Wheat yield and acreage prediction using LISS-III and AWiFS sensors data of indian remote sensing...

115

Rivista Italiana di Telerilevamento - 2010, 42 (3): 115-127Italian Journal of Remote Sensing - 2010, 42 (3): 115-127

Wheat yield and acreage prediction using LISS-III and AWiFS sensors data of indian remote sensing satellite of

Rupnager district of Punjab, India

Joydeep Mukherjee1, Girmay Gebru1, Anil Sood2, Raj Kumar Mahey3, Santanu Kumar Bal1, Harpreet Singh1 and Prabhjyot Kaur Sidhu1

1Department of Agricultural Meteorology, Punjab Agricultural University, Ludhiana, 141004, India E-mail: [email protected]

2Punjab Remote Sensing Centre, Punjab Agricultural University Campus, Ludhiana, 141004, India3Department of Agronomy, Punjab Agricultural University, Ludhiana, 141004, India

AbstractPrediction of crop yield is important for advanced planning and taking various policy decisions. Conventional techniques of crop monitoring and yield estimation through ground survey and crop cutting experiment is costly and time consuming. The present study was carried out in Rupnagar district of Punjab, India. IRS LISS-III and AWiFS data of February and March 2006 of the district was analyzed through supervised Maximum Likelihood classification approaches. Crop acreage was estimated through district boundary mask approaches. Wheat yield estimated through regression analysis between yield and vegetation spectral indices (RVI, NDVI, DVI, TVI, SAVI and IPVI). The study has shown that, wheat acreage of district is estimated better from LISS-III data (RD = -1.5 percent) compared to AWiFS (Advanced Wide Field Sensor) (RD = 9.41 percent). Among vegetation indices, RVI of both LISS-III and AWiFS had higher correlation coefficient with wheat yield of district (r = 0.89 and 0.91, respectively). Keywords: yield prediction, LISS-III, AWiFS, Punjab, wheat, remote sensing, yield.

Stima delle superfici e delle produzioni di grano con dati acquisiti dai sensori LISS-III e AWiFS del satellite IRS nel Distretto di Rupnager, Stato di Punjab, India

RiassuntoLa stima delle produzioni agricole è importante per supportare le attività di pianificazione e le scelte dei decisori politici. Le tecniche convenzionali di monitoraggio e stima delle produzioni basate su rilievi a terra sono eccessivamente dispendiose sia in termini di tempo sia in termini di costo. Il presente lavoro è stato effettuato nel Distretto di Rupnagar, nello Stato Punjab, in India. Sono stati utilizzati dati rilevati dai sensori LISS-III e AWiFS acquisiti nei mesi di Febbraio e Marzo 2006 dal satellite IRS. I dati sono stati elaborati utilizzando un classificatore supervisionato Maximum Likelihood. La superficie coltivata è stata stimata utilizzando una maschera dei confini del distretto oggetto di studio. La produzione di grano è stata stimata utilizzando analisi di regressione tra la produzione e una serie di indici vegetazionali (RVI, NDVI, DVI, TVI, SAVI e IPVI). I risultati ottenuti mostrano che le stime di superficie prodotte con i dati LISS-III sono migliori (RD = -1,5 %) di quelle prodotte con dati AWiFS (RD = 9,41 %). Tra gli indici di vegetazione utilizzati, l’indice RVI, colcolato sia con i dati LISS-III sia con

Mukherjee et al. Wheat yield and acreage estimation

116

i dati AWiFS, ha presentato il coefficiente di correlazione più elevato con le produzioni di grano presenti nel distretto (rispettivamente r = 0,89 e 0,91).Parole chiave: stima delle produzioni, LISS-III, AWiFS, Punjab, grano, telerilevamento, produzione.

