ORI GIN AL PA PER

Increased frequency of flash floods in Dire Dawa,Ethiopia: Change in rainfall intensity or human impact?

Paolo Billi • Yonas Tadesse Alemu • Rossano Ciampalini

Received: 3 July 2014 / Accepted: 8 December 2014 / Published online: 14 December 2014� Springer Science+Business Media Dordrecht 2014

Abstract In the first decade of the twenty-first century, Ethiopia has been subjected to an

increased frequency of flash floods, especially in the town of Dire Dawa. The results of

international organizations studies point to no evidence of a climate-driven change in the

magnitude/frequency of floods, though increases in runoff and risk of floods in East Africa

are expected. Flash floods are posing constraints to the economic growth and the devel-

opment process of a low-income country such as Ethiopia, and, in order to mitigate such

hazard, it is crucial to understand the relative roles of two main factors: rainfall intensity

and land use change. This study analyses the recent trends of rainfall intensity across

Ethiopia and investigates the relative role of rainfall intensity and land use change in

augmenting the frequency of flash flooding of the town of Dire Dawa by the Dechatu

River. Results indicate that the increase in rainfall intensity is a more important factor than

land use change in controlling the increased frequency of flash flood in Dire Dawa.

Keywords Flash flood � Rainfall intensity � Climate change � Land use change � Ethiopia

1 Introduction

Ethiopia is a country with great geographic diversity. Much of its land consists of a large

plateau at an elevation higher than 2,500 m a.s.l. with high mountains and deep gorges,

P. Billi (&)Dipartimento di Fisica e Scienze della Terra, Universita di Ferrara, Via G. Saragat 1,44122 Ferrara, Italye-mail: bli@unife.it

Y. T. AlemuDepartment of Geography, University of Dire Dawa, P.O. Box 1362, Dire Dawa, Ethiopia

R. CiampaliniSchool of Earth and Ocean Sciences, University of Cardiff, Park Place, Cardiff CF10 3AT, UK

123

Nat Hazards (2015) 76:1373–1394DOI 10.1007/s11069-014-1554-0

river valleys, and lowland plains. The rainy season in Ethiopia is concentrated in the

4 months between mid-June and mid-October when about 80 % of the annual precipitation

is received as torrential downpours. Ethiopia’s topographic and climatic characteristics

have made the country vulnerable to high floods that resulted in destruction, casualties and

damages to economic, livelihoods, infrastructure, services, and health systems.

In Ethiopia, flood disasters and the toll paid in terms of human lives and property

damage show an increasing trend (Fig. 1) (Alemu 2009; Adhikari et al. 2010).

Flash floods are formed from excess rain falling on upstream watersheds, flow down-

stream with massive concentration, high speed, and typically occur suddenly (Lin 1999).

Heavy downpours in mountainous highlands can lead to surges of water that turn dry river

beds or flood plains into raging torrents in minutes. Therefore, the damages of such floods

become particularly pronounced and devastating when they pass across or along human

settlements and infrastructures.

According to IPCC (2012), the second half of the twentieth century is dominated by

inter-annual to inter-decadal rainfall variations and many statistics indicate that trend

estimates are spatially incoherent (Manton et al. 2001; Peterson et al. 2002; Griffiths et al.

2003; Herath and Ratnayake 2004). Moreover, Alexander et al. (2006) highlighted that,

though statistically significant trends toward stronger precipitation extremes were generally

found, the observed changes in precipitation extremes are much less spatially coherent and

statistically significant compared with observed changes in temperature extremes. Recent

studies (IPCC 2012) show rather an increase than a decrease in extreme precipitation, but

there are also wide regional and seasonal variations and trends in many regions that are not

statistically significant.

If for East Africa, this is due to a lack of the literature on changes in heavy precipitation

(IPCC 2012), the same is patent in Ethiopia for the same reason, though Easterling et al.

(2000) and Seleshi and Camberlin (2006) report decreasing trends in heavy precipitation

over parts of Ethiopia during the period 1965–2002.

The literature on the impact of climate change on river floods (including also flash

floods) is scarce, even though the changes in heavy precipitation discussed above may

imply flood changes in some regions (IPCC 2012).

0

20

40

60

80

100

120

140

0

100

200

300

400

500

600

700

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

2003

2006

2010

Affec

red

peop

le (x

103 )

Fata

li�es

Fatali�es Affected people

Fig. 1 Fatalities and people affected by floods in Ethiopia between 1981 and 2010 (Alemu 2009; EM-DAT2014)

1374 Nat Hazards (2015) 76:1373–1394

123

IPCC Technical Paper VI (IPCC 2008) concluded that no evidence, based on instru-

mental records, has been found for a climate-driven globally widespread change in the

magnitude/frequency of floods during the last decades (Rosenzweig and Tubiello 2007),

though increases in runoff and increased risk of flood events in East Africa are expected. Di

