-parametric k-nearest neighbour method (Eigirdas et al. 2013). In automated approaches, spectral differences between healthy and damaged trees are used for identifying damage extent. High absorption in the blue and red parts of the spectrum and moderately increased reflectivity in the green part of the spectrum are typical for healthy vegetation. Chronic damage causes deterioration of chloroplasts, and this change in tree physiology results in leaves turning yellow, and then the maximum reflectivity moves from the green part to the red part of the spectrum (Zarco-Tejada et al. 2001). High and stable reflectivity in the near infrared range of the spectrum is typical for healthy vegeta-tion. Our measurements of poplar leaves showed high reflectivity for dry and brown foliage as well. Unlike healthy vegetation it was not uniform but was gradu-ally increasing in a range from 700 to 1100 nm, simi-larly to the pattern observed by Ahern (1988). Slaton et al. (2001) proved relation between infrared reflectance and changes of leaf intra-cellular and cellular struc-ture. Studies on vegetation response to stress showed that within wavelength range from 400 to 850 nm the maximum difference in reflectance between control and stressed vegetation was close to 700 nm (Carter and Knapp 2001; Masaitis et al. 2013).

Delineation of tree crowns is necessary for a forest damage classification. Many approaches have been so far proposed to separate tree crowns. Pitkänen (2001), Pouliot at al. (2002), Šumbera (2003), Wang et al. (2006) proposed the automated delineation of tree crowns for black and white, infrared and multispectral aerial im-ages. The authors assumed that the top of the tree can

be detected at the highest brightness value. The edge of the crown, on the contrary, represents the minimum brightness value.

Hirschmugl et al. (2007) used stereo images to de-rive Digital Model of Terrain (DMT). This was used to distinguish individual tree crowns and combined with procedures based on image spectral properties, namely the maximum and minimum reflectance on the top and edge of the crown. Wolf and Heipke (2007) proposed a procedure for automated delineation of tree crowns from infrared images combined with DMT-based fuzzy logic and approximation of tree crown by ellipsis. Bijker et al. (2010) modelled the probability profiles of single tree crowns with Gaussian functions, resulting in delineation of tree crown objects. Jing et al. (2012) proposed a method for individual tree crown delineation based on multi-scale filtering and imagery segmentation. Pontius et al. (2008) used supervised classification – Spectral Angle Mapper method – to identify non-forested areas.

Differences of radiation reflectivity depend on tree species. In the Danube inundation area, there prevail poplars and willows. Higher reflectivity in the visible infrared part of the spectrum is typical for willows and white poplars when compared to that of black poplar. This makes the model for damage estimation harder to derive. The problem was solved by application of multiple regression analysis in previous cycles of for-est monitoring in the Danube inundation. Apart from aerial photographs, there were also inserted into the model as a dependent variable, data on a representation of tree species taken from the forest management plan (FMP) (Raši and Bucha 2001) or damage classification was carried out separately for willow and poplar forests (Bucha et al. 2009).

The aim of this study was to improve the procedure of floodplain forests health evaluation based on com-mercially available aerial photographs. Specifically, our objectives were to:1. delineate the part of tree crowns on which damage

would be evaluated;2. objectify digital classification of forest condition by

transformation of original bands into the compo-nent optimised for damage estimation;

3. use the component (derived in the objective 2) inconjunction with ground control data to predict for-est health condition over large contiguous areas.



Tomáš Bucha, Martin Slávik60


Study area

The evaluated area of the Danube left bank inundation is located between the original Danube riverbed and the canal of the Gabčíkovo Waterworks (fig. 1). The area is characteristic of developed river branch system, vast complexes of floodplain forests and alluvial mead-ows. Forest area is approximately 3000 ha. According to FMP (2005), floodplain forest species composition within the area observed is as follows: 62% cultivated poplars (Populus sp.), 10% domestic poplars (Populus

nigra L. a P. alba L.), 17% willows (Salix alba, S. fra-gilis L.), 6% ash trees (especially Fraxinus angustifolia VAHL.), and 5% other deciduous trees, especially Eng-lish oaks (Quercus robur L.), acacias (Robinia pseudo-acacia L.), sycamore maples (Acer pseudoplatanus L.) and alders (Alnus sp.).

Aerial photography

The basic parameters of the aerial photography are described in tab. 1. Organisational and administrative works, as well as all the legal requisites of the aerial imaging, were ensured by the company Photomap Ltd.

17°20’0”E 17°25’0”E 17°30’0”E 17°35’0”E

17°20’0”E 17°25’0”E 17°30’0”E 17°35’0”E

47°5

0’0”

N47

°55’

0”N

48°0

’0”N

47°5

0’0”

N47

°55’

0”N

48°0

’0”N

Fig. 1. Study area


Košice. Photography was performed on 10th Septem-ber 2011 by the company ARGUS GEO SYSTÉM, Ltd. Hradec Králové using a digital camera Vexcel Ultra-CamX.

Tab. 1. Selected parameters for aerial photography in 2011

Locality Hrušov – SapScale 1 : 41 800Endlap 60%Sidelap No – photography in 1 stripDate of Photography 10th September 2011 Time of Photography 11:00– 15:00

Type of camera UltraCamX – digital multispectral camera

Spatial resolution 30 x 30 cm

Radiometric resolution Blue – Green – Red – Near-infrared band

Total imaging area ca. 20 000 ha

The photographs were supplied in tif format in 8 bit resolution, in the 3rd level of processing, which means that the multispectral channels (blue, green, red and near-infrared) were spatially modified from 90 × 90 cm into 30 x 30 cm resolution according to the panchro-matic band. The area of Hrušov – Sap was depicted in 30 aerial photographs obtained with overlap 60%. All the photographs were ortho-rectified using Image sta-tion software. The total root-mean square error (RMSE) did not exceed the value of ± 1 m at individual photo-graphs after their transformation to the national S-JTSK coordinate system. All 30 photographs were used for creation of stereo-pairs for visual assessment of tree defoliation for the purposes of derivation of regressive model for damage estimation. The 13 photographs, spe-cifically the images 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26 and 28 (i.e. every second photo from Slovak territo-ry) were used for the classification of forests condition.

Tree defoliation estimation

Defoliation (loss of assimilation organs) was used as the main indicator and basic visual symptom of tree health. The indicator expresses percentage ratio of missing as-similation organs to the ideal status of the assimilation apparatus, i.e. to a sample fully foliaged tree in given conditions.

The aim of tree defoliation assessment was to ob-tain support information for the evaluation of the aerial photographs. Defoliation was assessed during field sur-vey and by visual interpretation of the photos on the monitor screen.

During field survey there were chosen 11 areas with healthy to severely damaged trees. The assessment of defoliation was performed on 114 trees by one evaluator at the end of August 2011, i.e. before image acquisition. Defoliation was assessed using binocular – in percents with rounding off to 5%, following the methodology of the international programme UN-ECE ICP Forests (2006). Only 6 from 114 trees were unambiguously identified on the aerial photos. These trees served as a calibration sample for visual interpretation of defolia-tion from photo stereo-pairs as well as were included into the regression model for damage estimation. De-foliation of further 32 trees was assessed based on ste-reo-pairs and methodology described in EU’s manual (Gross 2000) by the evaluator who carried out a terres-trial assessment. In total, 38 trees were finally used in deriving the regression model for estimation of forest damage based on the aerial photos: 15 willows, 15 pop-lars and 8 dead trees.

Spectral measurements

Spectral reflectivity measurements were conducted with Licor-1800 spectroradiometer. External integra-tion sphere LI-1800– 12 was used for the measure-ments in a spectral range 400– 1100 nm with a 2 nm step. Leaves of black poplar, white poplar, grey poplar, white willow, quaking aspen and dry bark of aspen were analyzed. The leaves were collected in the ar-boretum Borová hora in Zvolen from the lower part of tree crown in the first decade of August. Measure-ments were carried out during app. 60 minutes after sample collection. The aim of the measurements was to compare determined spectral reflectivity with DN values from the aerial photographs and to use the in-formation obtained for verification of the proposed method of forest damage classification.

Creation of the forest mask based on aerial photographs

Distinguishing forest from other land cover categories, that are overall called forestless, is a significant opera-tion because all the characteristics of forest health are


SzmitP

Rectangle


The value 0.5 (middle of the interval 0– 1) was selected based on factual occurrence of various object shapes – from compact to non-compact. The size criterion de-termined the size of the resultant object. The selected value of 100 resulted to number of objects from 390 to 670 in individual photographs. This range allowed ef-fective manual classification of various forestless cat-egories. The “line selection” eCognition functionality which enables selection of several objects with an only draw of a cursor was used. This method of object se-lection was used particularly for the classification of heterogeneous segments as village with gardens, agri-culture areas with hedgerows and individual trees and non-arable areas with shrub succession. Homogenous segments with crops with similar spectral properties as forests as well as water bodies outside forest complexes were also selected by this method and then classified manually into the forestless category.

Automated classification was used for water bodies and shadows occurring spatially inside forest complex-es. The size of these segments is usually smaller what makes the manual classification slow. So water bodies as shadows are distinctly spectrally different from for-est when their DN value in infrared band is lower then 90. The segment with DNIRED < 90 was assigned to theforestless category.

Distinction of tree-crown: pixel-based classification approach

Subsequently, we distinguished tree crowns from other categories (shadows, small gaps in vegetation, roads and other non-forest objects) inside the segments with forest. Generally, the tree crown is not spectrally homogenous. DN values gradually decrease towards crown edges. This is caused by the fact that the top of tree crown is more illuminated and crown edges are more shadowed. We distinguished tree crowns by applying pixel-based approach. We used the method of Isodata that belongs to the group of unsupervised classifications. Isodata clas-sification was performed independently for each of 13 photographs. We specified 9 output classes and the min-imum cluster size: 100 pixels (fig. 5c). Subsequently, we assigned a land cover type to each class. We reclassified the result into the values of 1 – forest and 0 – other cat-egories. The forest mask was obtained with reciprocal combination of outputs from object- and pixel-based ap-proaches to the classification of each of 13 photographs.

First we evaluated the quality of forest masks visually, especially comparing them with photographs. Tree crown were underestimated on the photographs 10 and 12. We enlarged the forest mask (surroundings of clas-sified tree crown) by about 1 pixel in these photos. For-est was overestimated on the photographs 8, 14, 16, 24 and 26, therefore we reduced the mask by about 1 pixel. Overestimation was caused by inclusion of shadowed parts of crowns into the mask. The technique of picture filtration was used for overestimation elimination with kernel size 3 × 3 pixels. No filtering was needed for the photographs 4, 6, 18, 20, 22, 28.

In the next step we created a mosaic from 13 masks. In overlap we took the mask with higher quality accord-ing to a visual examination.

Accuracy of the forest mask

The accuracy of the forest mask was verified on the sample of 100 systematically selected points in a grid 1 × 1 km on the mosaic a subset of segments with for-est. Correctness of each point assignment into forest or forestless category was visually assessed. The result was processed in a form of error matrix.

regressIon Model for dAMAge estIMAtIon

The cornerstone for the derivation of tree damage from the aerial photograph is the intensity of reflected elec-tromagnetic radiation recorded by a digital multispec-tral camera. A close relation between defoliation evalu-ated in the field survey and the value of spectral reflec-tivity recorded with sensors have been proved in several works (e.g. Hildebrandt et al. 1991; Gross et al. 2000; Raši and Bucha, 2001; Bucha et al. 2009). Forest health evaluation was carried out in two stages.

First stage

In the first stage we derived two components using orthogonal transformation of original R, G, B and IR bands. Mathematic apparatus for the components der-ivation was the same as described by Jackson (1983). There is a difference in the way of defining the physical meaning of the derived components. In Jackson’s ap-proach, the first component represents brightness. It was derived from the points with low and high reflectivity, represented by moist and dry soil. The second compo-




nent (greenness) represents the amount of green vegeta-tion. In our approach the first component (new synthetic channel – further NSC1) emphasizes spectral variability of individual tree species. The second component (fur-ther NSC2) is optimized for damage estimation.

We chose fully foliaged woods (black poplar and willow) with very different reflectivity to compute the NSC1 and we calculated the differences of DN values:

bi = (Xw – Xbp)i

where: i – 1st to 4th channel of the aerial photo (IR, R, G, B), X – DN value of willow and black poplar.

Low DN values represent black poplars and Eu-roamerican poplars that have low reflectivity in the spectrum observed (B, G, R and IR). The highest DN values are represented by willows and white poplars which usually indicate high reflectivity. DN values (low and high) were calculated as mean values for several poplar and willow crowns distinguished manually on the aerial photos.

We standardized the vector b = (b1, b2, b3, b4) into the unit vector by dividing each of his elements by the normalization factor B:

A1,i = bi/B

where:

B bii

=

∑ 2

1

4 1 2/

A1,i – are sought coefficients for the first component derivation:

NSC1 = A1,1 × X1 + A1,2 × X2 + A1,3 × X3 + A1,4 × X4

where Xi are DN values of pixel in i-th channel.

While deriving the NSC2 we defined DN value rep-resenting the maximum damage (dead tree) and calcu-lated the difference to any of DN values lying on the line between Xw and Xbp. We chose the value of Xw for a calculation:

gi = (Xdead_ tree – Xw)i – D2,1 × A1,i

where D X X Adead tree w ii

2 1 11

4

, ,= −( ) ×∑

Orthogonality of b and g vectors is ensured with described process. Standardization of g vector was car-ried out by means of normalisation factor G, and trans-formation coefficients were calculated as follows:

A2,i = gi / G

where G gii

=

∑ 2

1

4 1 2/

Calculation of the NSC2 was carried out as follows:

NSC2 = A2,1 × X1 + A2,2 × X2 + A2,3 × X3 + A2,4 × X4

Calculated values of NSC2 component represent a perpendicular distance from the line of DN values of component NSC1 defined by the fully foliated trees of willow and poplar on its borders. The distance from the line is proportional to the extent of the tree damage in the given pixel.

Second stage

The second stage of forest health evaluation concerned obtaining data on tree defoliation during the field sur-vey or based on visual interpretation on the monitor screen carried out in 3D environment (evaluation of de-foliation on stereo-pairs of photographs). For the needs of vegetation health classification we evaluated 38 trees on stereo-pairs of photographs.

Predominant and dominant trees were chosen. Each tree crown was manually vectorised and mean spectral reflectivity was calculated. This spectral char-acteristic was interconnected with defoliation data, and thus we obtained a set of pair values for further analysis. We derived the regression model by means of simple linear regression analysis between the data obtained both in the first stage (NSC2 component de-rived from aerial photographs) and in the second stage (defoliation evaluated in the terrain or from stereo photographs). Finally, we estimated the damage for each picture element (pixel) of NSC2 using regression equation (Šmelko 1990).

results

Forest mask accuracy

The forest mask accuracy was verified on the sample of 100 systematically selected picture elements (tab. 2)


and expressed in a form of error matrix (Congalton et al. 1999).

Tab. 2. Error matrix of the forest mask

Ground truth:Classification result Forest Non-

forest ∑User’s Accu-racy

Error of com-mission

Forest 26 3 29 89.7% 10.3%Non-forest 8 63 71 88.7% 11.3%∑ 34 66 100Producer’s Accuracy 76.5% 95.5% 89.0%Error of omission 23.5% 4.5%

Tab. 2 shows that forest is undervalued. In 8 out of 34 cases, forest was incorrectly incorporated into the category of forestless, which represents 23.5% under-estimation.

On the other hand, there were 3 incorrect cases out of 29 of forest classifications when forestless was in-correctly categorized to forest. This represents 10.3% overestimation of category forest

The overall accuracy of forest mask creation was 89.0%. We consider this result as acceptable, but not op-timal. It points out a need for further refinement of the forest mask delineation algorithm from the photographs with very high resolution.

Transformation of aerial photograph bands for optimization of damage classification

Measuring leaf spectral reflectivity in laboratory conditions

In laboratory conditions, with spectral measuring in wave spectrum from 450 to 1100 nm there was obtained of assimilation apparatus reflectivity of the main tree species in the floodplain forest (fig. 3).

It follows from the results of measurements that wavelengths in the red spectrum (630– 690 nm) and in the infrared spectrum (750 to 770 nm) are suitable for the distinction of dead trees from healthy ones.

Bark reflectivity from dead poplar was higher in the blue and green part of the spectrum than the reflectivity of willow and poplar leaves’ top side and lower (similar) than the reflectivity of the bottom side of willow (pop-lar) leaves. This fact can noticeably influence utilization of the blue and green channel in damage classification,

for example in case of photographs taken in windy con-ditions. Then, there can be expected higher influence of leaf movements on overall reflectivity and spectral display of leaves and bark (dead trees) may become un-distinguishable in the aerial photograph.

450 550 650 750 850 950 1050

Salix alba – topSalix alba – bottomPopulus nigra – topPopulus nigra – bottomBark from deadPopulus tremulea

11500.0

0.1

0.2

0.3

0.4

0.5

0.6

Wavelenght (nm)

Re�e

ctan

ceFig. 3. Spectral reflectivity of the main tree species leaves and tree bark. Top: top of leaf; bottom: bottom of leaf

Derivation of transformation coefficients from DN values of the aerial photograph

Tab. 3 shows DN values (reflectivity) for willow, poplar, and dead trees, based on which we derived coefficients for computing NSC1 and NSC2 components.

Tab. 3. Input DN values derived from the aerial photograph

IR Red Green BlueDN values for poplar, willow and dead tree

according to multispectral bandsSalix alba 211.2 116.9 120.1 95.6Populus nigra 207.8 75.8 81.2 61.2Dead trees 182.2 132.4 123.9 102.5

Calculation of transformation coefficient of 1st and 2nd component

NSC1 0.045 0.615 0.585 0.525NSC2 –0.968 0.209 –0.136 –0.011

Similar relationships between DN values of the aer-ial multispectral bands and reflectivity measured with a Licor apparatus were recorded for white willow, black poplar and dead tree bark. In the visible part of the spec-trum (R, G and B) reflectivity of Populus nigra < Salix alba < dead tree. In the IR channel (755 nm) the relation is as follows: dead tree < Populus nigra < Salix alba.