IntroductionWheat is an important staple cereal of the world both from the point of view of gross acreage cultivated and grain production annually. Agriculture is the backbone of the economy of most of the nations in the world by providing livelihood for 60 per cent of the world population [FAO, 2000]. It is cultivated on about 213.6 M ha of land with a total production of 576.31 M tones and average productivity is 26.98 Q/ha [FAO, 2000]. Moreover, it is the dominant grain of world commerce because it is easily transported, stored and used to produce a large variety of foods that include many kinds and types of breads, cakes, noodles, breakfast foods, biscuits, cookies, and confectionary items [Gooding and Davies, 1997]. India is the second largest producer of wheat occupying land area of 26.74 M ha with annual production of 74.25 M tones which makes 12.51 and 12.9 per cent of area and wheat production of the world, respectively [FAO, 2000]. Punjab is a major wheat growing state of India, which contributes 20.2 per cent of the wheat production of the country from just 1.53 per cent of the total geographical area of the country. Wheat is cultivated on an area of about 3.38 M ha with the annual production of about 15 M tones and productivity of 46.96 Q/ha. The state contributes 70 per cent of wheat to the national pool of food grains [Sood et al., 2000]. Although crop production has shown a remarkable increase as a result of the green revolution of 1970s, the year-to-year variability of production remains a major concern for a country inhabited by more than one billion (16 per cent of world population) in just only 329 M ha (2.1 per cent of world land surface). Meeting the ever-increasing food demands - supply balance is one of the most important priority of a nation which needs meticulous planning based on elaborate information on various aspects related to agriculture and the potential crop harvest. In India, crop statistics is collected through ground survey and crop cutting experiments. These techniques are however unable to meet the much desired requirement of pre harvest forecast for advanced policy and decision because of the following major limitations: field surveys are complex, expensive, time consuming, tedious and require a lot of manpower [Verma et al., 2003]; are often subjective, influenced by personal bias, prone to large errors due to incomplete ground observation [Singh et al., 2003] and more importantly, production estimates are available late after the crops are actually harvested [Mahey et al., 1995]. Hence objective, standardized and possibly cheaper and faster methods of crop growth monitoring and early crop yield estimation are imperative. A large number of pre harvest production estimation techniques are currently under use. These ranges from sophisticated crop growth simulation models to simple empirical statistical techniques. Crop simulation models require verified computer (software) programs and a detailed input data, which is not usually readily available. Hence empirical regression models based on weather and spectral observations are most feasible alternatives.With the above points in view, this study was proposed with the following objectives:

117


− To compare IRS LISS-III and AWiFS sensors derived vegetation indices for wheat yield prediction.

− To compare Indian Remote Sensing Linear Imaging Self Scanning III (IRS LISS-III) and Advanced Wide Field Sensors (AWiFS) data for wheat-area estimation.

Materials and methodologyStudy area descriptionThe study is conducted in Rupnagar district of Punjab, India located between 30°32’ - 31°24’ latitude North and 76°18’ - 76°55’ longitude East. The district adjoins Una District (Himachal Pradesh) in the north-east, Hoshiarpur District in the north-west and Ludhiana and Patiala Districts in the south-west. The district comprises seven blocks namely Rupnagar,The district comprises seven blocks namely Rupnagar, Kharar, Chamkaur Sahib, Anandpur Sahib, Morinda, Majri and Nurpur bedi (Fig. 1).

Figure 1 - Location of the Study area, Rupnagar, Punjab, India.

The climate of Rupnagar district is characterized by its hot summer and cold winter. In general, the year may be divided into four seasons: 1) cold winter season (middle of November to February), 2) hot summer season ( May to June), 3) the southwest monsoon season ( July and middle of September) and 4) the post-monsoon or transition season (mid September to middle of November). The soils of the district vary in texture generally from loamy to silty clay loam except along the Sutlej River and chaos where some sandy patches may be found.Agriculture, like rest of Punjab, is the main activity in the district. Main food crops grown include wheat, Paddy and maize covering an area of 87,000, 47,000 and 24,400 ha respectively. Sugarcane and cotton are also grown to a limited extent. The area is characterized by two growing seasons-kharif and rabi. Wheat is grown in rabi season


118

mainly under irrigation (80 per cent).Multi date (February and March 2006) cloud free IRS LISS-III and AWiFS digital data acquired first week of March, 2004 which coincides with maximum vegetative stage of wheat was analyzed. The FCC (False Colour Composite) of AWiFS scene covering the district is presented in Figure 2.The spatial resolution of LISS-III (23.5 m) is higher than spatial resolution of AWiFS (56 m) but the temporal and radiometric resolution of AWiFS is much better than that of LISS-III. Both the sensors have same spectral resolution (four bands from green to middle infrared). Specifications of IRS LISS-III and AWiFS data are given in Table 1.

Table 1 - Specifications of LISS-III and AWiFS Sensors.