Baldassarre et al. (2010) found no evidence that the magnitude of African floods has

increased during the twentieth century, whereas Conway et al. (2009) concluded that

robust identification of hydrologic change was severely constrained by data limitations for

sub-Saharan Africa. The reason of that stands in the limited number of flow gauge data in

space and time and the confounding effects of land use changes and engineering inter-

ventions (Di Baldassarre et al. 2010; IPCC 2012). In fact, while the primary cause of

flooding is abnormally high rainfall, there are many human-induced contributory causes

such as land degradation, deforestation of catchment areas, increased population density

along riverbanks, poor land use planning, zoning and control of flood plain development,

inadequate drainage, particularly in cities, and inadequate management of discharges from

river reservoirs (Mulugeta et al. 2007; Di Baldassarre et al. 2010). In Ethiopia, a research in

Awash River basin by Terefe et al. (2006) indicates that human factor plays a crucial role

in causing frequent flood disasters in upper, middle, and lower Awash.

Moreover, projections of flood changes at river basin scale are scarce in the scientific

literature. These same limitations and uncertainties are presently effective also in Ethiopia.

In the last decade, an increasing occurrence of floods is reported for Africa in general,

sub-Saharan Africa, and several arid and semiarid areas of Ethiopia. In fact, according to

Conway and Schipper (2011), whereas historically floods have never been a major eco-

nomic hazard in Ethiopia, recent years have seen significant socio-economic disruption due

to flooding.

According to Mulugeta et al. (2007), the accuracy and lead times of flood forecasts in

sub-Saharan Africa are limited or questionable, thus, new research and collaborative efforts

are needed to advance flood management in the future.

In developing countries, studies on ephemeral streams flash floods are uncommon and

very few data are available to design appropriate risk mitigation countermeasures and

warning systems. In 2006, the town of Dire Dawa experienced a typical flash flood that,

following the heavy rain on the upland areas of eastern Harerge highlands, within a few

hours turned the dry bed of the Dechatu River into a swelling and devastating river that

caused several casualties and property damage for millions of Euros. The problem of flash

floods in the semiarid area of Dire Dawa is not new, but their frequency has significantly

increased in the last decades (Alemu 2009; DDAEPA 2011).

In response to the fact that few studies in African climate have considered temporal

variations in the frequency and distribution of daily rainfall events of different magnitude

(Hulme 2003) and to the lack of information about the relationship between rainfall

intensity, land use change, and flash floods in Ethiopia, this study aim was threefold: (1) to

analyze the recent trends of rainfall intensity across Ethiopia in order to ascertain whether

the increased frequency of flash floods can be associated with an increase in daily pre-

cipitation amounts; (2) to estimate peak discharges for the Dechatu River and assess the

evidence for changing flood behavior; (3) to compare the general rainfall intensity trend

with the variation in the Dechatu River catchment in an attempt to investigate the causes of

such flash flood hazard worsening in this area and to discern the role of climate change

compared with that of land use change and human impact in general, i.e., a situation that is

common in many dry lands of developing countries.

Nat Hazards (2015) 76:1373–1394 1375

123

2 Study area

2.1 The climate of Ethiopia

According to the National Meteorological Agency of Ethiopia (NMA 1989), Ethiopia

shows a variety of climatic conditions. This is mainly due to its rugged physiography

characterized by large contrasts in altitude and vicinity to the Indian Ocean.

The NMA (1989), through the analysis of the rainfall pattern of a large number of

meteo-stations, demonstrated that in Ethiopia, there are different rainfall regimes. They can

be summarized in the following main categories:

1. Mono-modal (single maximum)

2. Bimodal (double maxima)

3. Diffused pattern.

The mono-modal regime is dominated by a single rainfall maximum with the wet

season running from February/March or April/May to October/November and from June/

July to August/September. The wet season decreases from ten months in the SW to only

two months in the NE. A good example is given by the rainfall pattern at Gore (Fig. 2).

The bimodal regimes is characterized by three seasons: (1) the dry season (bega in the

local language) from mid-October to January (occasional rains are recorded in places);

(2) the small rainy season (belg), from mid-February to mid-June, affected by a large inter-

annual variability; (3) the main, big rain season (kiremt), from mid-June to mid-October

that, on average, accounts for about 70 % of the annual precipitation. An example of this

regime is given by Alamata (Fig. 2). In some areas, the two peaks may be almost

equivalent as in Dire Dawa (Fig. 2).

In the diffused pattern regime, the rainfall pattern is irregular without a well-defined

rainy season as it is observed in the Danakil depression. Aseb in Eritrea is a good example

of that (Fig. 2).