Calculation of the components was carried out as a linear combination of original aerial photograph bands (fig. 5b). We used transformation coefficients for the photographs 4 to 24:

NSC24– 24 = –0.968 × IR + 0.209 × R + – 0.136 × G – 0.011 × B [1]

For the photographs 26 to 28 we used transforma-tion as follows:

NSC226– 28 = –0.852 × IR + 0.064 × R + – 0.089 × G + 0.512 × B [2]

where: R – red channel, G – green channel, B – blue channel, IR – infrared channel.

The reason for application of two equations was the spectral difference between photographs that is prob-ably related to the process of their pre-processing. Dif-ferent contribution of the blue channel in equations [1] and [2] is mainly visible.

Derivation of regression model for damage estimation

The results of regression analysis carried out on pair values of the first (NSC2 values) and second stages (de-foliation) are regression models (fig. 4), with the help of which we estimated damage for each picture element (fig. 5d – right bottom) of the first stage:

Defoliation4– 24 = 292.7 – 1.393 × NSC24– 24 [3]

The correctness of the classification can be assessed based on the parameters of regression analysis (cor-relation coefficient, standard error of regression line). At relation [3] the correlation coefficient was r = 0.93 and standard error syx = ±13.3%, average deviation di = 10.6%, the range of selection n = 38 trees. Interpre-tation of regression line mean error can be illustrated by the following example: if derived defoliation is 30%, in reality it can deviate in a range ±13.3%, i.e. from 16.7% up to 43.3% with 68% reliability.

When continuous decline rating was rounded to the nearest integer for class comparison (5 classes – see tab. 4 for class range), the model was able to predict decline with 58% accuracy.

210200190

95% con�dence level

1801701601501401301200

20

40

60

80

100

Component NSC2

Defo

liatio

n (%

)

Fig. 4. Graph of a regression model [3] between pair value of first and second stage of defoliation estimation

The parameters (coefficients a, b) of linear regres-sion in equation [4] for the photographs 26 and 28 was deduced in a simplified way, only from trees with no defoliation and dead trees (i.e. tree with 0% and 100% defoliation).

Defoliation26– 28 = 386.0 – 2.780 × NSC226– 28 [4]

Due to simplify way of damage estimation a corre-lation and mean error of regression line are not derived.

Classification of forest damage

Information about damage of forests in the area affect-ed by the Gabčíkovo waterworks is presented in tab. 4 and 5. The scale in tab. 4 results from the classification used in the ICP Forests Program (UN/ECE ICP Forests, 2006). Defoliation was assessed for each pixel classified as forest in the process of forest mask creation.

Defoliation calculated for each pixel according to equations [3] and [4] was a basis for computation of damage in forest compartments. Intensity of damage at a level of forest compartment was calculated as arith-metic mean of all forest pixels in the compartment. The summary results are presented in tab. 5.

In the area investigated, there were only 2.3% of medium damaged stands – mainly in the north-western part of the study area, where a decrease of groundwa-ter level caused degradation of woody plants. Heavily damaged stands with defoliation above 50% were not observed at all.


A B

C D

Fig. 5. A: multispectral aerial photograph, combination of bands: IR/R/G. Cultivated healthy poplar trees are shown in red. Red-gray colour with variable texture indicates medium to heavily defoliated tree crowns. Deadwood is shown in colours from gray to turquoise. B: NSC2 component derived by orthogonal transformation of original bands of aerial photograph. C: Classification of aerial photographs into 9 categories by Isodata method. D: Classification of tree crown defoliation. Green: healthy vegetation. Red: damaged vegetation




Tab. 4. Summary of the results on classification at a pixel level in line with UN-ECE ICP Forests methodology (whole observed area)

Defoliation class

Loss of assimilation organs (%)

% of the pixels in defoliation

class

Description of damage

0 0– 10 38.0 No defoliation

1 11– 25 37.3 Slight defoliation

2 26– 60 23.3 Moderate defoliation

3 61– 90 1.3 Severe defoliation

4 91– 100 0.1 Dying and dead

Total 109

Tab. 5. Summary results of classification at a forest compartment level: frequency of compartments in the defoliation classes in the year 2011

Stand defoliation

(%)

Description of damage

Number of compartments

Frequency distribution

(%) 0– 10 Healthy stands 38 5.3

11– 20Healthy stands with first signs of damage

482 67.1

21– 30 Slightly damaged stands 181 25.2

31– 40 Moderately damaged stands

16 2.2 41– 50 1 0.1

51– 100 Severely damaged stands – –

Together 718 100.0Not evaluated stands 77 –

dIscussIon

A multitask procedure was applied for large area survey of forest condition from multispectral photographs with high resolution (30 × 30 cm). Firstly, we focused our re-search on accurate distinguishing forest from other land cover classes for the reason that all the characteristics of forest health were subsequently derived only from pixels classified as forest. We suggested the method

combining object- and pixel-based approaches for the following purposes:

Object-based approach allowed a time efficient classification of large non-forest segments with the use of combined digital and visual classification. One photograph was processed in 100 minutes. The whole area from Dobrohošť to Sap (13 photographs), includ-ing Hungarian Danube area was classified into forest and forestless categories in 22 hours (i.e. 3 working days). Further rationalisation is possible by applying a fully digital algorithm. However, at average size of segment of about 4.8 ha, there can be anticipated a quite complex algorithm due to heterogeneity inside the seg-ments. Therefore, fast visual classification of the seg-ments based on expert knowledge could be considered as a practical and operative solution.

Pixel-based approach was applied as unsupervised classification with Isodata methods for tree crown de-lineation inside the segments classified as forest. We successfully excluded the most of non-forest vegetation (crops, meadows, bushes, water, shadows) with overes-timation error 10.3% in the category forest. Underesti-mation error of forest category equal 23.5% shows a ne-cessity for developing an improved algorithm focused in particular on tree crown delineation.

Secondly our research was focused on to the meth-od for prediction of forest damage. We proposed the procedure based on orthogonal transformation of aerial multispectral bands instead of commonly used spectral indices (Carter 1993, 1994; Sims and Gamon 2002; Vo-gelmann et al. 1993). The purpose of the transforma-tion was a reduction of four dimensional data into two dimensional space with the aim to optimize informa-tion contents. The first component (NSC1) of orthogo-nal transformation emphasizes spectral variability of individual tree species. The main transformation output is NSC2 component optimized for damage estimation. Consequently, this is a biophysical indicator, the values of which could be recalculated to defoliation through re-gression with field assessment of defoliation on sample of trees. In the derived regression model concerning damage estimation, the value of correlation coefficient was r = 0.93 (r-square = 0.86) and the standard error syx

= ±13.3%. Validation of the continuous decline predic-tion carried out by Pontius et al. (2008) resulted in r-square = 0.71 and RMSE = 0.582 for ash forest. Eigirdas et al. (2013) achieved the highest correlation coefficients


as follows: 0.58 (spruce), 0.60 (pine) and 0.39 (birch) be-tween field-estimated and predicted crown defoliations from Vexcel UltraCam D photographs with 0.5 × 0.5 m resolution. They improved the accuracy by aggregating predicted and field estimated tree crown defoliation val-ues up to the sample plot level. Correlation coefficients between plotwise average values of field-estimated and predicted defoliations were around 0.8.

When continuous decline rating was expressed in 5 defoliation classes, our model was able to predict de-cline for calibration data with 58% accuracy and with one-class tolerance with accuracy 95%. In case of 10 classes with 10% defoliation step, the accuracy would be 39%. Pontius et al. (2008) declared 63% accuracy us-ing linear regression model with 6 variables (6 chloro-phyll and canopy water content sensitive indices) for the assessment of ash decline (10 defoliation classes) from hyperspectral imagery.

When compared to the previous monitoring survey (Raši and Bucha 2009) the method proposed is advanta-geous because of providing possibility to perform forest damage assessment based solely on NSC2 component derived from aerial data, excluding the need of addi-tional information from the forest management plan.

The method was also verified at an operational level. The study shows that the proposed approach can be used to produce detailed maps of flood forest decline embrac-ing relatively large areas, such as 3000 ha of forest evalu-ated in the Slovak side of the Danube in this study. For-est monitoring carried out in 3-year cycle of repetition has proved to be sufficiently accurate and – at the same time – economically satisfactory. The total cost of photo-graphs of 20 thousand ha area was 4000 €, i.e. 0.2 €/ha.

Yet, the study showed some problems with process-ing aerial imagery. Although all the photographs were obtained from one flight line, it was necessary to apply two models for NSC2 derivation. Consequently, there was a need to derive two forest decline prediction models, one for the photographs 4– 24 and the second – for the photo-graphs 26– 28. The reasons of that could be various, e.g. different quality of images pre-processing or different at-mospheric conditions during image acquisition in the up-per part of the study area. This suggest the need to exam-ine an use of original multispectral bands in 90 × 90 cm resolution for the classification, instead of pre-processed photographs at the level 3, i.e. spatially modified into 30 × 30 cm resolution according to the panchromatic band.

From forestry point of view, the continuous mon-itoring in the study area disproved the hypothesis on large-scale and harmful effects of water construction operation on forests situated between Danube original bed and supplying, resp. drainage channel. A version of hydro-technical measures with damming of arm sys-tem and water supplying object with volume of water approx. 30 m3/s, ensures, in most of the area, suitable conditions for healthy development and expected pro-duction of wood in floodplain forests.

conclusIons

The paper describes the improved procedure of survey-ing floodplain forest condition in the Gabčíkovo region based on the classification of digital multispectral pho-tographs. First, we proposed easily practicable method combining object- and pixel-based approaches for de-lineation of the part of tree crowns for damage evalu-ations. Next, the NSC2 component, optimized for the damage estimation, was created by orthogonal trans-formation of the original bands of multispectral photo-graphs. Then, we used the component in conjunction with ground control data to predict forest health condi-tion over large contiguous areas. The results confirmed that NSC2 component was a suitable biophysical indica-tor to provide precise information about forest condi-tion. Furthermore, forest damage assessment was based solely on NSC2 component without using additional in-formation from the forest management plan.

AcKnowledgeMent

This work was supported by the Cross-border Coopera-tion Programme Hungary-Slovakia 2007– 2013 within the project: Innovative methods of inventory and moni-toring of Danube floodplain forests using 3-D Remote Sensing technology (HUSK/1101/1.2.1/0141).

references

Ahern F. 1988. The effects of bark beetle stress on the foliar spectral reflectance of lodgepole pine. Inter-national Journal of Remote Sensing, 63, 61– 72.




Bijker W., Ardila J.P., Tolpekin V.A. 2012. Change de-tection and uncertainty in fuzzy tree crown objects in an urban environment. In: GEOBIA 2010 pro-ceedings, 29 June–2 July 2010, Ghent, Belgium, IAPRS, ISPPRS, Vol. 38– 4/C7.

Bucha t., Barka I., Bartko M. 2009. Zhodnotenie zdra-votného stavu lužných lesov v inundacnej oblasti Dunaja v úseku Dobrohošt – Sap z leteckých sní-mok. Lesnícky časopis – Forestry Journal, 55 (2), 145– 164.

Carter G.A. 1993. Responses of leaf spectral reflectance to plant stress. American Journal of Botany, 80, 239−243.

Carter G.A. 1994. Ratios of leaf reflectances in narrow wavebands as indicators of plant stress. Interna-tional Journal of Remote Sensing, 15, 697−703.

Carter G.A., Knapp A.K. 2001. Leaf optical properties in higher plants: linking spectral characteristics to stress and chlorophyll concentration. American Journal of Botany, 88 (4), 677– 684.

Ciesla W.M. 2000. Remote Sensing in Forest Health Protection. USDA Forest Service, FHTET Report No. 00– 03.

Congalton R.G., Green K. 1999. Assessing the accuracy of remotely sensed data: Principles and practices. Lewis Publisher, Boca Raton.

Eigirdas M., Augustaitis A., Mozgeris G. 2013. Predict-ing tree crown defoliation using color-infrared or-thophoto maps. iForest, 6, 23– 29.

Gross C.P. 2000. Remote sensing application for forest health status assessment. 2nd edition, Commission of the European Communities.

Hildebrandt G. 1991. Remote Sensing Application for Forest Health Status Assessment. Commission of the European Communities, Walphot S.A, 1991.

Hirschmugl M., Ofner M., Raggam J., Schardt M. 2007. Single tree detection in very high resolution remote sensing data. Remote Sensing of Environment, 110 (4), 533– 544.

Jackson R.D. 1983. Spectral Indices in n-Space. Remote Sensing of Environment, 13, 409– 421.

Jing L., Hu B., Noland T., Li J. 2012. An individual tree crown delineation method based on multi-scale segmentation of imagery. ISPRS Journal of Photo-grammetry and Remote Sensing, 70, 88– 98.

Klobučar D., Pernar R., Lončarič S., Subašič M., Seletkovič A., Ančič M. 2010. Detecting Forest

Damage in Cir Aerial Photographs Using a Neural Network. Croatian Journal of Forest Engineering, 31, 157– 163.

Masaitis G., Mozgeris G., Augustaitis A. 2013. Spectral reflectance properties of healthy and stressed conif-erous trees. iForest, 6, 30– 36.

Meddens A.J.H., Hicke J.A., Vierling L.A. 2011. Evalu-ating the potential of multispectral imagery to map multiple stages of tree mortality. Remote Sensing of Environment, 115, 1632– 1642.

Morgan J.L., Gergel S.E., Coops N.C. 2010. Aerial pho-tography: A rapidly evolving tool for ecological management. Bioscience, 60, 47– 59.

Olofsson K., Hagner O. 2006. Single tree detection in high resolution satellite images and digital aerial images using artificial neural network and geometric-optical model. Proceeding from international workshop 3D Remote Sensing in Forestry, 14– 15 February 2006, Vienna, Austria, 205– 209.

Pitkänen J. 2001. Individual tree detection in digital aerial images by combining locally adaptive bi-narization and local maxima methods. Canadian Journal of Forest Research, 31, 832– 844.

Pontius J., Martin M., Plourde L., Hallett R. 2008. Ash decline assessment in emerald ash borer-infested regions: A test of tree-level, hyperspectral tech-nologies. Remote Sensing of Environment, 112, 2665– 2676.

Pouliot D.A., King D.J., Bell F.W., Pitt D.G. 2002. Auto-mated tree crown detection and delineation in high-resolution digital camera imagery of coniferous for-est regeneration. Remote Sensing of Environment, 82, 322– 334.

Raši R., Bucha T. 2001. The evaluation of forest stands defoliation in the region of water construction system Gabčíkovo using aerial photos. Lesnícky časopis – Forestry Journal, 47 (2), 165– 177.

Sims D.A., Gamon J.A. 2002. Relationships between leaf pigment content and spectral reflectance across a wide range of species, leaf structures and devel-opmental stages. Remote Sensing of Environment, 81, 337−354.

Slaton M.R., Hunt R.E. Jr, Smith W.K. 2001. Estimating near-infrared leaf reflectance from leaf structural characteristics. American Journal of Botany, 88 (2), 278– 284.


Strasser T., Lang L., Pernkopf L., Paccagnel K. 2012. Object-based class modeling for assessing habitat quality in riparian forests. Proceedings of the 4th GEOBIA, 7– 9 May 2012, Rio de Janeiro, Brasil, 555– 560.

Šmelko Š. 1990. Zisťovanie stavu lesa kombináciou odhadu a merania dendrometrických veličín. VPA, VŠLD Zvolen, 1990/6.

Šumbera S., Židek V. 2002. Automated tree top identi-fication using colour infrared aerial photograps of high spatial resolution. Ecology, 21 (3), 229– 238.

Vogelmann J.E., Rock B.N., Moss D.M. 1993. Red edge spectral measurements from sugar maple leaves. International Journal of Remote Sensing, 14, 1563−1575.

Wang Y., Soh S.Y., Schultz H. 2006. Individual tree crown segmentation in aerial forestry images by mean shift clustering and graph-based cluster

merging. Intrernational Journal of Computer Sci-ence and Network Security, 6 (11), 40– 45.

Wolf B-M., Heipke Ch. 2007. Automatic extraction and delineation of single trees from remote sens-ing data. Machine Vision and Applications, 18 (5), 314– 330.

Zarco-Tejada P.J., Miller J.R., Noland T.L., Mohammed G.H., Sampson P. 2001. Scalling-up and Model In-version Methods with Narrow-band Optical Indices for Chlorophyll Content Estimation in Closed For-est Canopies with Hyperspectral Data. IEEE Trans-actions on Geoscience and Remote Sensing, 39 (7), 1491– 1507.

UN/ECE ICP Forests. 2006. Manual on methods and criteria for harmonized sampling, assessment, monitoring and analysis of the effects of air pollu-tion on forests: Part II Visual Assessment of Crown Condition. UNECE, Geneva (www.icp-forests.org).




ORIGINAL ARTICLE

Received 25 March 2013 / Accepted 6 June 2013

DOI: 10.2478/ffp-2013-0008

Changes in runoff as an indicative measure of water retention status in the Białowie˝a Primeval Forest

Jan Tyszka , Andrzej Stolarek Forest Research Institute, Department of Forest Ecology, Sękocin Stary, Braci Leśnej 3, 05– 090 Raszyn, Poland, phone: 48 22 7150526, fax: 22 7150507, e-mail: [email protected]

AbstrAct

The study concerned changing status of water retention within an unique nature complex of the Białowieża Primeval Forest. The retention was evaluated based on an analysis of research results on water outflow from the lowland catch-ment of the river Łutownia – representative for physiographic conditions as well as those of habitats and tree stands in Poland’s part of the Forest. The catchment in the lowland hydrometric dissection at the Pogorzelce village – Old Białowieża Wilderness, covers the area of 120.1 km2 with natural flow conditions preserved. A unified sequence of measurements carried out since 1966 using unchanged methodology with respect to precipitation, air temperature and forest management status allowed to separate different phases of changes occurring in water resources of this protected forest complex. There were appraised periodical runoff changes and cycles including a decrease in water resources which occurred at the turn of the century. The retention capacity of the catchment was determined based on periodical changes in rainfall-runoff relations and the ongoing process of deteriorating water conditions of forest ecosystem was interpreted.