Specification AWiFS LISS-III

Spectral Band (μm)

0.52 - 0.59 0.52 - 0.59

0.62 - 0.68 0.62 - 0.68

0.77 - 0.86 0.77 - 0.86

1.55 - 1.77 1.55 - 1.77

Spatial Resolution (m) 56 23.5

Swath width (km) 740 140

Revisit time (days) 5 24

Quantization (bit) 10 8

Image processing and classification Ground Truth (GT) data was collected during fourth week of February to first week of March, which coincides with the maximum growth stage of wheat. Larger homogenous sites of wheat grown under different physiographic conditions were identified using previous season FCC and transferred on 1:50,000 scale base map. These sites were visited and agronomic observations like growth stage of crop, crop density, variety etc were recorded for each site. The geographic locations of important sites in the wheat-growing areas were recorded by using global positioning systems (GPS). The multi date data acquired at maximum growth stage of wheat were classified following standard Maximum Likelihood (MXL) supervised classification algorithm. The major steps involved in acreage estimation are given in Figure 3.IRS LISS-III and AWiFS digital data was registered from Survey of India (SOI) topographic map (Scale 1:250,000) using the geo-referencing facility of ENVI 4.7 (ITT solutions, USA) image processing software. For this large number of GCPs (70 – 80), which well distributed in the entire image were identified on the image and the corresponding SOI topographic maps. Using these GCPs and second order polynomial with nearest neighbor (NN) re-sampling procedure, a geo-referenced image was generated. Then image-to-image registration was performed on AWiFS. Second order polynomials were applied for both map-to-image as well as image-to-image registration. Root mean square error for map to image georefrencing was less than one pixel for LISS-III and for image to image georefererncing it was less than 0.5 pixel for AWiFS.

119


Figure 2 – FCC and corresponding classified images of AWiFS (left) and LISS-III (right) through block boundary mask, Rupnagar district.


120

Figure 3- Flow diagram for supervised image classification and yield estimation.

Maximum Likelihood Classification (MXL) procedure has been used for classification of data within the district, block and sample segments. For this digital boundary mask generated for the district, blocks and segments were superimposed on the LISS-III and AWiFS data and all pixels within their respective boundaries were classified. If the unclassified pixels were greater than five per cent some more training sites within unclassified areas were marked and entire data was reclassified using additional classes. After classification of the images, histogram showing mean, maximum, minimum and

121


standard deviation of digital number values of each band belonging to wheat classes was displayed and the block mean DN values were taken for computation of various vegetation indices using formulas shown in Table 2.

Table 2 - Formulas used for computation of different vegetation indices.

Index Name Formula Reference

RVI Ratio Vegetation Index NIR / Red Jordan [1969]

NDVI Normalized Difference Vegetation Index (NIR - Red) / (NIR + Red) Rouse et al. [1974]

IPVI Infrared Percentage Vegetation Index NIR / (NIR + Red) Crippen [1990]

DVI Difference Vegetation Index NIR - Red Tucker [1979]1979]

TVI Transformed Vegetation Index (NDVI + 0.5)0.5 Deering et al. [1975]

SAVI Soil Adjusted Vegetation Index ((NIR - Red) / (NIR + Red + L)) * (1+L) Huete [1988]

The steps involved in crop acreage estimation using remote sensing data include (a) Study area extraction, b) crop discrimination, c) estimation of area under the crop in the study area and assessment of accuracy of crop identification. Maximum Likelihood (MXL) supervised classification was performed using training statistics of wheat.Wheat yield estimated by using spectral signature compared with the historical yield record of BES. The relative deviation (RD) were computed as a measure of accuracy as follows

RD 100 1Sensor estimate

Sensor estimate BES estimate#= -

6 @

Results and discussionWheat acreage estimation by LISS-III and AWiFS sensorsAnalysis of the digital data has shown that, in both LISS-III and AWiFS, boundary mask approach gave better estimate over sample segment approach. The RD of acreage estimate from BES estimate by using boundary mask approach was -1.51 per cent compared to acreage estimates by using sample segments of LISS-III ranging between -1.98 to -3.70 per cent. Similarly, the RD of acreage estimate from district boundary mask on AWiFS was 9.41 per cent compared to RD of 9.84 to 11.31 per cent by using sample segments of AWiFS.More over, in both LISS-III and AWiFS the precision of acreage estimate was improved as the sample segment size decreases. In LISS-III data 5 km x 5 km sample segment gave RD of -1.98 percent while the estimate was with a RD of -3.70 per cent when 10 km x 10 km segment is used. Similarly from 5 km x 5 km sample segment size of AWiFS, wheat acreage was estimated with a RD of 9.84 per cent while from a segment of 10 km x 10 km it was estimated with a RD of 12 per cent of the BES estimate.Comparing sample segment sizes, it was found that smallest sample segment gave better


122

estimates compared to larger ones. In LISS-III data, the RD of wheat acreage estimate was -3.7, -2.63, and -1.95 per cent for using 10 km x 10 km, 7.5 km x 7.5 km and 5 km x 5 km sizes, respectively. Similarly, RD of acreage estimate was found 11.31, 10.95 and 9.84 per cent for using sample segments sizes 10 km x 10 km, 7.5 km x 7.5 km and 5 km x 5 km of AWiFS. This might be because of better random distribution in the study area. Similar observation was found by Vyas and Kalubarme [1990] in which they reported larger sample segments size with small sampling fraction with few segments resulted in very high coefficient of variation. Dutta et al. [1994] also observed NDVI obtained from near complete enumeration showed consistently better correlation with yield as compared to NDVI derived from sample segments. Area estimated from block boundary mask aggregated to district was also closer to BES estimate compared to sample segment approach and was also comparable to estimates through the district boundary mask. The RD was found to be -1.75 and 10.23 per cent for LISS-III and AWiFS, respectively (Tab. 2).Although the accuracy from sample segment approach was lower, taking 10 per cent sample fraction has certain advantages over compete enumeration through district boundary mask. The volume of the data to be analyze is reduced by 90 per cent and hence reduce the computer time for supervised Maximum Likelihood Classification (MXL) and also the ground truth data requirement (Tab. 3).

Table 3 - Wheat acreage estimation by LISS-III and AWiFS from block boundary mask approach and relative deviation from BES estimation of Rupnagar district.

Sampling ApproachWheat Acreage (ha) Relative Deviation (%)

LISS-III AWiFS LISS-III AWiFS

District boundary mask 85,690 96,100 -1.51 9.41

10 km x 10 km 83,890 98,150 -3.70 11.31

7.5 km x 7.5 km 84,770 97,700 -2.63 10.95

5 km x 5 km 85,330 96,500 -1.98 9.84

Comparison of sensorsThe FCCs and corresponding classified image of LISS-III and AWiFS through district and block boundary mask is shown in Figure 4.The study has shown that wheat acreage of the district was estimated from 83,890 ha to 85,690 ha from LISS-III data while it was estimated from 96,100 ha to 98,150 ha from AWiFS data against the 87,000 ha BES (Bureau of Economics and Statistics) estimate. Over all accuracy remote sensing estimate was found with in -4 to 12 percent of the BES estimate.The estimates of crop acreage from LISS-III was found with a RD (Relative deviation) of -4 per cent from BES estimate while it was within RD of 12 per cent in case of AWiFS. This might be attributed to the higher spatial resolution of LISS-III (23 m) compared to AWiFS (56 m). This agrees with the findings of earlier studies by Potdar et al. [1991], Dadhwal et al. [1991].Another surprising finding is that LISS-III consistently underestimated while AWiFS consistently overestimated the acreage under wheat of the district in all sample segment sizes. The underestimation in case of LISS-III could be due to difficulty to identify unirrigated

123


wheat fields with its lower radiometric resolution (8 bit) as compared to higher radiometric resolution of AWiFS (10 bit). On the other hand, an overestimation of AWiFS might be related to its lower spatial resolution compared to LISS-III which leads to a problem of mixed pixels (i.e. non wheat parcels are included in the AWiFS pixels). A similar result was reported by Yadav et al. [1995].Full district of Rupnagar is covered by 675,063 pixels of AWiFS and 3,833,409 pixels of LISS-III, which means LISS-III require more computer space.The swath width of AWiFS is 740 km compared to 140 km of LISS-III which means one scene of AWiFS covers total geographical area of 547,600 km2, while one scene of LISS-III covers geographical area of 19,600 km2. That is one scene of AWiFS covers an area of as much area as covered by 28 LISS-III scenes. Moreover, AWiFS the higher temporal resolution (5 days) compared to LISS-III (24 days), would provide an opportunity for monitoring the production.One pixel in AWiFS represents an area of 3136 m2 or 0.31 ha while one pixel of LISS-III represents an area of just 525 m2 or 0.0525 ha. It means LISS-III has higher discrimination capability compared to AWiFS i.e. land cover types or parcels smaller than 0.31 ha are averaged out in case of AWiFS while in LISS-III they can be sensed separately.AWiFS has higher radiometric resolution of 10 bit data i.e. it records reflectance from an object in the interval of 210 (1028) compared to of 28 (256) bit in LISS-III. In this regard AWiFS can discriminate objects which have smaller difference in intensity of reflectance than do LISS-III.