Key words

runoff, natural forests, conditions of changes in retention status

IntroductIon

Water resources in the summer 6-month period are cru-cial for balanced functioning of forest ecosystems. The magnitude of changes in rainfall-runoff relations indi-rectly indicates catchment retention capacity. During periods of weather oscillations, changes in river outflow volume reflect well the status of water resources, which are generated as a result of the effects of precipitation at the start and then those of evapotranspiration and reten-tion. Rainfall- runoff relations are shaped by specific retention properties of individual forest habitats, which

are the reason why discharge changes are not synchro-nized with the cycle of precipitation incidence. During the period of warmer winters, retention volume is af-fected not only by changes in precipitation distribution but also by increased rainfall. Long-term identification of discharge of the Łutownia river with its catchment representative for Białowieża Forest conditions allows regular evaluation of the status of retention in the sum-mer 6-month period. Periodical relations between run-off in the summer and winter 6-month periods can be used for forecasting water conditions in the vegetation season. For the past decades, escalating changes in for-

Changes in runoff as an indicative measure of water retention status in the Białowieża Primeval Forest 73

est water conditions have been reflected in groundwa-ter level decrease and deterioration of river discharge. There is a need for identification of the current state and trends of changes in water resources as well as insight-ful analyses of the reasons involved in these processes. Unstable hydrothermal conditions manifested as a ten-dency towards decreasing Sielianinow’s index are the main causative factors in the occurrence of disturbanc-es which are threatening nature and proper functioning of forest ecosystems in the Białowieża Primeval For-est (Malzahn and Chomotowska 2009). A substantial role in shaping water resources is played by changing (especially during the periods of weather anomalies) water demands of forest stands, and these depend on forest habitat types (Tyszka 2008). Growing destabili-zation of water conditions impacts sustainability of the whole valuable and protected forest complex, which is designated as Natura 2000 site 2000 04 comprising the Białowieża National Park (UNESCO World Herit-age Site) and 6 nature reserves “Natural Forests of the Białowieża Primeval Forest”. There are also long-es-tablished game areas which now are designated among others for the activities connected with reintroduction of the European bison. For conservation of forest eco-systems’ natural values and appropriate interpretation of changes in biodiversity, it is crucial to identify ter-ritorial water resources of forest habitats, and river run-off is the easiest to evaluate indicative measure of the ongoing changes.

MeterIAl And Methods

The aim of the study carried out was to determine changes of water conditions in the forest river catch-ment with comparatively well preserved natural habitats of the Białowieża Primeval Forest. The issue of water resources is now especially important due to disrupted balance of hydrological relations in this forest complex, exceptional on a world scale. A relatively easy to deter-mine measures of ongoing changes are runoff of catch-ment and its periodical changeability, and these were used as parameters to assess water resources status in the Białowieża Forest. Runoff of catchment is resultant of the process of shaping other water cycle components, i.e. precipitation, evapotranspiration and retention. The catchment of the Łutownia river, situated in the cen-

tral part of the Polish side of the Forest was regarded as representative of its hydrological conditions (Pierzgal-ski et al. 2002a). Unified and continuous data on pre-cipitation, water levels and discharge were at disposal. These have been collected since 1966 in the natural hydrometric profile preserved by the village Pogorzelce (Old Białowieża Wilderness) following the guidelines elaborated by the Institute of Meteorology and Water Management (IMGW) (Pasławski 1973). Based on aver-age daily water levels as well as the results on volume of water flow under changeable conditions (Byczkowski 1999), there were determined periodical curves of flow intensity and then daily discharges (dcm3/s) were calcu-lated. Uninterrupted registration of changes of other pa-rameters of water cycle constituents was carried out only for three previous years, thus conclusive analyses of the hydrological processes would be too early at that stage.

The amount of catchment average rainfall (mm) was determined by means of polygons of equal rainfall meth-odology based on the data from four observation stations located in the Forest (fig. 1). Based on daily averages, there were determined rainfall and runoff for the win-ter 6-month period (November–April) and the summer 6-month period (May–October). There was ascertained the magnitude of periodical changes of half-year and yearly runoff as well as the runoff irregularity coeffi-cient. Taking into account the effect of air temperature there were determined mutual relations between rainfall and runoff in the summer 6-month period (data obtained from IMGW- Białowieża station). The relation between Sielianinow’s hydrothermal index and the volume of losses were determined as the difference between the amount of rainfall (P) and runoff (H). Indicative reten-tion capacity of Białowieża Forest stands (R1) was deter-mined as the difference between runoff in the summer 6-month period (HL) up to the extreme values of runoff coefficient (c = HL/PL) along with the difference between runoff in winter and summer 6-month periods shaped by winter runoff volume (R2). During statistical inter-pretation (regression equations) of the results, there was taken into account so far gained knowledge on forest and hydrological parameters (Tyszka 2008).

The forest complex analyzed is located on the Pol-ish side of the Białowieża Forest, nearby the border with Belorussia, within the catchments of the river Narewka (right side tributary of the river Narew) and the river Leśna (right side tributary of the river Bug) (fig. 1). The



Jan Tyszka, Andrzej Stolarek74

sources of both rivers are located beyond Białowieża Forest boundaries, thus the rivers do not reflect specifics of the forest complex investigated. On the other hand, hydrological conditions of the river Łutownia catch-ment can be presumed as representative (Pierzgalski et al. 2006) for the purpose of this study. The catchment area up to the Pogorzelce cross-section is 120.1 km2,

which is 19% of the area of the Białowieża Forest on the Polish side.

The river Łutownia starts from the Derlicz Peat Bogs, located by the village Nowosady, and its estu-ary to the river Narew is located inside the Białowieża National Park. Łutownia’s tributaries are periodical wa-

Fig. 1. Position of the Łutownia river catchment within the Polish side of the Białowieża Forest


tercourses, and only the rivers: Dobitka and Krynica infrequently dry out.

In the 19th century, the Łutownia was regulated for the purpose of its utilization as a water route. In the six-ties of the 20th century the river was elutriated. Now it has close to natural character with traces of river bed flattening in the downstream part.

A flat and slightly corrugated plain formed on bot-tom moraine structures of Middle-Polish glaciations is a prevailing land type within the Łutownia catchment. Glacial tills often reside on a layer of ice-marginal silts of large thickness. The layer of soil was formed on a considerable area from glacial sand and grit, and within marshy river valleys – from low bogs. Average catchment slope is 2.6 m/1 km at average terrain eleva-tion 166 m a.s.l. and the height difference 28 m.

Characteristics of habitat and tree stand conditions of the catchment area investigated indicate that almost entire area is covered with a mixed stand with a high production potential. Taking into account generally poor habitats of Middle European Plain terrains, there can be assumed that the area investigated was to our advantage in view of research undertaken due to definite dominance of riparian and alder forest sites over coniferous habitats (9.9%). Stand species composition shows the highest per-centage of spruce 35.6% followed by oak – 17.5%, pine – 13.4% and birch – 11.6%. Other admixture decidu-ous tree species cover 21.9% of the area. The process of wetland drying out has been ongoing in the area, and in 1983– 2010 there was observed a decreasing trend in the groundwater level (Pierzgalski et al. 2002b). During the last decades, overall share of wetland habitats decreased to 28.2% (counting 16.6% peat bog habitats) (Majer 2004). Soil moisture changes caused dieback of spruce stands and expansion of the hornbeam. Higher air tem-peratures, air pollution decrease and limitations of tim-ber acquisition have resulted in a considerable increase of stand stock to more than 300 m3 of large timber wood/ha.

results

Water-gauge readings within the hydrometric cross-section Pogorzelce

As a result of unsupported hydrometric profile, the river Łutownia has unstable conditions of flow, thus its meas-urements must be conducted often and regularly. Chang-

es of flow conditions in the river bed are caused not only by overgrowing vegetation or else ice cover, but also by secondary effects of changes in land elevations as well as river bed slopes. Natural character of the river bed and high fertility of river valley habitats cause abundant overgrowth by deepwater and river bank plants. Water impoundment due to vegetation increases in the period May-August and lasts until the incidence of negative temperatures. Other elements of flow disturbance are connected with the effects of water dames constructed by beavers as well as decayed trees fallen down into the river. Every year, the effects of the above result in periodical water impoundments of tens of centimetres. The flow rate curves obtained for the summer period indicate an increase of the water stages – from 15 cm to about 25 cm. The maximum elevations are observed in April, and also as transitory incidents – after rainfalls in summer months. Long-term average monthly water levels indicate a tendency to decrease after April snow melting and to increase in the period starting in August and ending in winter. Starting from the year 1974, there was observed a long-term unidirectional change of the river bed level, which raised due to silting by 25 cm (fig. 2). This was assessed based on river bed probing every time during flow measurements.

1970 1975 1980 1985 1990 1995 2000 2005 2010 Year

Rive

r bed

leve

l (cm

)

50

60

70

80

90

100

Fig. 2. Shallowing trend in the Łutownia river within the Pogorzelce cross-section

During the annual cycle, the level of the river bed indicates small and short-lived fluctuations. Monthly average changeability of river bed levels registered dur-ing the period of observations amounts for a few centi-metres, along with the tendency to shallow during the hydrologic year (fig. 3). The lowest river bed level is usu-ally observed in November being a result of increased




intensity of river erosion owing to dying vegetation. In general low level of the river bed is also observed at the beginning of snow melting period, when ice cover pulls out (February–March) and after the period of river rising (June–August). Silting of the river bed is usually observed during dry months (May-September). In view of that, river bed movements cause a decrease of the amplitude of extreme water levels.

XI XII I II III IV V VI VII VIII IX X

Rive

r bed

leve

l (cm

)

Month

63

64

65

66

67

68

69

70

71

Fig. 3. Average monthly changes of Łutownia’s river bed within the cross-section Pogorzelce

1964–1970 1971–1980 1981–1990 1991–2000 2001–2009 70

100

130

160

190

220

h (c

m)

Years

WhZ

NhL

WhZ – high winter levelWhL – high summer levelShZ – middle winter levelShL – middle summer levelNhL – low summer level

ShL

ShZ

WhL

Fig. 4. Process of decade changes in typical water levels in the Łutownia river

Even though river flow rate decreases in the Po-gorzelce hydrometric cross-section, disturbances of water outflow in the river bed (which started in the year 1990) caused the increase of low and middle river water levels. High water levels were unchanged during the summer 6-month period and indicated an increas-ing trend during the winter 6-month period (fig. 4). As a result, a decrease of the runoff rate did not lead to drying out of habitats along Łutownia’s lower stream (1.4 km section) flowing inside the Białowieża Nation-al Park.

Ruling out any activities on river stream restora-tion within the areas under strict protection of nature resources obstructs conducting hydrometric measure-ments. However, thank to the implementation of proven methods of interpretation of runoff and comparisons of the results obtained with those reported on runoff of the catchment in the area investigated as well on sub-catchments of the Łutownia river (fig. 1), there can be assumed that our results were affected only by the mi-nor error. On a regular basis, river discharge volume is the constituent of the water balance which can be as-sessed most exactly, even under the conditions of natu-ral hydrometric cross-sections. This is possible thank to recent development of technical methods of flow meas-urements.

Long-term changes in river discharge

Several studies have been conducted on discharge with-in forest catchments in north-eastern Poland (Kuchars-ka et al. 1984; Olszewski 1986; Ciepielowski et al. 1992; Byczkowski and Mandes 1998; Tyszka 2007). The re-sults of these works confirmed the specific character of runoff and the unquestioned role of forest areas in the water cycle. Based on the example of the Łutownia river we analyzed the volume and timing of runoff from the forest of a primeval type. Evaluation of the tendencies of changes ongoing in retention resources of the catch-ment was carried out with the use of long-term results (since 1959) on precipitation and discharge. Hydrologi-cal data from the period 1959– 1962, i.e. just before elu-triation of the river bed, indicated that at annual rainfall lower than normal by 31.5 mm, overall runoff stayed at the similar level in the following years. Advantageous retention capacity of the catchment under the condi-tions of not disturbed water cycle was confirmed by higher (more than a dozen %) runoff observed in the


summer 6-month period as well as that in the winter 6-month period – in the same way lower. After 1966, runoff relied upon overall increase of air temperature and weather anomalies (Pierzgalski and Tyszka 2005). There is a possibility that after the year 2005, small retention objects introduced in the catchment area af-fected Łutownia’s discharge.

After the year 1983, there was observed a growing rainfall deficiency during the summer 6 -month period which resulted in decreased runoff. Until the year 2010, balance losses in the Łutownia catchment increased by approximately 3.1 mm/year. This resulted in lowering groundwater level with an extent depending on habitat wetness (Pierzgalski et al. 2002b). Average hydrother-

Tab. 1. Average hydrothermal parameters of the Łutownia river catchment in 1966– 2011

Period Rainfall P (mm)

Runoff H (mm)

P–H (mm) c = H/P

Air temperature

T (oC)

Sielianinow’s index

kL = P/0,1nt

Duration (days)

vegetation period snow cover

Nov–Apr 235.2 69.0 166.2 0.297 –0.2 29.12– 27.02May–Oct 395.2 37.4 357.9 0.093 13.7 1.58 7.05– 24.09Nov–Oct 630.4 106.4 524.1 0.169 6.8

66–68 69–72 73–76 77–80 81–84 85–88 89–92 93–96 97–00 01–04 05–08 09–11 0

100

200

300

400

500

600

700

800

mm

m

m

0

100

200

300

400

500

600

Years

66–68 69–72 73–76 77–80 81–84 85–88 89–92 93–96 97–00 01–04 05–08 09–11 Years

P–H P H

Year November–October

Summer 6-month period May–October

Fig. 5. Average values of rainfall (P) , runoff (H) and balance losses (P-H) within the Łutownia river catchment in 4-year periods



Jan Tyszka, Andrzej Stolarek78Ta

b. 2

. Per

iodi

cal h

ydro

logi

c pa

ram

eter

s with

in th

e Łu

tow

nia

catc

hmen

t up

to th

e Po

gorz

elce

cro

ss-s

ectio

n in

196

6– 20

11

Year

sR

ainf

all

P (m

m)

Run

off

H (m

m)

P–H

H/P

Hig

h di

scha

rge

WQ

(m3 /s

)

Mid

dle

disc

harg

e in

hig

h w

ater

leve

l ran

geSW

Q (m

3 /s)

Dis

parit

y in

dex

WQ

/SW

Q

Low

di

scha

rge

NQ

(m3 /s

)

Mid

dle

disc

harg

e in

low

w

ater

leve

l ran

geSN

Q (m

3 /s)

Dis

parit

y in

dex

NQ

/SN

Q1

23

45

67

89

1011

12

1966

– 196

8W

inte

r28

6.2

58.1

228.

10.

203

3.96

1.27

3.12

0.07

50.

175

0.42

9Su

mm

er37

4.7

38.2

336.

50.

102

6.48

0.87

7.45

0.08

10.

150

0.54

0Ye

ar66

0.9

96.3

564.

6

1969

– 197

2W

inte

r23

1.3

68.3

163.

00.

295

4.33

1.17

3.70

0.00

00.

219

0.00

0Su

mm

er40

1.4

47.0

354.

40.

117

4.59

1.25

3.67

0.04

30.

120

0.35

8Ye

ar63

2.7

115.

351

7.4

1973

– 197

6W

inte

r18

3.7

75.0

108.

70.

408

5.43

1.90

2.86

0.07

50.

210

0.35

7Su

mm

er44

8.5

49.6

398.

90.

111

9.63

1.87

5.15

0.03

40.

098

0.34

7Ye

ar63

2.2

124.

650

7.6

1977

– 198

0W

inte

r23

9.6

100.

713

8.9

0.42

07.

072.

083.

400.

056

0.28

60.

196

Sum

mer

423.

670

.835

2.8

0.16

78.

331.

585.

270.

019

0.15

60.

122

Year

663.

217

1.5

491.

7

1981

– 198

4W

inte

r22

6.3

101.

012

5.3

0.44

67.

641.

894.

040.

043

0.30

10.

143

Sum

mer

353.

948

.030

5.9

0.13

66.

491.

334.

880.

025

0.13

10.

191

Year

580.

214

9.0

431.

2

1985

– 198

8W

inte

r21

6.7

47.7

169.

00.

220

5.39

1.55

3.48

0.01

50.

080

0.18

8Su

mm

er38

5.5

30.0

355.

50.

078

4.46

1.09

4.09

0.02

80.

061

0.45

9Ye

ar60

2.2

77.7

524.

5

1989

– 199

2W

inte

r22

0.8

71.1

149.

70.

322

4.03

1.37

2.94

0.04

30.

172

0.25

0Su

mm

er36

7.6

26.2

341.

40.

071

2.62

0.79

3.32

0.01

60.

070

0.22

9Ye

ar58

8.4

97.3

491.

1

1993

– 199

6W

inte

r26

6.4

75.2

191.

20.

282

4.66

1.39

3.35

0.01

90.

166

0.11

4Su

mm

er37

6.0

29.2

346.

80.

078

3.00

0.67

4.48

0.00

90.

084

0.10

7Ye

ar64

2.4

104.

453

8.0

1997

– 200

0W

inte

r24

2.4

62.0

180.

40.

256

3.62

1.32

2.74

0.02

20.

154

0.14

3Su

mm

er36

3.2

23.1

340.

10.

064

2.74

0.59

4.64

0.00

10.

038

0.02

6Ye

ar60

5.6

85.1

520.

5


mal parameters with reference to the period of obser-vation are presented in tab. 1.

Climatic conditions of the Białowieża Primeval Forest indicate characteristics of cool and dry conti-nental climate with average annual temperature 6.8oC and average summer 6-month period temperature 13.7oC (data from Białowieża meteorological station). Average annual rainfall (P) for 46 years of observation was 630.4 mm, and showed advantageous distribution with 63% rainfall in the summer 6-month period. Run-off (H) amounted to 106.4 mm, and runoff irregular-ity coefficient c = H/P was 0.169. Average balance loss P–H per year was 524 mm being higher than average precipitation in dry years.

Long-term observations on runoff allowed iden-tification of certain regularities in the process of changes. Hydrologic characteristics of the catchment in subsequent 4-year periods of cyclical changes are presented in tab. 2.

In the last decade of observations, there was ob-served the increase of discharge irregularity coeffi-cient in the high range of water levels WQ/SWQ as well as discharge in the low range NQ/SNQ. Fig. 5 shows an increasing tendency of the difference be-tween rainfall and runoff since 1981, both in the sum-mer and winter 6-month periods. The results indicate great periodical changes in the process of shaping water resources status.

Reducing effects of climatic and hydrologic drought is crucial for stable development of wetland habitats. The analysis of occurrence of periodical low rainfall in the summer 6-month periods indicates that low rainfall showed once in 3– 6 years and its volume was smaller by 21% when compared to normal rainfall. Except for the year 2000, there was observed a general increase of minimal rainfall in the summer 6-month period (fig. 6).

Opposite tendencies were indicated in the case of minimal runoff in the summer 6-month period (fig. 7). Even though minimal rainfall was relatively high, de-creasing trends were observed for low (NQ) and mid-dle (SNQ) discharges in a range of low water levels until the year 2005. The lowest values of discharge in 4-year periods were observed in the turn of the cen-tury, when water discharge from the Łutownia river almost vanished around the year 2000.

12

34

56

78

910

1112

2001

– 200

4W

inte

r22

1.0

36.6

184.

40.

166

3.05

0.92

3.32

0.03

80.

095

0.40

0Su

mm

er38

7.3

13.5

373.

80.

035

2.16

0.36

6.00

0.00

00.

026

0.00

0Ye

ar60

8.3

50.1

558.

2

2005

– 200

8W

inte

r23

9.3

49.9

189.

40.

209

2.23

0.52

4.29

0.07

00.

133

0.52

6Su

mm

er38

2.8

22.8

360.

00.

060

1.56

0.32

4.88

0.03

90.

062

0.62

9Ye

ar62

2.1

72.7

549.

4

2009

– 201

1W

inte

r24

9.0

82.6

166.

40.

332

3.22

0.63

5.11

0.07

80.

201

0.38

8Su

mm

er47

8.0

49.9

428.

10.

104

2.72

0.38

7.16

0.02

50.

086

0.29

1Ye

ar72

7.0

132.

559

4.5

Max

Win

ter

286.

210

1.0

228.

10.

446

7.64

2.08

5.11

0.07

80.

301

0.52

6Su

mm

er47

8.0

70.8

428.

10.

167

9.63

1.87

7.45

0.08

10.

156

0.62

9Ye

ar72

7.0

171.

559

4.5

Min

Win

ter

183.

736

.610

8.7

0.16

62.

230.

522.

740.

000

0.08

00.

000

Sum

mer

353.

913

.530

5.9

0.03

51.

560.

323.

320.

000

0.02

60.

000

Year

580,

250

,143

1.2




Distribution of average monthly values (SNQ) and minimal (NNQ) discharges (m3/s) from the catchment of the Łutownia river were as follows:

Month SNQ NNQ1 2 3

November 0.15 0.02December 0.14 0.02January 0.17 0.02February 0.20 0.02

1 2 3

March 0.25 0.06April 0.26 0.02May 0.14 0.01June 0.10 0 01July 0.08 0.01August 0.08 0.00September 0.08 0.01October 0.12 0.02

∆PL = –84 mm (21% PLn)

1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011

5 4 36 35 65 5 3200

240

280

320

360

400

440

P L [m

m]

Years

PLn = 400 mm

Fig. 6. Trend and cyclical nature of occurrence of minimum periodical rainfall in the summer 6-month period (PL) in the Łutownia catchment

1966–1968

1969–1972

1973–1976

1977–1980

1981–1984

1985–1988

1989–1992

1993–1996

1997–2000

2001–2004

2005–2008

2009–2011 0,00

0,04

0,08

0,12

0,16

0,20 low water level

middle low water level

m3 /

s

Years

Fig. 7. Periodical discharges from the Łutownia river catchment, characteristic of the summer 6-month period (May-October)


Minimal outflows occurred once in about 3 years in August, once in 4 years in September, once in 5 years in June and July, and once in about 9 years in May.

Catchment retention capacity

Forests have higher retention capacity when compared to that of agricultural lands. While discussing catch-ment retention capacity, there should be taken into ac-count both the volume of precipitation water and dura-tion of its maintenance in the soil profile, in ground and on the land surface including snow cover and vegetation. Good retention capacity of the primeval forest provides relatively stable conditions for nature expansion. In view of the process of retention capac-ity development, important roles are played by soil permeability and moisture as well its suction power (pF) (Król 1979; Maciaszek 1998). According to Ma-ciaszek (1998) water capacity of forest soils ranges from 37.55% for acidic brown soils to 82.8% for bog soils. Retention of water in the forest soil profile of 1 m amounts from 106 to 204 mm. In prevailing Po-land’s mineral and permeable soils water amounts are easy reinstated, however maintained shortly. Reten-tion capacity in forest catchments with mineral soils depends not only on hydro-physical properties of the soil but also on the thickness of the zone of soil aera-tion (Suliński 1989; Tyszka 2008). After precipitation or snow melting, catchment recharge with groundwa-

ter can sometimes last for a long time. Detailed inves-tigation of water retention has to include recognition of all retention types including interception capacity of vegetation. In case of the Łutownia river catchment, we firstly determined overall approximate retention value as a result of the analysis of rainfall and summer runoff in subsequent decades (fig. 8). Relations rain-fall-runoff were disturbed especially in 1970– 2000 decades, i.e. during the period of changing conditions of the discharge process when rainfall-runoff rela-tions were disturbed by durable retention. For exam-ple, large runoff was observed in the 80-ties, during the period of low rainfall (PL) and high air tempera-tures (TL).

Retention capacity (ΔR1) can be evaluated based on runoff changes in the summer 6-month periods in the years with differentiated values of runoff coef-ficient c = HL/PL. Largest retention capacities were observed in the periods when small runoff occurred at high rainfall. At runoff 400 mm (close to normal), the difference between the years with edge values of runoff coeficient, i.e. hydrologically dry or wet, was 56 mm. This value was 10 mm higher when evaluated for separate years (fig. 9).

Summer runoff is influenced not only by current rainfall, but also by the process of evapotranspiration and water recharge from melting snow. Fig.10 illustrates the relation between winter/summer rainfall (PZ/PL) and

PL = 424 mmTL = 13.2°C

PL = 482 mmTL = 14.5°C

PL = 381 mmTL = 14.1°C

PL = 401 mmTL = 14.0°C

PL = 388 mmTL = 14.4°C

1970–1979 1980–1989 1990–1999 2000–2009 2010–2011

H L (m

m)

Decades

0

10

20

30

40

50

60

70

Fig. 8. Runoff in the summer and winter 6-month periods in 2010 and 2011 with reference to hydrothermal conditions




summer/winter runoff (HL/HZ) and shows that winter re-tention reserves are decisive of summer runoff if winter rainfall (PZ) is higher than summer rainfall (PL) of about

0.6 mm. At equal values of rainfall in both 6-month pe-riods, the increase of runoff can reach 20– 30%.

The above relation is best confirmed by the dif-ference between winter and summer runoff (HZ–HL)

∆R1 =

56

mm

extremaly dry years(c < 5%)

wet years(c = 10÷15%)y = 0.1465x – 14.405R2 = 0.83

dry years(c = 5÷10%)y = 0.1095x – 16.005R2 = 0.72

extremely wet years(c > 15%)

0

10

20

30

40

50

60

70

80

250 300 350 400 450 500 550 600 650

H L (m

m)

PL (mm)

Fig. 9. Retention capacity (ΔR1) of forest habitats determined based on the difference of runoff in the years with edge values of runoff coefficient in the summer period c = HL/PL (%)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

PZ/PL

H L/H Z

mean Hz = 94 mm mean Hz = 47 mm

Fig. 10. Summer runoff as the effect of rainfall recharge in the summer and retention during the winter 6-month-period


which increases under the influence of increased winter runoff (Hz) (fig. 11). The difference can be treated as an index value of forecasted runoff in the summer 6-month period, which periodically changes in different phases of water resources fluctuations.

A decrease of winter runoff by 42.8 mm occurred be-tween the beginning (1966– 1982) and the end (2001– 2006) of observations and the difference (∆R2) between winter and summer runoff was decreased by the similar quantity (fig. 11). Periodical character of retention changes was as-sociated with hydrothermal conditions (fig. 12). There can be distinguished the periods with regular dependencies between summer runoff and air temperature:

– warm and wet 1970-ties, when runoff was decreas-ing with rising air temperature,

– cool 1980-ties with surplus water balance and char-acteristic of increasing with temperature runoff,

– summers, with close to normal temperatures rang-ing from 13.2– 14.6oC and summer runoff (HL) 11– 42 mm which was increasing with temperature,

– dry summers after 2000 with very low runoff which was increasing in summers with higher temperatures,

– wet summers after 2000 when runoff volumes and temperatures were similar to those observed at the beginning of investigation. The results indicate that changes in hydrothermal

conditions are of periodical character, and their recogni-tion allows forecasting approximate summer discharge based on known volume of winter discharge.

–40

–20

0

20

40

60

80

100

120

H =

H Z – H

L (m

m)

10 30 50 70 90 110 130 150

HZ (mm)

y = 1.2369x – 44.079R2 = 0.8583

y = 1.1604x – 64.781R2 = 0.9547

mea

n

1966–1982 except for dry years – { }

1983–2000

2001–2006

2007–2011

1969

1976

1966

2007

2008

2009

2011

2010

84.241.4 63.4 68.3

Fig. 11. Periodical changes of runoff in the winter 6-month period (HZ) as index value of catchment retention (ΔR2 = HZ – HL)




Two periods of time: 1983– 2000 and 2001–2006 with similar rainfall and temperatures were chosen for the analyses in terms of finding other than climatic reasons of the decrease of retention resources. The dif-ference between winter and summer runoff in the peri-ods investigated was 17 mm. This can be attributed to increased water needs of forest as a result of growing volume of large timber in stands.

dIscussIon

Determination of the role of forests in the catchment water cycle has become an issue focusing the atten-tion of hydrologists as well as foresters, and has not yet been fully understood. External conditions which shape forest-water relations are so much environmen-tally complicated that actually each forest complex has to be approached separately with taking into ac-count its hydro-climatic conditions as well as features of natural environment. The results of broad research undertaken in a range of geographical conditions, for-est habitats and catchment areas of differentiated physi-

ographic characteristics often concluded with opposite results. Divergent results on hydrological functions of forests were obtained by several researchers with rec-ognized standing and professional experience (Dębski 1951; Lull, Sopper 1965; Bac 1968; Rachmanow 1970; Baumgartner 1971; Calder 2007). The direct changes in climatic conditions for forest functioning have recently become an additional concern in understanding rela-tions between runoff and forest management within the catchment area. The results of attempts towards statisti-cal interpretation of these relations are most often appli-cable only at a local level and are burdened with a broad range of measurement data which have to be constantly registered.

In the present study we attempted to evaluate water retention status based on the changes observed within the area of unique natural forest habitats of the Białowieża Primeval Forest. We used a narrow range of climatic and hydrological parameters however these were registered as unified measurement series during the long-term period. The data obtained served for in-terpretation of the reasons of observed changes as well as evaluation of water retention capacity – treated as the

0

20

40

60

80

12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5

2010

2011

y = 26.514x – 332.94R2 = 0.69

y = 18.115x – 228.77R2 = 0.58

tL (°C)

H L (m

m)

dry years after 2000 (P = 363; t = 14.6) wet years after 2000 (P = 482; t = 14.6) normal years (P = 370; t = 14.1) wet years 70 (P > 400; t = 13.3)

normal years (P = 390; t = 13.7) wet years 80 (P > 400; t = 14.7)

Fig. 12. Effects of temperature on runoff in the summer 6-month period in differentiated hydrothermal conditions


main feature of hydro-physical properties specific for forest ecosystems. The study was carried out with the aim to determine extreme changes in water resources, the knowledge of which can be used in the protection of nature valuable ecosystems. There were taken into account hydrological processes resulting from weath-er fluctuations and directional climatic trends, which are important in view of catchment runoff treated as a measure of territorial changes in water resources.

The results showed that the observed trend of de-creasing water resources was accompanied by the in-crease of air temperature growing by 0.2oC every 10 years since the 1970-ties. Ongoing changes are indi-cated by among others decreasing Sielianinow’s hydro-thermal index, the value of which was about 1.8 in the 1970-ties and 1.45 in the years 2000– 2009. The latter is associated with storm rainfalls and long-term droughts (Malzahn and Chomutowska 2009). The decrease of hy-drothermal index (measured by the difference between rainfall and water runoff) results in water balance loss-es in the summer 6-month period (fig. 13) (Malzahn et al. 2012).

y = 215.66x + 43.63R2 = 0.8058

y = 217.48x – 3.9661R2 = 0.9637

y = 250.18x – 16.151R2 = 0.9361

2000.5 1.5 2.5 3.02.01.0

250

300

350

400

450

500

550

600

P L – H

L (m

m)

years 1972–1983

years 1984–2000years 2001–2008

k

Fig. 13. Periodical differentiation of water balance losses during the summer 6-month period (PL–HL) within the Łutownia river catchment according as Sielianinow’s hydrothermal index

Besides current precipitation, the process of shap-ing discharge is influenced by the factors associated with water retention in forest habitats and fulfilling wa-ter needs of forest vegetation. The type of habitats and the intensity of tree biomass production affects directly evapotranspiration – the constituent of water balance (Suliński 1993; Kędziora and Ryszkowski 1998). Evap-

otranspiration increases with rising temperature and precipitation. In the Białowieża Primeval Forest, during the research period, rainfall increased by 1.34 mm/year. This is a result of both moisture deficiency and needs for water in increasing tree biomass following the im-provement of habitat fertility, decrease of soil pollution and limitations of logging. The changes of hydrologi-cal parameters of the Łutownia river catchment which is situated in the central part of the Białowieża Primeval Forest indicate that after 1983 there was shown a ten-dency to decrease water resources leading to limitation of possibilities to meet water needs of tree stands. Such phenomenon was also observed in many other forest complexes in Poland (fig. 14), however rainfalls in 2010 impeded its further progress.

290 310 330 350 370 390 410 430 450100

150

200

250

300

350 Z

(m3 /

ha)

periods with precipitationhigher than annual average periods with precipitationlower than annual average

PL – HL (mm)

Fig. 14. Changes of water balance losses (P-H) in the summer 6-month period as a result of growth of stand volume (Z) in lowland forest catchments (Tyszka 2007)

The occurrence of serious hydrological droughts within the area of the Białowieża Primeval Forest is a warning sign signalling that prevention activities should be undertaken towards avoiding degradation of valuable wetland habitats. In view of deficiency in pre-cipitation during the summer 6-month period, we can-not anticipate overall improvement of water conditions in the whole Primeval Forest. Countermeasure activi-ties should be focused on slow changes in moisture of the most sensitive habitats. However, water enrichment in one habitat generally leads to water depletion in the




other one. Inadvertent activities of beavers, especially after the year 1995, are leading to increased water reten-tion in river valleys, and on the other hand – to desta-bilization of water conditions within the areas located downstream. If the trend of dry years is maintained, there will have to be taken into consideration changes in stand species composition towards stands with lower water demands followed by possible extinction of some valuable flora and fauna species.

The results on relations between rainfall and run-off obtained in this study confirmed the view of Dębski (1951) that in Middle European Plain climatic condi-tions, forests play differentiated functions in the catch-ment water cycle depending on wet, dry or normal years. In the Białowieża Primeval Forest normal annual precipitation is 630 mm. Precipitation in the summer 6-month period 350 mm leads to the increase of water use by forest stands, supplementation of water retention in soil and litter as well as water storage inside organ-isms and on the surface of plants. So as to assure stable water supply for forest stands under climatic conditions of the Middle European Plain, the increase of tempera-ture by 1oC should be compensated by 14% increase of precipitation (Tyszka, Żakowicz 1998).

During last decades the above condition has not been fulfilled, thus the increase of balance losses has been observed. Suliński (1989) carried out a 3-year study on the effect of forest stands on catchment water properties within the highland catchment of the river Traczówka comprising pine stands. Runoff of 63% in the summer 6-month period explained ground water changes observed. In the years of observations, regis-tered summer runoff differed by 61 mm which was a re-sult of rainfall changes. Discharge volume was influ-enced by stand age which affected the depth of ground-water retention. Thus, the pine stand regulated retention capacity of the catchment investigated. The intensity of tree transpiration, changing with age, can also have ad-ditional influence on discharge volume (Boczoń 2004).

More radical methods of research on hydrological effects of forests, including evaluation of consequences of forest removal were applied in dissimilar nature con-ditions in the U.S.A. (Swank and Crossley 1988). The results indicated an increase of runoff and maximum flows, followed by deterioration of soil retention proper-ties and changes in evaporation conditions.