The relationship between vegetation indices and wheat yield It has been observed that yield of wheat had a strong correlation with NDVI, DVI, RVI, SAVI, TVI and IPVI. The correlation coefficient varied between 0.81 and 0.91 for vegetation indices derived from AWiFS data while the correlation coefficient was between 0.78 and 0.89 vegetation indices derived from LISS-III data. Among the vegetation indices, RVI had highest correlation coefficient with wheat yield in both from AWIFS data (r = 0.91) and LISS-III data (r = 0.89). The correlation coefficient between the vegetation indices and wheat yield is shown in Table 4.Among all the various vegetation indices, RVI (r = 0.89) had higher correlation coefficient followed by NDVI (r = 0.86) and TVI (0.86), SAVI (r = 0.86), IPVI (r = 0.86) in LISS-III data. DVI had lowest correlation coefficient. Similar relationship was obtained from AWIFS data. Moreover, the correlation coefficients between vegetation indices from AWiFS and wheat yield were higher than correlation coefficients between vegetation indices from LISS-III and wheat yield. Highest wheat yield (4267 kg/ha) recorded in block Morinda was associated with the highest vegetation indices in both LISS-III and AWiFS.

Correlation matrixCorrelation matrix was developed to identify most suitable indices for model development. Indices derived from both sensors correlated with each other in a similar pattern. From the correlation matrix it can be revealed that, all vegetation indices are highly inter-related to each other that choice of indices for the models was made on basis of higher correlation with the wheat yield. As a result RVI, NDVI, DVI and TVI were selected for regression analysis.


124

Figure 4 – FCC (left) and corresponding classified images (right) through district boundary mask approach from AWiFS (top) and LISS-III (bottom) March, 2004 data.

125


Table 4 - Correlation coefficient between vegetation indices from AWiFS and LISS-III and wheat yield.

Vegetation IndicesCorrelation Coefficient

AWiFS LISS-III

RVI 0.91 0.89

NDVI 0.88 0.86

DVI 0.81 0.78

TVI 0.90 0.86

SAVI 0.90 0.86

IPVI 0.90 0.86

Regression analysisEight spectral regression models, four from each sensor, were developed with most correlated vegetation indices (Tab. 5). From the result it can be observed that models developed from RVI, NDVI, TVI and DVI of LISS-III explained yield variability by 89, 86, 86 and 78 per cent, respectively. And similarly, models developed from RVI, NDVI, TVI and DVI of AWiFS explained yield variability by 91, 90, 90 and 81 per cent, respectively. The prediction of grain yield is found to be more accurate when spectral models developed from RVI of both LISS-III and AWiFS were used. RVI was found best indices for estimation of yield. The study showed that grain yield can be explained up to 91 per cent and 89 per cent from using RVI of AWiFS and LISS-III data, respectively. This suggests wheat yield can be predicted from single date data of both AWiFS and LISS-III spectral data.

Table 5 - Regression equations developed using various indices derived from LISS-III and AWiFS data.

Model Correlation Coefficient (r) Standard Error of Estimate (kg/ha)

AWiFS

Y = -47.15 + 15.43RVI 0.91 207

Y = -167 + 296NDVI 0.90 257

Y = -663.22 + 641.90TVI 0.90 225

Y = -136.74 + 0.84DVI 0.81 298

LISS-III

Y = -14.29 + 17.18 RVI 0.89 236

Y = -23.3 + 121NDVI 0.86 224

Y = -200.46 + 237.70TVI 0.86 225

Y = -7.14 + 0.53 DVI 0.78 317


126

ConclusionIRS LISS-III and AWiFS data of February and March 2004 was analyzed following supervised Maximum Likelihood image classification technique. Wheat acreage was estimations through complete enumeration and different sample segment sizes were compared.Different wheat yield models were developed through correlation regression analysis between spectral indices (RVI, NDVI, DVI, TVI, SAVI and IPVI). Wheat acreage was estimated better from LISS-III data than from AWiFS which could be because of the higher spatial resolution of the former sensor resulting in better crop discrimination capability and hence better acreage estimation. Smaller sample segment sizes for same sampling proportion gave better estimate than bigger sample segment sizes. Yield estimation was better from AWiFS data compared to LISS-III, which could be because of the higher radiometric resolution of the former sensor. Among the various vegetation indices RVI had higher correlation with wheat grain yield in Rupnagar district. The prediction equation developed by stepwise regression techniques revealed yield can be estimated successfully up to 90 percent accountability from remote sensing approach.