According to the results obtained by Okoński and Miler (2010), retention properties of a catchment de-pend to a big extent on tree stand characteristics. These are determined by the effective precipitation, i.e. that part of precipitation which decides about the volume of rising waters. This could be important contribution to SCS-CN method which is used for hydrological fore-casting in agricultural catchments. The overall reten-tion status analyzed every six months can give a basis for the interpretation of differentiated effects on water rising caused by intensive precipitation or snow melt-ing.

All the results on hydrological functions of for-est give insight to the issue of the effects of forests on shaping the water cycle, however due to limited range of representativeness they have not so far been commonly applied in practice. Determination of retention capacity of forest is important both for sustainability of forest ecosystems and enchancement of the role of forests in regulating river discharge.

conclusIons

– Decreased river discharge volumes as well as the disturbed process of shaping the annual runoff cycle indicate deterioration of appropriate water condi-tions for functioning ecosystems of the Białowieża Primeval Forest. The minimal discharges are dis-tinctive in terms of substantial volume decrease and earlier occurrence in the annual cycle.

– The decreasing trend of river runoff within the area of the Białowieża Primeval Forest observed in the years 1983– 2009 was foremost a result of low rain-fall during this period of time. Current decrease of river runoff should be also interpreted as a conse-quence of increased air temperatures, increased wa-ter needs of stands in changing habitat conditions as well as lately established small retention objects.

– Retention capacity of the catchment can be evaluat-ed based on the differences between runoff in winter and summer 6-month periods or else between runoff in dry and wet years. During periodical drought in vegetation months, precipitation deficiency can be substituted to a big extent by after-winter retention resources.


– The maximum water retention capacity of the habi-tats within the Białowieża Primeval Forest amounts for about 100 mm when assessed based on run-off differentiation in 6-month periods (excluding changes in water use by tree stands).

– Runoff from the catchment serves as a good meas-ure of the status and changeability of the conditions needed for proper functioning of forest habitats. Feasibility and precision of measurements allow us-ing the results on the changes observed for evaluat-ing water retention status as well as short-term fore-casting catchment water conditions. Enhancement of water cycle monitoring in forest catchments is in line with UE’s law recommendations.

references

Bac S. 1968. Rola lasu w bilansie wodnym Polski. Folia Forestalia Polonica, Series A Forestry, 14, 5– 65.

Baumgartner A. 1971. Einfluss energetischer Faktoren auf Klima. Produktion und Wasserumsatz in be-waldeten Einzugsgebieten. 15th IUFRO- Congress in Gainsville (Flo), USA.

Boczoń A.2004. Zużycie wody przez dęby. Monografie Komitetu Gospodarki Wodnej PAN, 23, 123.

Byczkowski A. 1999. Hydrologia. Wydawnictwo SGGW, Warszawa.

Byczkowski A., Mandes B. 1998. Influence of catch-ment forested area on selected flows according to the regression analysis of those flows versus flow generation factors, case study for the north-east Poland. International Scientific Conference „Forest and Water” Cracow, 25– 29 May, 107– 114.

Calder J.R. 2007. Forest and water-ensuring forest ben-efits outweigh water costs. Forest Ecology and Management, 251, 110– 120.

Ciepielowski A., Gutry-Korycka M., Tyszka J. 1992. Wykorzystanie cech lasu do opisu matematycznego odpływu ze zlewni. Prace Instytutu Badawczego Leśnictwa „Gospodarka wodna w lasach”, 155– 164.

Dębski K. 1951. Wpływ lasu na stosunki hydrologicz-ne. Wiadomości Służby Hydrologicznej i Meteoro-logicznej, 3/4, 57– 70.

Kędziora A., Ryszkowski L. 1998. Modification of evapotranspiration by different plant patches. Inter-

national Scientific Conference „Forest and Water” Cracow, 25– 29 May, 149– 164

Król H. 1979. Zdolności retencyjne gleb różnych sied-lisk leśnych. Sylwan, 1, 23– 30.

Kucharska K., Tyszka J., Ciepielowski A. 1984. Rola lasów w wyrównaniu odpływu z małych zlewni w północno-wschodniej Polsce. Gospodarka Wod-na, 4, 107– 110.

Lull H.W., Sopper W.E. 1965. Prediction of overage annual and seasonal streamflow of physiographic units in the northeast. International symposium on forest hydrology, Pennsylvania.

Maciaszek W. 1998. Soil as a natural water reservoir. International Scientific Conference „Forest and Water”, Cracow, 25– 29 May, 290– 299.

Majer A. 2004. Zmiany w siedliskach hydrogenicznych w ocenie Biura Urządzania Lasu i Geodezji Leśnej. Seminarium „Zagrożenia leśnych siedlisk hydro-genicznych w Puszczy Białowieskiej”. Białowieża 21.05.2004.

Malzahn E., Chomutowska H. 2009. Ocena zmian czynników abiotycznych w Puszczy Białowieskiej. Zeszyty Problemowe Postępów Nauk Rolniczych, 540, 247– 258.

Malzahn E., Pierzgalski E., Tyszka J., Janek M. 2012. Zmiany warunków klimatycznych i wodnych w śro-dowisku lasów naturalnych Puszczy Białowieskiej. Sprawozdanie z realizacji badań – temat BLP-361, Biblioteka IBL.

Okoński B., Miler A.T. 2010. Adaptacja metody SCS-CN dla obliczania opadu efektywnego w zlewniach leś-nych. Monografie Komitetu Inżynierii Środowiska PAN, vol. 68, tom 1 „Hydrologia w inżynierii i go-spodarce wodnej” (ed.: B. Więzik), 143– 152.

Olszewski J. 1986. Rola ekosystemów leśnych w mody-fikacji klimatu lokalnego Puszczy Białowieskiej. Ossolineum, Wrocław.

Pasławski Z. 1973. Metody hydrometrii rzecznej IMGW. Instrukcje i podręczniki nr 115.

Pierzgalski E., Tyszka J., Boczoń A. 2002a. Zróżnico-wanie odpływu w małych zlewniach rzecznych w rejonie Puszczy Białowieskiej. Prace Instytutu Geografii Akademii Świętokrzyskiej, 7, 133– 145.

Pierzgalski E., Boczoń A., Tyszka J. 2002b. Zmienność opadów i położenia wód gruntowych w Białowie-skim Parku Narodowym. Kosmos, 51 (4), 415– 425.




Pierzgalski E., Tyszka J. 2005. Water outflow during drought years in watersheds with various forest co-ver. Annals of Warsaw Agricultural University, 36, 21– 28.

Pierzgalski E., Tyszka J., Stolarek A. 2006. Zmienność odpływu wody ze zlewni rzeki Łutowni (Puszcza Białowieska) w latach 1966– 2000. Leśne Prace Ba-dawcze, 1, 21– 36.

Rachmanow W. 1970. Zavisimost recznogo stoka ot le-sistosti bassiejnow. Moskwa.

Suliński J. 1989. Szczegółowy bilans wodny zlewni po-toku Traczówka (24 km2) w Puszczy Niepołomic-kiej. Acta Agraria et Silvestria, Series Silvestris, 28, 105– 123.

Suliński J. 1993. Modelowanie bilansu wodnego w wy-mianie miedzy atmosferą, drzewostanem i gruntem przy użyciu kryteriów ekologicznych. Zeszyty Na-ukowe AR w Krakowie, 179, 1– 10.

Swank W.T., Crossley Jr. D.A. 1988. Forest Hydrology and Ecology at Coweeta. Ecological Studies, vol. 66. Springer-Verlag New York. Inc.

Tyszka J., Żakowicz S. 1998. Changes in soil retention and geochemical properties of water in selected for-est ecosystems of Northern Poland. International Scientific Conference „Forest and Water”, Cracow, 25– 29 May, 198– 208.

Tyszka J. 2007. Kształtowanie się zasobów wodnych lasu w zmieniających się warunkach klimatycz-nych. In: Quo vadis, forestry? (ed.: Z. Sierota). Instytut Badawczy Leśnictwa, Sękocin Stary, 440– 448.

Tyszka J. 2008. Hydrologiczne funkcje lasu w małych zlewniach rzecznych. Prace IBL. Rozprawy i Mo-nografie, 10, 1– 215.



ORIGINAL ARTICLE

DOI: 10.2478/ffp-2013-0009

Phenology of development and population characteristics of the small spruce bark beetle Ips amitinus (Eichh.) in the Karkonoski National Park

Andrzej Mazur, Robert Kuźmiński Poznań University of Life Sciences, Faculty of Forestry, Department of Forest Entomology, Wojska Polskiego 71c, 60– 625 Poznań, Poland, phone: +48 61 8487885, e-mail: [email protected]

AbstrAct

In the years 2005– 2006, in the Karkonoski National Park there were conducted observations on infestation of spruce trees by bark beetles (Col., Curculionidae, Scolytinae). Data on bark beetle species composition and frequency of occurrence of individual species were collected. The data on development phenology of the small spruce bark beetle Ips amitinus in the upper subalpine spruce forest sites (1000– 1250 m a.s.l.) indicated higher frequency of occurrence of this species (26.3%) when compared to the European spruce bark beetle Ips typographus. The study included analyses of population size, density, fecundity and mortality of I. amitinus. The rate of development in I. amitinus in the Karkonosze Mts. is similar to that observed in the Alps. The number of laid eggs observed was low and the reproduction success was very high at minimal mortality. Establishment of feeding galleries and egg laying lasted several weeks and 1/3 of feeding galleries were found in the second series of trap trees.

Key words

Ips amitinus, development phenology, population traits, Karkonoski National Park

IntroductIon

The small spruce bark beetle Ips amitinus (Eichhoff 1872) is a species closely related to the European spruce (eight-toothed) bark beetle Ips typographus (Linnaeus 1758) and resembles the latter morphologically and biologically. Both species frequently co-habit same ar-eas in tree stands or on dying trees together with other spruce bark beetle species. In the ecological sense these form a community of cambiophagous insects compet-ing for environmental resources.

While there is available abundant information on I. typographus, one of the most important bark beetle species causing considerable economic losses, data on the small spruce bark beetle are not systematically presented. I. amitinus is described most frequently as a species accompanying I. typographus (Grodzki 1997, 2009). Only scarce studies have been devoted solely to the small spruce bark beetle (Annila and Nuorteva 1977; Stauffer and Zuber 1998; Jurc and Bojović 2004; Økland and Skarpaas 2008; Witrylak 2008; Holuša et al. 2012; Kicińska et al. 2012).

Received 20 May 2013 / Accepted 10 June 2013


Andrzej Mazur, Robert Kuźmiński90

The small spruce bark beetle is found mainly in central Europe, including mountainous and alpine re-gions. In western Europe it reaches Belgium, the Neth-erlands and France.. In the east it is found in Russia (including its northern provinces), while in the south it reaches lowland areas in former Yugoslavia, Italy and Greece. In the 20th century the small spruce bark bee-tle started to expand its range in a northerly direction. In the 1930’s it appeared in Estonia, while in the early 1950’s it was first reported in Finland (Koponen 1975; Annila and Nuorteva 1977; Biermann and Thalenhorst 1977; Grodzki 1998; Økland and Skarpaas 2008). In the mountainous regions, where I. amitinus is typically found more often than in the lowlands, it is observed up to the mountain forest limit and its highest reported range is altitude 2250 m a.s.l. (Nierhaus-Wunderwald and Forster 2004). This species was accidentally intro-duced with wood to the British Isles, Sweden, the USA and New Zealand (Lundberg 1995; Lindelöw 2000; Brockerhoff et al. 2006).

In Poland, the small spruce bark beetle is common-ly found throughout the country, however to date it has not been reported in the Baltic coast, the Pomeranian Lake District and in the Świętokrzyskie Mts. (Bura-kowski et al. 1992).

Observations on the occurrence of spruce bark bee-tles in the mountain forest limit in the Karkonoski Na-tional Park conducted in the years 2001– 2003 (Mazur et al. 2006; Mazur et al. 2008), showed a significant role and abundant occurrence of I. amitinus in bark beetle community colonizing upper alpine spruce forests in the Karkonosze Mts. The dominant role of the small spruce bark beetle was shown on stand edges in the case of stands with artificial spatial structure, aged approx. 70– 140 years. In older stands, characteristic of spatial structure similar to natural, single spruces were infest-ed by communities of bark beetles with a predominant share of I. typographus. The results of the above study indicated a need for further observations of tree infesta-tion by bark beetles and it was attempted to explain the dominant role of the small spruce bark beetle (Mazur et al. 2006; Mazur et al. 2008).

In 2005 and 2006, in continuation of the aforesaid studies, further observations were conducted in the Karkonoski National Park (KNP) aiming at examina-tion of:

– small spruce bark beetle biology and phenology of development,

– small spruce bark beetle population characteristics.


Field observations were conducted in the Karkonoski National Park from the second half of May, 2005 to 8 September, 2005. In 2006, inspections were conduct-ed on 5–7 June and 5– 7 July.

For data collection there were used trap trees (se-ries 1 and 2), produced from trees infested by bark bee-tles (description of the trees used in the analyses, includ-ing diameters and numbers of mother galleries is con-tained in Kuźmiński and Mazur, 2013). Observations on the trap trees were conducted in the western part of the park in the Szrenica, Śnieżne Kotły and Przełęcz pro-tection zones. Inspections carried out in 2006 included the Śnieżka protection range. Bark beetle-infested trees were also identified in the upper subalpine coniferous forest zone throughout the park area, and these were not assigned as trap trees, but used for analyses of bark beetle population traits, i.e. the number of laid eggs, the length and number of mother galleries, as well as devel-opment success and mortality.

Observations were carried out on 22 trap trees (spruce long logs), 8 rollers 2.4 m long and 9 rollers 1.2 m.long. On the logs and bolts there were marked 1-meter long sections and the diameters and numbers of feeding galleries were assessed for individual bark bee-tle species. In addition, mother galleries extending from the mating chamber were counted in selected long logs together with the number of established egg galleries and pupa cells at the ends of larval galleries. Most anal-yses were conducted after long logs and rollers had been debarked. Until the time of debarking a rate of feeding gallery development was examined by removing the bark from trees on the area of approx. 10 × 15 cm.

The method of data collection from trees follows recommendations on studying cambio- and xylopha-gous insects (Starzyk 1987).

Development phenology and population traits were also studied on bark beetle infested trees not used as conventional and unbarked traps (tab. 1).

Phenology of development and population characteristics of the small spruce bark beetle… 91

results

Small spruce bark beetle biology and phenology

In the course of observations on small spruce bark bee-tles there was found that the number of egg galleries in the mother gallery ranges from 9 to 48. Calculated on the basis of these data mean number of egg galleries per one mother gallery was 20.92.

Tab. 2. Percentage shares of small spruce bark beetle feeding galleries with different numbers of mother galleries extending from one mating chamber

Number of mother galleries 2 3 4 5 6 7

Share [%] 7.1 25.0 42.8 7.1 7.1 10.7

In single feeding galleries of small spruce bark beetles there extended 2– 7 mother galleries from the mating chamber. For example, a total of 185 of small spruce bark beetle feeding galleries were recorded in one analyzed tree (no. 10, forest compartment 151a).

The proportions of feeding galleries with specific num-bers of mother galleries are given in table 2. Develop-ment phenology of the small spruce bark beetle based on observations collected in the Karkonoski National Park is presented below (see also tab. 1).

Assessment of population parameters for Ips typographus and Ips amitinus

Population parameters were assessed mainly on the basis of data collected on spruces (standing and fallen) colonized by bark beetles and not used as trap trees. These trees were inventoried at altitudes of 750– 1220 m a.s.l. (tab. 3).

Population size was typically limited to single trees or foci (bark beetle nests), and for I. amitinus it ranged from 0 to 185 feeding galleries per tree (on average 51), whereas for I. typographus – from 0 to 245 (on aver-age = 35.32). The highest level was recorded on the half circumference of a felled spruce (from the Sowia Do-lina area, diameter in the butt end = 70 cm, stem length = 26.3, 803 I. typographus feeding galleries).

Tab. 1. Phenology of development of small spruce bark beetle observed in 2005 and 2006 in the Karkonoski National Park

Date of control Development stageFirst half of MayCool period, typically in mid-May after a relatively sunny, but cool beginning of the month in 2005

No activity, lack of entrance holes

Second half of MayRelatively cool and wet weather in 2005

Infestation and beginning of establishment of bark beetle mother galleries, reaching 3– 4 cm in length Simultaneous emergence of numerous eight-toothed bark beetle and six-toothed spruce bark beetle

1 –15 June Further and more intensive establishment of mother galleries; gallery length: up to 7 cm; appearance of larval galleries of several mm in length

Second half of JuneFirst half of July

Development of larvae in galleriesLarvae complete development by preparing pupa cells

Second half of July

Larvae still feed in feeding galleries, most of them turn into pupa stage and first light-coloured beetles are found. Young, discoloured beetles start secondary feeding under the bark On 2nd series trap trees, there is observed the beginning of feeding gallery establishment by the small spruce bark beetle; six-toothed spruce bark beetle predominates in tree crowns – also in establishing feeding galleries

Mid-August Completion of development; some young beetles leave their galleries, other beetles remain in feeding galleries

First half of September1st series trap trees: beetles still remain in feeding galleries, some remain in feeding galleries for overwintering2nd series trap trees: mother galleries and well-developed larval galleries, no pupa cells




Tab. 3. Results of survey on spruce infestation by European spruce bark beetle (I. typographus) and small spruce bark beetle (I. amitinus) in accordance with altitude in the Karkonoski National Park

Location[compartment] Altitude Species

of bark beetleNumber

of mother galleriesNumber

of exit holesControl area

[cm2]22b 1200 Ips typographus 13 – 22 × 54 = 118865d 1100 – 31 30 × 10 = 30067g 1220 – 33 30 × 20 = 60071i 1100 26 – 20 × 20 = 40071i 1100 13 – 20 × 20 = 40071k 1150 26 – 20 × 20 = 40071k 1150 23 – 20 × 20 = 40071r 1200 1 12 20 × 20 = 40072j 1120 – 16 45 × 20 = 90072j 1120 – 42 45 × 20 = 90082c 1150 – 12 5 × 10 = 5082c 1150 – 7 6 × 8 = 4882c 1150 – 13 6 × 10 = 6083f 1150 2 – 38 × 22 = 83683f 1150 7 – 70 × 20 = 140083f 1150 – 15 36 × 20 = 720

155f 1200 5 6 15 × 10 = 150155f 1200 8 14 15 × 10 = 150155f 1200 9 9 15 × 10 = 150155f 1200 12 – 30 × 50 = 1500133j 750 4 – 30 × 44 = 1320133j 750 13 – 30 × 44 = 1320133j 750 3 – 9 × 5 = 45163b 1100 7 37 25 × 15 = 375 72f 1150 Ips amitinus 14 – 40 × 20 = 80072f 1150 13 – 40 × 20 = 80072f 1150 13 – 40 × 20 = 80071k 1150 13 – 20 × 20 = 40071r 1200 7 – 20 × 20 = 40072j 1120 – 22 45 × 20 = 90072j 1120 – 53 45 × 20 = 90073w 1150 – 6 40 × 10 = 40073w 1150 – 27 20 × 40 = 800139p 1030 24 – 10 × 50 = 500139p 1030 21 – 35 × 17 = 595163l 1200 11 – 10 × 10 = 100163l 1200 8 9 10 × 10 = 100194g 1000 32 – 25 × 100 = 2500


Population density for I. amitinus ranged from 1.75 to 11 galleries per 100 cm2 – on average 2.16 gal-leries per 100 cm2 (tab. 4). Depending on altitude the area covered by one mother gallery ranged from 23.1 to 52.8 cm2. In other words, the greatest density of mother galleries burrowed by small spruce bark beetles was observed at approximate altitude 1200 m a.s.l., while it was the lowest in trees growing at altitudes from 1100 to 1199 m a.s.l. Also, the number of I. amitinus exit holes per unit area was higher in trees growing at altitude 1200 m than that observed in trees growing at lower altitudes (however, the latter conclusions are based on the limited number of observations).

Tab. 4. Mortality and reproduction success of European spruce bark beetle (I. typographus) and small spruce bark beetle (I. amitinus) at different altitudes in the Karkonoski National Park

Species of bark beetle

AltitudeMother gallery/

cm2 Exit holes/cm2

range mean range mean

Ips typographus

750 15.0– 330.0 134.2 – –1100– 1199 15.4– 418.0 40.5 4.2– 56.3 18.91200– 1220 16.7– 400.0 73.7 10.7– 33.3 19.6

Ips ami-tinus

1000– 1099 20.8– 78.1 46.7 – –1100– 1199 30.8– 61.5 52.8 13.4– 66.7 24.6

1200 9.1– 57.1 23.1 Individu-al data 11.1

Population density of I. typographus was great-est in trees growing at altitudes from 1100 to 1199 m a.s.l. (tab. 3 and 4) where there were recorded on aver-age: 1 mother gallery/40.5 cm2 and 1 exit hole/18.9 cm2 of the bark. For comparison, in tree stands located at 750 m a.s.l., there were observed on average 1 mother gallery/134.2 cm2 of the bark, and in trees growing at altitude at least 1200 m – 1 gallery/73.7 cm2 and 1 exit hole/19.6 cm2.

Fecundity: for I. amitinus there were found on aver-age 20.9 laid eggs/mother gallery, and for I. typogra-phus at an altitude 1000 m a.s.l. – 51 eggs/gallery.

Mortality: when determined for I. typographus at altitude 1000 m a.s.l. mortality was 88.5%, whereas at altitude 1200 m a.s.l. it ranged from 81.5% through 92.0% to 100% in different tree sections (tab. 4).

The analyses of small spruce bark beetle mortal-ity under natural conditions were conducted only in 72j,

73w and 163l forest compartments, and not in other sites, because there the observations were conducted on the trap trees – debarked before completion of bark beetle development. The mean number of exit holes observed was 42.25/100 cm2, being close to the theoretical num-ber of beetles developing per a unit area with no effect on mortality (the mean number of galleries amounting to 2.16 times the mean number of eggs laid in a mother gallery of 20.9 gives 45.36 beetles developing per 100 cm2 area). Small spruce bark beetle mortality calculated based on the above data was 0.93%.

dIscussIon

The small spruce bark beetle Ips amitinus is a typical element of bark beetle community and it accompanies the European spruce bark beetle Ips. typographus in spruce stands. Increasingly often, next to Pityogenes chalcographus, it begins to play a role of co-dominant or else dominant species, particularly in mountainous areas (Starzyk et al. 2000; Mazur et al. 2006; Plašil and Cudlín 2006; Økland and Skarpaas 2008; Witrylak 2008).

Development phenology in I amitinus compared to that of I. typographus was investigated in Finland (An-nila and Nuorteva 1977). The results of the studies on spruce bark beetle biology, ecology and phenology of development carried out in Poland in the lower subal-pine forests of the Beskid Sądecki Mts. were presented by Witrylak (2008).

In small spruce bark beetle feeding galleries ob-served in the Karkonoski National Park, the number of mother galleries ranged from 2 to 7. Almost 50% of the galleries were feeding galleries with four ones extend-ing from the mating chamber. The number of mother galleries recorded in a single feeding gallery did not dif-fer from the data available in literature.

I. amitinus females lay eggs in egg galleries bored on mother galleries’ sides. According to Swiss data the number of laid eggs ranges from 30 to 60 (Nierhaus-Wunderwald and Forster 2004). The number of egg gal-leries recorded in feeding galleries of small spruce bark beetle population in the Karkonosze Mts. ranged from 9 to 48, and on average 20.92. Observed values are lower than those reported in subject literature.

I. amitinus embryogenesis typically lasts 2 weeks, while larval development takes 3 to 4 weeks. During




this period of time larvae bore larval galleries termi-nated with oval pupa cells and go through 3 larval stages. Pupae development lasts 2 weeks, and con-sequently full development of the small spruce bark beetle is completed within the period of 8 to 11 weeks. Under alpine conditions at altitude 1500 m a.s.l., de-velopment of this species lasts from the second half of May to early September (Nierhaus-Wunderwald and Forster 2004).

The comparison of the results recorded in the Karkonoski National Park with data reported in litera-ture shows that the course and rate of development of the small spruce bark beetle in the Karkonosze Mts. are fully comparable with the results on this species popu-lation obtained under alpine conditions.

Development phenology of bark beetles, counting I. typographus and I. amitinus, depends on temperature conditions in a given season. Diurnal temperatures in-fluence the development rate of specimens, while the so-called heat sum required for development of speci-mens determines the duration of development (Szujecki 1983; Plašil and Cudlín 2006 and literature sources cit-ed therein). Particularly, spring mass emergence of bark beetle adults after hibernation, establishment of feeding galleries and the course of swarming may be prolonged as a result of low temperatures. Observations reported by Kuźmiński and Mazur (2013) indicate that the small spruce bark beetle, apart from swarming culmination that occurs in the second half of May and the turn of May and June, may colonize trees also in the later pe-riod. On many trap trees, some feeding galleries with poor (delayed) development of mother galleries was ob-served. This may indicate mass incidence of sister gen-eration, which is consistent with the results presented by Witrylak (2008).

The long period of spruce colonization by small spruce bark beetles may be explained in two ways. Firstly, weather conditions may have a significant effect on the length of feeding gallery establishment period. However, this effect may pertain to the first 4 weeks, i.e. the turn of May and June. A large number of feed-ing galleries initiated and observed in the second half of July may indicate establishment of the sister genera-tion (although this is not suggested by the appearance of feeding galleries, particularly the high number of mother galleries), or else establishment of feeding gal-leries by bark beetles overwintering in the larval stage

and pupating only in the spring. As it is indicated by ob-servations from 2005, such trends in small spruce bark beetle population (i.e. abundance and establishment of feeding galleries throughout the season) are much stronger than those in I. typographus.

When comparing development phenology of the small spruce bark beetle described above with develop-ment of I. typgraphus in the Karkonosze Mts. (Mazur et al. 2006) we may their considerable similarity.

The occurrence of the second generation of I. ty-pographus and I. amitinus, particularly under moun-tainous conditions of the Karkonosze, seems impossi-ble due to the duration of development of spruce bark beetle specimens lasting until September. However, manifestation of some feeding galleries with truncated mother galleries and the small number of larval galler-ies observed may indicate potential emergence of sister generations. However, the number of sister generations does not seem high, since it has been reported by dif-ferent authors that considerable numbers of “parental” specimens remain in feeding galleries produced by the first generation extending mother galleries.

conclusIons

– Phenology and development rate of the small spruce bark beetle in the Karkonosze Mts. are analogous with those observed in this species under alpine con-ditions. The mean number of laid eggs found in the Karkonosze Mts. is lower. The reproduction success of the small spruce bark beetle assessed based on the number of exit holes is very high, and conse-quently its mortality is very low. Yet, there should be stressed that mortality results were obtained based on the analyses of individual trees only, thus any generalization may be burdened with error.

– Colonisation of trees and establishment of feed-ing galleries by the small spruce bark beetle in the Karkonosze Mts. is a long process (lasting several weeks). In the second half of June new feeding gal-leries were established on many trees, and more than 1/3 of all feeding galleries of this bark beetle species were observed the second series traps. This indicates advisability of applying also second series of conventional trap trees in forest protection prac-tice.


references

Annila E., Nuorteva M. 1977. Dates of attack and emer-gence of Ips amitinus Eichh. (Col., Scolytidae) in Finland. Ann. Ent. Fenn. 43 (1), 28– 30.

Biermann G., Thalenhorst W. 1977. Zur Kenntnis des kleinen Buchdrukers, Ips amitinus (Eichh.) (Col., Scolytidae). Anzeiger Schädlingskunde und Pflan-zenschutz Umweltschutz, 50 (2), 20– 23.

Brockerhoff E.G., Bain J., Kimberley J.M., Knížek M. 2006. Interception frequency of exotic bark and ambrosia beetles (Coleoptera: Scolytidae) and re-lationship with establishment in New Zealand and worldwide. Canadian Journal of Forest Research, 36, 289– 298.

Burakowski B., Mroczkowski M., Stefańska J. 1992. Chrząszcze Coleoptera Ryjkowcowate prócz ryjkowców – Curculionoidea prócz Curculionidae (Beetles Coleoptera Spruce wood borers except for true weevils – Curculionoidea except for Curcu-lionidae). Katalog Fauny Polski, Part 23, vol. 18, 1– 234.

Grodzki W. 1997. Changes in the occurrence of bark beetles on Norway spruce in a forest decline area in the Sudety Mountains in Poland. In: Proceedings: Integrating cultural tactics into the management of bark beetle and reforestation pests (eds.: J.-C. Grégoire, A.M. Liebhold, F.M. Stephen, K.R. Day, S.M. Salom. USDA, Forest Service General Tech-nical Report, Ne-236, 105– 111.

Grodzki W. 1998. Szkodniki wtórne świerka – kornik drukarz i kornik drukarczyk (Secondary pests of spruce – eight-toothed bark beetle and small spruce bark beetle). Biblioteczka Leśniczego, 95.

Grodzki W. 2009. Entomofauna of dying young spruce Picea abies (L.) Karst. in the area after forest de-cline in the Izerskie Mountains. Folia Forestalia Polonica, Series A Forestry, 51 (2), 161– 170.

Holuša J., Lukášová K., Grodzki W., Kala E., Matoušek P. 2012. Is Ips amitinus (Coleopetra: Curculioni-dae) Abundant in Wide Range of Altitudes? Acta Zoologica Bulgarica, 64 (3), 219– 228.

Jurc M., Bojović S. 2006. Bark beetle outbreaks dur-ing the last decade with special regard to the eight-toothed bark beetle (Ips amitinus Eichh.) outbreak in the Alpine region of Slovenia. In: Biotic damage in forests. Proceedings of the IUFRO (WP7.03.10)

Symposium held in Mátrafüred, Hungary, 12– 16 September 2004 (eds.: G. Csóka, A. Hirka, A. Kotlay).

Kicińska A., Kuźmiński R., Mazur A. 2012. Morfo-logia kornika drukarczyka (Ips amitinus Eichh.) i zmienność osobnicza populacji w Karkonoszach (Morphology of small spruce bark beetle (Ips amitinus Eichh.) and individual variation in the Karkonosze populations). In: Ochrona lasu – wy-brane problemy historyczne i współczesne (Forest protection – selected historical and contemporary problems). Publication dedicated to Prof. dr. hab. Jacek Michalski on his birthday. Wydawnictwo Uniwersytetu Przyrodniczego w Poznaniu, Po-znań, 95– 120.

Koponen M. 1975. Distribution of Ips amitinus Eichh. (Coleoptera, Scolytidae) in Finland in 1950– 1973. Ann. Ent. Fenn., 41, 65– 69.

Kuźmiński R., Mazur A. 2013. Zasiedlanie drzew przez zespół korników (Col. Curculionidae, Scolytinae) w górnoreglowych borach świerkowych w Karko-noskim Parku Narodowym (Infestation of trees by bark beetle community (Col. Curculionidae, Scoly-tinae) in upper subalpine spruce forests in the Kar-konoski National Park). In preparation.

Lindelöw Å. 2000. Bark- and wood-living insects in timber imported to Sweden – aspects on the risks for establishments of new species. Aktuelt fra skog-forskningen, 4, 5– 10.

Lundberg S. 1995. Catalogus Coleopterororum Suecie. Naturhistoriska riksmuseet Stockholm.

Mazur A., Łabędzki A., Raj A. 2006. Characteristics of forest stands in the uppermost forest boundary zone of the Karkonoski National Park. In: Insect outbreaks in managed and unmanaged forests (ed.: A. Kolk). Forest Research Institute, Warszawa, 47– 58.

Mazur A., Łabędzki A., Raj A. 2008. Obserwacje nad występowaniem kornika drukarza Ips typogra-phus (L.) i gatunków towarzyszących w ekosyste-mach leśnych Karkonoskiego Parku Narodowego w latach 2001– 2003 (Observations on occurrence of eight-toothed bark beetles Ips typographus (L.) and accompanying species in forest ecosystems of the Karkonoski National Park in the years 2001– 2003). In: Monitoring ekosystemów leśnych w Karkonoskim Parku Narodowym (Monitoring of




forest ecosystems in the Karkonoski National Park) (eds.: A. Mazur, A. Raj, R. Knapik). Karkonoski Park Narodowy, Jelenia Góra, 91– 127.

Nierhaus-Wunderwald D., Forster B. 2004. Zur Biolo-gie der Buchdruckerarten. Merkblatt fur die Praxis, 18 (3), 1– 8.

Økland B., Skarpaas O. 2008. Draft pest risk assess-ment report on the small spruce bark beetle, Ips amitinus. Commissioned report from Norwegian Forest and Landscape Institute, 10/2008.

Plašil P., Cudlín P. 2006. Management des Borkenkäfers Ips typographus L. im Nationalen Naturschutzgebi-et Praděd in der Tschechischen Republik. Mitteilun-gen der Deutschen Gesellschaft für Allgemeine und Angewandte Entomologie, 15, 167– 172.

Starzyk J.R. 1987. Methods of quantitative studies on cambio- and xylophagous insects. In: IV Sympo-sium on the protection of forest ecosystems. Wyd. SGGW-AR, Warszawa.

Starzyk J.R., Graboń K., Hałdaś E. 2000. Cambio-and xylophagous insects in spruce (Picea abies (L.) Karst.) stands of the Upper San River Valley in the Bieszczady Mountains National Park (East-ern Carpathians). Zeszyty Naukowe AR Kraków, Leśnictwo, 29, 57– 72.

Stauffer C., Zuber M. 1998. Ips amitinus var. montana (Coleoptera, Scolytidae) is synonymous to Ips ami-tinus: a morphological, behavioural and genetic re-examination. Biochemical Systematics and Ecol-ogy, 26, 171– 183.

Szujecki A. 1983. Ekologia owadów leśnych. PWN, Warszawa.

Witrylak M. 2008. Studies of the biology, ecology, phe-nology, and economic importance of Ips amitinus (Eichh.) (Col., Scolytidae) in experimental forests of Krynica (Beskid Sądecki, southern Poland). Acta Scientiarum Polonorum Silvarum Colendarum Ra-tio et Industria Lignaria, 7 (1), 75– 92.



ORIGINAL ARTICLE

DOI: 10.2478/ffp-2013-0010

Growth of Scots pine (Pinus sylvestris L.) on forest and former agricultural lands in Krynki Forest District

Jacek Zakrzewski1 , Katarzyna Leosz2, Agata Jędrzejuk3

1 Warsaw University of Life Sciences, Department of Forest Botany, Nowoursynowska 159, 02-776 Warsaw, Poland, phone: +48 22 5938028, e-mail: [email protected]

2 Krynki Forest District, Poczopek, 16-113 Szudziałowo, Poland3 Warsaw University of Life Sciences, Faculty of Horticulture and Landscape Architecture,

Department of Ornamental Plants, Nowoursynowska 166, 02-787 Warsaw, Poland

AbstrAct

The paper shows differences in morphological and anatomical features of the Scots pine (Pinus sylvestris L.) grow-ing on former agricultural and forest lands. It was found that at the same age and in the same climatic conditions Scots pines from former agricultural land had larger stem dimensions and needle sizes as compared with the trees growing on forest land. These results lead to an interesting conclusion connected with future afforestation and refor-estiation in the Krynki Forest District.