AcknowledgementsAuthors are thnkful to the Vice Chancellor, PAU and the Director, PRSC, Ludhiana for poviding necessary support.

ReferencesCrippen R. E. (1990) – Calculating the Vegetation Index Faster. Remote Sensing of

Environment 34: 71-73.Dadhwal V. K., Parihar J. S., Madhavy T. T., Ruhal D. S., Jawal S. D., Khera A. P. (1991)

– Wheat acreage estimation for Haryana using satellite. J of Indian Soc of Remote Sensing 19: 1-15.

Deering D. W., Rouse J. W., Haas R. H., Schell J. A. (1975) – Measuring forage production of grazing units from Landsat MSS data. Proceedings of the 10th International Symposium on Remote Sensing of Environment Ann Arnon, Michigan, pp1169-1178.

Dutta S., Sharma S.A., Kherra A.P., Ajai Y. M., Hooda R.S., MathiKumar K. E., Manchandra M.L. (1994) – Accuracy assessment in Cotton acreage estimation using Indian Reemote Sensing data. J Photogramm Eng and Remote Sensing 49: 21-26.

FAO (2000) - Production Year Book. Rome, 54.Gooding, M. J. and Davies, W. P. (1997) – Wheat production and Utilization systems, quality

and the environment. CAB international, Walingford Oxford OX108DE, UK, pp335.Huete, A. R. (1988) – A soil-adjusted vegetation index (SAVI). Remote Sensing of

Environment 25: 53-70.Jordan, C. F. (1969) – Derivation of leaf area index from quality of light on the forest floor.

Ecology 50: 663-666.Potdar, M. B., Sharma, R., Dubey, R. C., Biswas, B. C. (1991) – Relative performance of

Landsat MSS, TM, and IRS LISS-I data for acreage estimation of rabi sorghum. J. of Indian Soc of Remote Sensing 19: 45-58.

Mahey R.K., Sidu S.S., Singh C. (1995) – Spectral Response of Sunflower in relation to growth and yield. Proceeding of national symposium on remote Sensing of environment

127


with special emphasis on green revolution. Nov: 22-24. Rouse J. W., Haas R. H., Deering D. W., Schell J. A., Harlan, J. C. (1974) – Monitoring

the vernal advancement and retrogradation (green wave effect) of natural vegetation. NASA/GSFC Type III Final Report, Greenbelt, MD, p371.

Singh R., Rai A., Misra P. (2003) – Use of GIS for sampling design for agricultural surveys. Http//WWW.GISdevelopment.net.

Sood A., Ray S.S., Patel L.B., Sharma P.K, Sushma P. (2000) – Agricultural scenarios in Punjab with special reference to cropping pattern changes. Scientific Note, RSAM/SAC/CS/SN/01/2000.

Tucker C. J. (1979) – Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment 8: 127-150.

Verma U., Rahul D. S., Yadav M., Khera,A. P., Hooda R. S., Singh C. P., Kalubame M. H., Hooda I. S. (2003). Wheat production forecasting using remote sensing and agroclimatic variables in Haryana. J of Indian Soc of Remote Sensing 31: 141-144.

Vyas S. P., Kulubarme M. H. (1990) – LISSSI data analysis for Sugar cane acreage estimation using different sampling approaches. Remote sensing for Agricultural Application, Proceedings of National Symposium on Remote Sensing for Agricultural Applications, Dec 6-8, 1990, New Delhi.

Yadav M., Hooda R. S., Mothikuner K. E., Ruhal D. S. Khera A. P., Singh C. P., Hood S., Verma U., Dutta S., Kalubarme M. H. (1995) – Cotton acreage estimation in Hissar and Sirsa districts of Haryana using LISS-III digital data. Proceeding of national symposium on remote sensing of environment with special emphasis on green revolution Nov 22-24, 1995.

Received 28/02/2010, accepted 27/09/2010

Wheat yield and acreage prediction using LISS-III and AWiFS sensors data of indian remote sensing...

Documents

Transcript of Wheat yield and acreage prediction using LISS-III and AWiFS sensors data of indian remote sensing...