Key words

Scots pine, height and radial growth, needle parameters, Krynki Forest District, forest and former agricultural lands

IntroductIon

The Krynki Forest District, where the study described below took place, is one of the districts with the biggest share of forest stands planted on former agricultural lands (so far on over 3 000 ha in this forest district), and at the same time one of few where forest stands of the Knyszyńska primeval forest with the so-called Supraśl pine enduring. The district contains most beautiful and imposing Scots pine specimens in Poland [Białobok, Boratyński 1993]. Coexistence of forest stands of such extremely different origins presents a unique oppor-tunity to compare growth of trees belonging to both populations. The current study is preliminary and can

be useful for planning the proper selection of planting material for further forestation of former agricultural lands as well as assessing perspectives of growing high quality forest stands in the future, whose value may match that of the primeval pine forest.


Two forest stands were selected for the study:1. On former agricultural lands in the Ostrów forest

range, compartment 88g, aged 45, on fresh mixed coniferous forest (BMśw). Characteristics: lowland flat terrain, mossy-bilberry soil cover, shrub layer

Received 20 May 2013 / Accepted 15 June 2013


Jacek Zakrzewski, Katarzyna Leosz, Agata Jędrzejuk98

– buckthorn, rowan, juniper on 60%, stand struc-ture – one storied pine stand, stocking – 0.7; medi-um density, moderate crown closure, pine diameter at breast height 18 cm, pine height 19 m, site index 1A, stand quality 1.2, average stock 355 m3/ha.

2. On forest lands (primeval) in the Sosnowik forestrange, compartment 456c, aged 43, on fresh conifer-ous forest (Bśw –Św So). Characteristics: lowland flat terrain, mossy soil cover, shrub layer – spruce, birch on 60%, stand structure – one storied pine stand, under wood – 8 spruce, 1oak; stocking – 0.8; medium density, moderate crown closure, pine di-ameter at breast height 17 cm, pine height 19 m, site index 1A, stand quality 1.2, average stock 267 m3/ha.

Diameter at breast height (DBH) and total height of fifty randomly selected trees from each stand were measured. From 10 trees on each area, samples of wood were taken from above the root collar with the use of Pressler drill. Dendrochronological calculations were carried out on them. Measurements of width of annual ring growth were taken with the accuracy 0.01 mm, us-ing electronic device by Biotronik (model BEPD4) con-nected to a computer with „Growth” software. From five trees on each area, samples were taken at DBH level (1.3 m), using a hammer driven cylindrical punch. The samples contained at least cambium area and the last an-nual growth of wood. They were preserved in 70% etha-nol, then cross-sections were prepared and placed on the

15 16 17 18 19 20 21 22 23 24 25 26 27 28

ForestAgricultural

14

12

10

8

6

4

2

0

Breast height diameter (cm)

Num

ber o

f tre

es (m

)

Fig. 1. Frequency of width at DBH level (1,3 m) in pine trees from forest and former agricultural lands in Krynki Forest District

19.0 19.5 20.0 20.5 21.0 21.5 22.0 22.5

ForestAgricultural

25

20

15

10

5

0

Height of trees (m)

Num

ber o

f tre

es (m

)

Fig. 2. Frequency of height of pine trees from forest and former agricultural lands in Krynki Forest District

Growth of Scots pine (Pinus sylvestris L.) on forest and former agricultural lands in Krynki Forest District 99

microscopic glass in glycerine. The prepared cuttings were examined and measured under the microscope (microscope Olympus BX 61 with digital camera DP70).

One tree was cut per each investigated area and samples of needles were taken: 100 needles from upper, middle and lower parts of canopy. Later length (accuracy 1 mm) and weight (accuracy 1 g) of raw needles were measured. The collected needles were also used to make cross-sections at half length. The cuttings were placed on the microscopic glass in glycerine. Needle width and length and the number of resin canals were assessed.

results

The measurements showed (fig. 1) that mean DBH in forest stands on former agricultural land was 24.3 cm, whereas on forest land – 20.4 cm. Calculated standard error allows to claim that the existing difference in mean DBH is negligible. Height measurements showed that, on former agricultural land, the lowest tree was 19 m and the heighest 22.5 m (fig. 2) As much as 38% of the specimens examined were 21 m high. On forest land, the highest tree was 21.5 m and the lowest – 19 m. Almost half of the trees were 20 m high, and only 20% went above this height. The average height of trees on former agricultural land was 20.86 m, and on forest land – 19.89 m. Calculated standard error doesn’t indicateimportant differences of average height in the exam-ined stands. The study on relationship of the height with DBH (fig. 3) showed that thinner primeval pine trees

were higher than those growing on former agricultural land. However, when DBH was larger, the tendency was reversed, and with the highest diameter on both land types were similar. The study on annual ring growth (fig. 4) showed that the biggest radial growths on both former agricultural and forest lands were recorded be-tween 1965 and 1972. In 1975– 1982, alternating chang-es in growth were indicated in both stands. Since 1983 the growth of trees has stabilized and trees from former agricultural land have achieved higher parametres. Av-erage ring width was 2.83 mm and 2.15 mm for for-mer agricultural – and forest land, respectively, and the percentage of late wood was 44.9% and 41.4%. Radial growths in the last annual ring calculated under the mi-croscope were 503.7 µm on former agricultural land and 462.2 µm on forest land.

22,5

21,5

20,5

19,5

18,5

17,515 16 17 18 19 20 21 22 23 24 25 26 27 28

18,0

19,0

20,0

21,0

22,0

Breast height diameter (cm)

Heig

ht (m

)

ForestAgricultural

Fig. 3. Height with regard to DBH in pine trees from forest and former agricultural lands in Krynki Forest District

7

6

5

4

3

2

1

01964 19681966 19721970 19761974 19801978 19841982 19881986 19921990 19961994 20001998 20042002 20082006

Year

Annu

al ri

ng th

ickn

ess (

mm

)

ForestAgricultural

Fig. 4. Radial growth of stem during the life of pine trees from forest and former agricultural lands in Krynki Forest District



Jacek Zakrzewski, Katarzyna Leosz, Agata Jędrzejuk100

Needles were longer in pine trees from former ag-ricultural land at all levels of canopy with the longest within the top and the shortest in the bottom part. It was found that the length of needles in both land types had close to normal location, and in primeval pine trees over 50% of needles had close to average length.

The weight of needles increased from the bottom to the top, being heavier in trees from former agricultural land, especially at the top of the canopy (respectively 1.758 g and 3.709 g per 100 needles).

On cross-sections at the middle length, the width and height of needles were measured as well as the number of reisin canals. It was found that the measured parametres increased their values towards the top of the canopy with higher ones in trees from former agricul-tural land.

dIscussIon

It was shown that the measured growth parametres at tree age 43– 45 years were higher in pine trees from for-mer agricultural land when compared with those from

primeval forest. On average, trees from former agricul-tural land are higher by 1 m and thicker by 4 cm. How-ever, it must be noted that the widths of primeval pines fit into a broader- and the heights in a narrower range of occurence compared to pines growing on former agri-cultural land. Dendrochronological study showed that primeval pines had notably narrower rings in juvenile wood (formed close to the core area), and this distin-guished them from trees from former agricultural land. In subsequent years pine growth stabilized, although a tendency for bigger growth on former agricultural land was observed. Similarly, average radial growths and the width of last ring pointed out bigger growth in pines from former agricultural land. This could be caused by higher contents of nitrogen and other miner-als in the soil of former agricultural land, which can have direct influence on better growth of trees (Fidler, Hohne 1965; Parsevnikov et al. 1985). The area around the forest stand studied is still farmed, which can also have influence on its fertility.

It cannot, however, be ruled out that the observed differences have genetic background, caused by years of natural selection. Hence, smaller and lighter needles

Tab. 1. Chosen features of pine trees’ needles (Pinus sylvestris L.) growing on forest and former agricultural lands in Krynki Forest District

LandsPart of

crown

Length of needles [cm]Frequency Mean lenght

of needles±SE3– 4 4– 5 5– 6 6– 7 7– 8 8– 9 9– 10

Forestlower

15 80 5 0 4.3 ± 0.41Agricultural 10 47 34 9 0 4.8 ± 0.77Forest

middle1 17 55 26 1 0 5.6 ± 0.73

Agricultural 0 15 26 30 20 9 0 6.3 ± 1.13Forest

upper0 4 71 25 0 0 0 6.7 ± 0.43

Agricultural 0 1 1 49 45 3 1 8.0 ± 0.57

LandsPart of

canopy

Weight of 100 raw needles [g]

Cross section of needles ± SEMean number of resin

canals ± SEWidth[μm]

Thickness [μm]

Forestlower

0.565 1401 ± 85.9 693 ± 63.3 8.1 ± 1.40Agricultural 0.847 1440 ± 84.1 757 ± 40.6 6.8 ± 0.92Forest

middle1.267 1720 ± 197.8 806 ± 103.9 10.6 ± 1.30

Agricultural 1.612 1842 ± 99.1 888 ± 57.7 11.1 ± 0.74Forest

upper1.758 1921 ± 139.4 851 ± 82.6 11.2 ± 1.30

Agricultural 3.709 2286 ± 66.8 1088 ± 44.6 13.6 ± 1.35

Growth of Scots pine (Pinus sylvestris L.) on forest and former agricultural lands in Krynki Forest District 101

can protect pine trees from damage related to weather conditions (e.g. heavy wet snowfall and so on). It can be stipulated that thick wood of primeval pine is more durable, and thus more resistant to pathogenes and mechanical damage (Fabianowski 1961; Adamowski 2008).

AcKnowledgeMent

The authors would like to thank sincerely forest rangers Adam Kiczuk and Piotr Jucha for help in carrying out this research.

references

Adamowski W. 2008. Sosny niebotyczne. Matecznik, 2, 2– 4.

Białobok S., Boratyński A. 1993. Biologia sosny zwy-czajnej. Sorus, Poznań–Kórnik.

Fabianowski J. 1961. Kilka uwag o badaniach dotyczą-cych ras sosny zwyczajnej w Polsce oraz o sośnie mazurskiej. Sylwan, 4, 21– 30.

Fiedler H.J., Höhne H. 1965. Beitrag zur Stickstoffdun-gung mittelalter Kiefernbestande (I). Archiv für Forstwesen, 14 (9), 909– 931.

Parsevnikov A.L., Seryj V.S., Buchvalov JU.M. 1985. Effektivnost azotnych udobrenij v chwojnych le-sach Komi ASSR. Lesnoe Chozjajstvo, 12, 21– 22.




ORIGINAL ARTICLE

Received 5 June 2013 / Accepted 22 June 2013

DOI: 10.2478/ffp-2013-0011

Radiative power of wildfires in Siberia on the basis of TERRA/Modis imagery processing

Evgenii I. Ponomarev Siberian Branch of Russian Academy of Sciences, V.N. Sukachev Institute of Forest, Laboratory of Forest Monitoring, Akademgorodok 50/28, 660036 Krasnoyarsk, Russia, phone: +7 391 2494092, e-mail: [email protected]

AbstrAct

Variety of radiation power of wildfire was investigated by processing TERRA/Modis imagery in 4 µm spectral band. Fire radiative power (FRP) was used for calibrating high-temperature event database obtained by the satellite technique. An analysis was performed on the database of Siberian wildfires for 2010– 2012. Dynamics of FRP was investigated for a number of wildfires including some cases of crown fire. FRP variation was evaluated for various forest zones of Siberia. Classification of wildfires was elaborated in terms of FRP value as a GIS-layer over the ter-ritory of Siberia.

Key words

wildfire, radiative power, FRP, high-temperature event

IntroductIon

One of the main natural disturbances in boreal forests of Russia are wildfires. The annual area of wildfires is 2– 17 million ha in Siberian part of Russia (Conard et al. 2002; Soja et al. 2004; Shvidenko et al. 2011). An increase of wildfire activity and expansion of burned areas have been observed during last years (Loupian et al. 2006; Sukhinin 2008; Shvidenko et al. 2011; Ponomarev 2012). For example in 2012, there were over 23 thousand wild-fires including 2200 large-scale ones within the area over 1500 ha. As a result, Siberian wildfires produced extreme emissions to the atmosphere in 2012 (Panov et al. 2012).

The use of satellite methods and products allowed to expand the area of monitoring, so as to increase the efficiency of wildfire detection, as well as to obtain more attributive information about wildfires, such as coordi-

nates, estimation of active burning area, total damaged areas, temporal characteristics of fires, etc. These are especially important for northern territories of Siberia. The satellite technique is a primary means for wildfire monitoring in this region.

An important feature of multispectral satellite im-agery usage is assessment of radiative energy of active fire. Information obtained is important for the analysis of fire impact on vegetation as well as assessment of emissions or forecast of post-fire conditions and regen-eration dynamics.

In the present study, fire radiative power (FRP) was investigated by using methodology of 4 µm TERRA/Modis imagery processing (Kaufman et al. 1998).

The aim of the study was to analyze FRP variety for wildfires in Siberia and classify wildfire database records in terms of FRP value.

Radiative power of wildfires in Siberia on the basis of TERRA/Modis imagery processing 103

AreA of InvestIgAtIon

Russia is one of the biggest countries with forested ter-ritory, having about two-thirds of 1.2 billion ha of glob-al boreal forests (FIRESCAN, 2013). A significant part of boreal forests is in Siberia. Siberia includes the ter-ritory of Asian part of Russia from the Ural Mountains up to the Far East.

The area of investigation contained a series of forest regions. In the west there is the Western Siberian plain swampy taiga; in the centre there are the Central-Sibe-rian upland taiga regions and the Angara forest region; in the eastern part there are the Trans-Baikal mountain forest regions, the Pre-tundra forests of Eastern Siberia and the Permafrost taiga forest region of Eastern Siberia (see fig. 6 for details). The total area of observation was approximately 980 million ha.


Characteristics of the process of burning in forests are highly variable and depend on a number of parameters such as forest and fuel type, average fuel load, weather conditions, duration of drying period, topography, etc. (Kurbatski and Sheshukov 1978). These factors deter-mine a level of weather fire danger and a level of natural fire hazard (Nesterov 1949).

Fire intensity and burning rate of forest litter deter-mine the type of wildfire. In boreal forests of Siberia up to 95% of fires are surface wildfires of low and moder-ate intensity and about 1– 5% of overall fires are crown ones of high intensity (Kurbatski and Sheshukov 1978; Andreev and Brukhanov 2011; Ivanov et al. 2011).

The intensity of wildfires (low, medium and high) corresponds to power radiative flux from the active burn-ing area. Based on subject literature, fire radiative power (FRP) can have a range 28– 750 kW/m2 per unit area of surface fire and depends on local environmental condi-tions and characteristics of forest fuel (Konev 1977; Va-lendik and Kosov 2008). Wide variety of fire radiative power was evaluated by means of experimental data on fire burning in Russian taiga and also in Canadian jack pine forests. There are estimates of the flux from fire edge in a range of 300– 3200 kW/m for surface fires, in a range of 4900– 16700 kW/m for intermittent/fully-de-veloped crown fires along with extreme values in a range

of 45200– 54000 kW/m for some cases of crown fires (Korovin and Andreev 1988; Stocks and Hartley 1995).

The analysis of wildfire radiation power was car-ried out using surface brightness temperature from 4 µm spectral band of TERRA/Modis imagery. The method and equation were proposed by Kaufman et al. (1998) (Justice et al. 2002):

FRP T T bg= × −( )−4 34 10 1948

48. (1)

where: T4 – brightness temperature of high-temperature

pixel in 21st channel of Modis radiometer (λ = 3.989 – 3.929 µm),

T4bg – the mean brightness temperature of background pixels.

An alternative FRP derivation was developed by Wooster et al. (2003), who normalized FRP over the area of integral flux (Mottram et al., 2005).

FRP S R R

R Tbg= × × −( )

= ⋅

20 25 4 4

4

.

σ(2)

where: R4, R4bg – the radiance of high-temperature pixel and

background pixels correspondingly to the Stefan-Boltzmann law,

σ = 5,6704·10– 8 W/(m2 × K4) is the constant, S – remotely sensed area of active fire.

The unit of FRP measure is W/m2 according to the equation (1). The dimension could be MW/pixel or MW when the equation (2) is used, which determines the inte-gral characteristic per pixel of image. These dimensions are equivalent representation of the same characteristic. The last ones could be converted into W/m2 and W/m in compliance with spatial and linear characteristics of ac-tive fire, respectively. So, in the present study the unit of measure was used according to the equation (2).

For analyses we used own database of satellite de-tected wildfires for 2010– 2012. It includes brightness temperature data as an attribute of daily observation of high-temperature events. Thus, a series of quick-looks was available for every wildfire during lifetime (fig. 1). We converted data into vector polygon layers for treat-ment in the geographic information system (GIS).



Evgenii I. Ponomarev104

Results of daily satellite monitoring allowed getting up to 5– 10 observations of active fire per day. FRP was calculated for every polygon of high-temperature event. Using this information daily and long-term dynamics of FRP was obtained (fig. 2) for different cases of wildfire. In total there were used more than 2.5 × 105 of cali-brated high-temperature events per year.

We performed the conjugate geospatial analysis us-ing GIS vector layers of wildfires calibrated according

to FRP and vegetation maps of Siberia (Isachenko 1988; Bartalev et al. 2003).

FRP database was classified for different forest types: wildfires in pine (Pinus sylvestris) forests (more than 19 000 high-temperature events per year), wild-fires in larch (Larix) forests (30,000 high-temperature events per year), wildfires in Siberian pine (Pinus siber-ica) plantations (over 5000 high-temperature events per year), wildfires in deciduous forests (over 8000 high-

FRP, MW< 25002500–40004000–50005000–10000

Fig. 1. Series of high-temperature events (white polygons) detected by TERRA/Modis over wildfire and the result of FRP classification. Landsat image of active fire is in the background

A B

FRP,

MW

Number of �re observation Number of �re observation

diurnal dynamicslong-term component

100

1,000

10,000

100,000

1 788 15 22 29 36 43 50 57 64 711 5 9 13 17 21 25 29 33 37 41 45 49 53

Fig. 2. FRP dynamics of: A – surface fire of medium intensity, B – extreme values of FRP of crown fire


temperature events per year), and burning non-forest areas (over 7 000 high-temperature events per year). Histogram analysis procedure was performed for each of examined samples. Sporadic extremely high values of FRP or wrong data, (less than 1% of the total) were excluded. As a result, estimates of FRP variety were ob-tained for wildfires in forests dominated by pine, larch, Siberian pine and deciduous stands, as well as for fires at non-forest areas (meadows and agricultural land).

Furthermore, FRPs for wildfires under same en-vironmental conditions were compared. The spatial-temporal filter in GIS was used. Forest fires were se-lected within the local area and during the same time. The area is located on the boundary of Central-Siberian upland taiga regions and the Angara forest region. It is dominated by pine plantations on burozem soil and larch and larch-pine stands on drained sod-podzol soils (Isachenko 1988; Isaev et al. 1994; Bartalev et al. 2003). Spatially wildfires were located within 35 km.

Next, using GIS technology the FRP polygon lay-er was compared with distribution of crown fires that were recorded by the Russian Service of Ground and Air Forest Protection (Avialesookhrana). The compari-son was carried out based on the spatial-temporal fil-ter. So sampling series of calibrated high-temperature events were selected corresponding to Avialesookhrana crown wildfire records. Totally there were considered 73 crown fires that occurred in 2010– 2012 in the An-gara forest region in Siberia.

Finally, FRP ranges were marked out for different Siberian forest zones by means of the geospatial analy-sis in GIS. Thus the radiative heat flux from active fire is determined by zonal differences of local climatic condi-tions and by natural fire danger. Comparisons for differ-ent fire seasons were done as well.

results And dIscussIon

Diurnal variation of FRP (fig. 2) was associated with stages of active fire spreading. The intensity of wild-fire was rising at the noon local time (maximums of the curve) and was reducing at night and on early morning (minimums of the curve).

For wildfires over Siberia’s territory, average FRP value was in range of 1200– 4500 MW (fig. 3). The low-est value corresponded to burning of non-forest terri-

tories. The mean value was 1900 MW for wildfires in pine stands. The mean FRP for fires in larch stands was 3800 MW, which is twice of the flux of fire in pine forests. The average value of FRP was dropping down to 51– 60% of the maximum during night-times. The same FRP reduction was found for the active fire in all types of tree stands, except for larch, where FRP value at night was not less than 80% of average for day-time.

FRP

mea

n, M

W

FRP

max

, MW

0

1000

2000

3000

4000

0

25,000

50,000

75,000

100,000

1 2 3Stands

4 5

maximum FRPmean FRP values day-timemean FRP values night-time

Fig. 3. Sporadic maximum and mean FRP values for the wildfire on territories dominated by: 1 – non-forest (meadows and agricultural lands), 2 – pine stands, 3 – larch stands, 4 – Siberian pine stands, 5 – deciduous stands

Extreme values of radiative power were marked out on the background of long-term and daily dynamics of FRP corresponding to high-intensity fires. Extremely high values of FRP were recorded in forests dominat-ed by pine stands (up to 48000 MW) and larch stands (more than 90000 MW) (fig. 3).

The differences of FRP were mostly the same with-in local area under the same environment condition. FRP values depended on dominating tree stands. Ex-amples of FRP series are presented for four fires operat-ing simultaneously on the site in Central-Siberia taiga in July 2012 (fig. 4, fig. 6, inset).

Weather fire danger index PV-1 (Nesterov 1949) varied from 4600 to 8000 (weather station North-Eni-sejsk) during the time of observation. These values cor-respond to the fourth and fifth (the highest level) classes of weather fire danger. Thus, the intensity of burning as well as fire radiative power were determined mainly by fuel and dominating stands.




FRP data were summarized daily during fire moni-toring. Mean values were calculated for different times of active fire observations on the same day. Maximum values per day were used to plot a graph of fire flux dynamics during the period 20 July – 1 August 2012. So maximums of active fire intensity are presented in fig. 4. It was found that FRP of fires in larch stands (20,000– 27,000 MW) was twice of that for fires in pines (10,000– 12,000 MW). Qualitatively, the result confirms average distribution obtained for Siberia in total (fig. 3). And differences of FRP values in pine and larch stands were stored under high level of fire hazard. Thus remote sensed energy characteristics of active fires could be calibrated for classifying wildfire types in different tree stands.

The FRP maximums were analyzed for series of ac-tive fires in the stage of crown burning. Up to 95% of crown fires corresponded to extremely high FRP values. The latter exceeded average FRP up to 2– 14 times (me-dian = 4.6). Thus, the proposed approach provided for accuracy of crown fires at a level 0.9– 0.95. On the other hand, there are a number of limitations such as: satellite imagery may not be in the phase of wildfire maximum intensity, weather conditions and shooting conditions can introduce errors in accuracy of the brightness tem-

perature and FRP calculating, the size of active burning zone is limiting significantly usage of satellite imagery with a resolution of 1000 m.

We carried out regression analysis of crown wild-fire FRP data versus the area of active burning. Ground-based data of the area were taken into account. The ratio did not depend on the fire season (fig. 5).

FRP,

MW

Crown �re area, ha 0 500 1000 1500 2000

0

10,000

20,000

30,000

40,000

2011

2012

y = 20.36x

Fig. 5. Crown fire size ratio (according to ground-based data) and FRP for selected wildfires in the Angara forest region in 2011 and in 2012

Quantity of high-intensity wildfires was evaluated for the territory of Siberia. Percentage of high- and extreme-intensity wildfires could be used to deter-

A B

Days Days

maximum values per daymaximum values per daymean FRP value per wild�re lifetime

1 3 5 7 9 11 1 3 5 7 9 11

FRP

1000

, MW

0

5

10

15

20

25

30

Fig. 4. FRP dynamics for selected wildfires under same environmental conditions: A – in pine stands, B – in larch and mixed larch-pine stands in Central Siberia taiga, July, 2012


mine and to characterize the intensity of the fire season (tab. 1).

Tab. 1. Annual ratio of high-intensity wildfires in Siberia according to FRP range (%)

YearFRP, MW Characteristic

of fire season intensity

5,000- –20,000

20,000-–35,000 > 35,000

2010 7.1 0.8 0.4 Low2011 18.7 2.9 1.0 Medium2012 29.8 7.7 2.8 Extreme

The difference of FRP ranges appears at the scale of forest regions of Western, Central and Eastern parts of Siberia. Similar ranges of variation of FRP were re-corded for the following regions: Pre-tundra forests and

Permafrost taiga forest region of Eastern Siberia (FRP max ~ 0.9 ± 0.1 × 104 MW, FRP mean ~ 4500 ± 350 MW), Western Siberian plain swampy taiga and Center-Siberian upland taiga (FRP max ~ 4.7 ± 0.4 × 104 MW, FRP mean ~ 1800 ± 200 MW), Angara and the Trans-Baikal mountain forest (FRP max ~ 2.3 ± 0.2 × 104 MW, FRP mean ~ 2100 ± 300 MW). The data for 2012 is pre-sented at the map of forest areas of Siberia as classifica-tion of wildfires in terms of intensity (fig. 6).

The result obtained could constitute the basis for improvement of wildfire intensity forecasting. Data on FRP, weather conditions, forest region and predominant tree species need to be taken into account as well. The proposed approach is needed for quantitative and quali-tative monitoring of wildfire causing emissions into the atmosphere (Sofiev et al., 2008; Panov et al. 2012; Yurova et al. 2013), as well as for assessing the global

SiberiaCity

Forest region boundaryFRP, MW

< 5,0005,000–14,50014,500–22,00022,000–40,000> 40,000

Fig. 6. Classification of 2012 wildfires in Siberia based on a FRP data. Forest regions of Siberia are: I – Western Siberian plain swampy taiga, II – Center-Siberian upland taiga regions, III – Angara forest region, IV – Trans-Baikal mountain forest regions, V – Permafrost taiga forest region of Eastern Siberia, VI – Pre-tundra forests of Eastern Siberia




carbon budget. It is very important to improve reliabil-ity of fire role estimates by using qualitative character-istics of damaged areas and estimating burned biomass (Boschetti and Roy 2009; Kumar et al. 2011). Moreover, this is a new approach to assess vegetation disturbances after fire. Typically, this problem is solved by using the variation of vegetation index before and after the fire. However, remote data on FRP can be the basis for al-ternative estimating along with assessing the type and kind of fire, post-fire damages, and forecasting of veg-etation recovering.

Thus, the use of the 4 μm remote imagery followed by FRP calculation allows to classify forest fires as well as collect additional information on active fire dynamics. Raw data can be provided only by series of high-temper-ature events detecting for each case of forest fire by TER-RA/Modis instrument or by an analogous technique.

Systematic research has not yet been carried out with the use of long-term database of forest fires in Siberia. However, the importance of this direction for the analysis of the fire situation in Siberia is unquestioned (Mottram et al. 2005; Berezin et al. 2013). Having the retrospective database of satellite detected wildfires for the period of 1996– 2013, we expect to implement geospatial analysis for annual data, as well as to calibrate remote sensing measurements by ground-based experiments.

conclusIons

Remote sensing data in infrared range was used to esti-mate the power of the radiative flux from active fires in Siberian forests. The analysis of the FRP dynamics dur-ing the time of the fire lifetime was performed. There were recorded diurnal (up to 40– 49% reducing in the FRP level) and long-term components that characterize changes in burning intensity during fire activity. Fires in larch forests were marked out for their diurnal varia-tion of FRP which was not more than 20%. It was found that against the background of the long-period and daily dynamics of FRP there may be sporadic extreme values corresponding to the case of crown fire ignition. The proposed approach provided for accuracy of crown fires at a level of 0.9– 0.95.

The limits of FRP variation were determined for wildfires under different conditions and in differ-ent forest types. For fires in pine forests, FRP values

were 10000– 12000 MW, and up to 48000 MW in the case of extreme heat radiation power. FRP values were 20000– 27000 MW for wildfires in larch stands, with a sporadic – extremely high – value over 90000 MW. FRP characteristics can be used to determine the type of fire, burning intensity estimate, and also as charac-teristics of fire season intensity.

The classification of wildfires based on FRP values was made for Siberia territory. The result was presented in the format of vector polygon coverage for GIS. Simi-lar ranges of variation of FRP values were recorded for the zones of Pre-tundra forests and Permafrost taiga forest region of Eastern Siberia, the Western Siberian plain swampy taiga and the Center-Siberian upland taiga regions; the Angara forest region and the Trans-Baikal mountain forest region.

AcKnowledgeMents

The author gratefully acknowledge financial support for this research from the National Aeronautics and Space Administration (NASA) and from the Russian Founda-tion for Basic Research.

references

Andreev Yu.A., Brukhanov A.V. 2011. Prevention, monitoring and control of wildfires (for example, the Altai-Sayan Ecoregion): A Reference Guide (in Russian), Krasnoyarsk.

Bartalev S., Belward A., Erchov D.V., Isaev A.S. 2003. A new SPOT4-VEGETATION derived land cover map of Eurasia. International Journal of Remote Sensing, 24 (9), 1977–1982.

Berezin E.V., Konovalov I.B., Ciais P., Broquet G., Wu L., Beekmann M., Hadji-Lazaro J., Clerbaux C., Meinrat O.A., Kaiser J.W., Schulze E.D. 2013. CO2 emissions from wildfires in Siberia: FRP measure-ment based estimates constrained by satellite and ground based observations of co-emitted species. Geophysical Research Abstracts, 15.

Boschetti L., Roy D.P. 2009. Strategies for the fusion of satellite fire radiative power with burned area data for fire radiative energy derivation. Journal of Geo-phisical Research, 114, doi:10.1029/2008JD011645.


Conard S.G., Sukhinin A.I., Stocks B.J., Cahoon D.R., Davidenko E.P., Ivanova G.A. 2002. Determining effects of area burned and fire severity on carbon cycling and emissions in Siberia. Climatic Change, 55, (1/2), 197– 211.

FIRESCAN Science Team. 2013. The Bor Forest Island Fire Experiment. Part 2. In: Prescribed Burning in Russia and Neighbouring Temperate-Boreal Eur-asia (ed.: J.G. Goldammer), 149– 231. Global Fire Monitoring Center (GFMC) / Kessel Publishing House.

Isaev A.S., Maltseva Yu.V., Semenоva V.N. 1994. At-las of the Krasnoyarsk Krai and the Republic of Khakassia (in Russian). Roskartographiya.

Isachenko A.G. 1988. Landscapes of USSR: scale 1 : 4000000 (in Russian). Moscow: GUGK, vol. 96.

Ivanov V.A., Ivanova G.A., Moskalchenko S.A. 2011. Handbook to extinguish wildfires (in Russian). Krasnoyarsk.

Justice C.O., Giglio L., Korontzi S., Owens J., Morisette J.T., Roy D., Descloitres J., Alleaume S., Petitcolin F., Kaufman Y. 2002. The MODIS fire products. Remote Sensing of Environment, 83, 244– 262.

Kaufman Y.J., Justice C.O. 1998. MODIS ATBD: Fire Products // EOS ID#2741.

Kaufman Y.J., Kleidman R.G., King M.D. 1998. SCAR-B fires in the tropics: Properties and remote sens-ing from EOS-MODIS. Journal of Geophisical Re-search, 103, 31955– 31968.

Konev E.V. 1977. Physical basis of vegetation fuel burn-ing (in Russian). Nauka, Novosibirsk.

Korovin G.N., Andreev N.A. 1988. Aircraft forest pro-tection (in Russian). Agropromizdat, Moscow.

Kumar S.S., Roy D.P., Boschetti L., Kremens R. 2011. Exploiting the power law distribution properties of satellite fire radiative power retrievals: A method to estimate fire radiative energy and biomass burned from sparse satellite observations. Journal of Geo-phisical Research, 116, doi:10.1029/2011JD015676.

Kurbatski N.P., Sheshukov M.A. 1978. About forest fires in Khabarovski Krai (in Russian). Lesnoe kho-zyaystvo, 4, 79– 83.

Loupian E.A., Mazurov A.A., Flitman E.V., Ershov D.V., Korovin G.N., Novik V.P., Abushenko N.A., Altyntsev D.A., Koshelev V.V., Tashchilin S.A., Ta-tarnikov A.V., Csiszar I., Sukhinin A.I., Ponomarev E.I., Afonin S.V., Belov V.V., Matvienko G.G., Lo-

boda T. 2006. Satellite Monitoring Of Forest Fires in Russia at Federal And Regional Levels. Mitiga-tion and Adaptation Strategies for Global Change, 11 (1), 113– 145.

Mottram G.N., Wooster M.J., Balster H., George C., Gerrard F., Beisley J. 2005. The use of MODIS-de-rived fire radiative power to characterize Siberian boreal forest fires. In: Proc. of the 31st international symposium on remote sensing of environment, St. Petersburg, Russia.

Nesterov V.G. 1949. Forest fire danger and methods of assessment (in Russian). Moscow, Leningrad.

Panov A., Chi X., Winderlich J., Birmili W., Lavrič J., Ponomarev E., Andreae M. 2012. Assessment of biomass burning emissions to the atmosphere from the Zotino Tall Tower Observatory (ZOTTO) in Central Siberia. Int. workshop on Impact of Climate Change on Forest and Agricultural Ecosystems and Adaptation Strategies. Siberian State University, Krasnoyarsk.

Ponomarev E.I., Valendik E.N., Kisilyakhov Y.K. 2012. Satellite Monitoring of Large Scale Wildfires in Si-beria. Int. workshop on Impact of Climate Change on Forest and Agricultural Ecosystems and Adapta-tion Strategies. Krasnoyarsk, SFU.

Shvidenko A.Z., Shchepashchenko D.G., Vaganov E.A., Sukhinin A.I., Maksyutov Sh. Sh., McCallum I., Lakyda I.P. 2011. Impact of Wildfire in Russia be-tween 1998—2010 on Ecosystems and the Global Carbon Budget. Doklady Earth Sciences, 441, 1678– 1682.

Soja A.J., Sukhinin A.I., Cahoon Jr. D.R., Shugart H.H., Stackhouse P.W. 2004. AVHRR-derived fire fre-quency, distribution and area burned in Siberia. International Journal of Remote Sensing, 25 (10), 1939– 1960, doi:10.1080/01431160310001609725.

Sofiev M., Vankevich R., Lanne M., Koskinen J., Kuk-konen J. 2008. On integration of a Fire Assimila-tion System and a chemical transport model for near-real-time monitoring of the impact of wild-land fires on atmospheric composition and air quality. In: Modelling, Monitoring and Manage-ment of Forest Fire (ed.: J. de la Heras et al.). WIT-press, 342– 351.

Stocks B.J., Hartley G.R. 1995. Fire behavior in three jack pine fuel complexes. Great Lake Forestry Cen-ter, Canada, Sault Ste. Marie, Ontario.




Sukhinin A.I. 2008. Space Monitoring and Analysis of Catastrophic Fires in Central Siberia and Far East. North-East Asia, A la Carte, vol.19, Center For North-East Asian Studies. Tohoku University, 19– 23.

Valendik E.N., Kosov I.V. 2008. Thermal radiation of forest fires and its impact on tree stands (in Rus-sian). Coniferous of Boreal Zone, 25 (/2), 88– 92.

Wooster M.J., Zhukov B., Oertel D. 2003. Fire radiative energy for quantitative study of biomass burning:

derivation from the bird experimental satellite and comparison to modis fire products. Remote Sensing of Environment, 86, 83– 107.

Yurova A.Yu., Paramonova A.V., Konovalov I.B., Kuznetsova I.N., Beekmann M. 2013. Forecast of thermal radiation and aerosol emissions from forest fires in the Central European region (in Russian). Atmosphere and Ocean Optics, 26 (3), 203– 207.

Folia Forestalia Polonica 2013, Vol. 55 (2)

Documents

Transcript of Folia Forestalia Polonica 2013, Vol. 55 (2)