0

50

100

150

200

250

300

350

J F M A M J J A S O N D

P (m

m)

Gore Alamata Dire Dawa Aseb

Fig. 2 Typical monthly rain patterns in Ethiopia as illustrated by four representative meteo-stations: Gore,mono-modal; Alamata, bimodal with a smaller peak in March/May and a more prominent peak in July/August; Dire Dawa, bimodal with two peaks almost equivalent; Aseb, diffused pattern

1376 Nat Hazards (2015) 76:1373–1394

123

The annual rainfall is also spatially highly variable with the deepest amounts in the

southwestern plateau between the Baro and Omo rivers with about 2,400 mm year-1 to the

eastern hyper-arid lowlands of Afar and Danakil where annual rainfall is commonly close

to zero.

According to NMA (1989) data, the largest proportion of the highlands is characterized

by a low (\30 %) variation coefficient of annual precipitation, whereas the low lands and

those areas with less annual rainfall such as eastern Tigray, central Ogaden, and Afar/

Danakil show a higher coefficient of variation as high as 70 % and above.

In Ethiopia, mean monthly maximum temperature (Tmax) varies at both space and time

scale. The highest temperatures (45 �C) occur in the Danakil depression with the secondary

maxima observed in the western and southern lowlands (NMA 1989), whereas the lowest

daily temperatures are recorded mostly on the highlands with values as low as 0 �C from

November to January. In coincidence with the main rainfall season, Tmax decreases all over

the country with the exception of the Danakil depression.

The annual temperature range is rather moderate all across the country, whereas the

daily temperature excursion is rather pronounced during the dry seasons reaching values as

large as 30 �C on the highlands and it is very much reduced during the rainy seasons with a

daily temperature range of even less than 10 �C.

2.2 The Dechatu River catchment

The Dechatu River drains the northern escarpment of the Harerge plateau, has a watershed

of about 660 km2 with the shape of an elongate triangle the apex of which points to N–NW

Fig. 3 Location map of the Dechatu River catchment

Nat Hazards (2015) 76:1373–1394 1377

123

(Fig. 3). Its highest point is 2,337 m a.s.l., whereas the lowest one coincides with the

confluence with the Lege Hare River at an elevation of about 1,100 m a.s.l., at the

downstream end of Dire Dawa. The flow gauge is located in a narrow reach, incised in the

bedrock, downstream of the confluence of the main tributaries that originate the Dechatu

River, and a couple of kilometers upstream of the town (Fig. 3). Only four rain gauges are

within or close to the river catchment; they are located in Dire Dawa, Kulubi, Dengego and

Alemaya (Fig. 3; Table 1).

The main stratigraphic units outcropping in the catchment are: (1) Precambrian meta-

morphic basement rocks, consisting mainly of gneiss with a variable density of thin granite

injections; (2) Triassic to lower Jurassic Adigrat sandstones: reddish, medium to coarse

sandstones and quartz conglomerates resting on the crystalline basement; (3) Antalo

limestone (Lower Cretaceous) including neritic, fossiliferous limestones and marls, rep-

resenting the most extensively outcropping formation in the catchment; (4) Cretaceous

Amba Aradam sandstones, consisting of patchy outcrops of fine to medium grained, friable

sandstone; (5) Quaternary alluvial and colluvial deposits (Merla et al. 1979).

The mean annual precipitation ranges from 1,022 mm at Kulubi, on the plateau margin,

down to 634 mm at Dire Dawa in the lowland, with both stations characterized by a

Table 1 Main characteristics of the meteo-stations considered in this study

No. Meteo-station Elevation (m a.s.l.) Annual rain (mm) Ip (mm/24 h)

1 Alamata 1,520 750.4 96.0

2 Alemayaa 2,047 764.1 118.0

3 Dengegoa 1,650 763.4 98.8

4 Dese 2,460 1,207.1 94.0

5 Dire Dawaa 1,260 638.7 122.3

6 Fitche 2,820 1,125.2 90.9

7 Gina Ager 3,160 1,693.2 170.8

8 Gode 295 239.9 174.0

9 Gonder 1,967 1,117.2 99.1

10 Gore 2,002 2,101.3 107.7

11 Hayk 2,030 1,174.0 132.8

12 Jimma 1,725 1,563.6 105.7

13 Kebri Dehar 550 325.3 128.0

14 Kobo 1,610 752.8 101.5

15 Konso 1,053 805.6 96.9

16 Kulubia 2,410 1,022.3 100.5

17 Mekele 2,070 598.7 95.5

18 Metahara 947 549.2 96.0

19 Moyale 1,097 685.0 167.6

20 Negelle 1,544 726.6 137.0

21 Nekemte 2,080 2,037.7 137.5

22 Robe Bale 2,480 876.3 112.3

23 Zikwala 2,980 1,124.8 132.6

Ip is the highest intensity ever recorded during the working time of the stationa Rain gauges within or near the Dechatu River catchment

1378 Nat Hazards (2015) 76:1373–1394

123

bimodal rainfall regime (see Sect. 2.1). In Dire Dawa, pan evaporation is about 3,140 mm

and in every month, potential evapotranspiration is greater than the average monthly

precipitation.

Dire Dawa is the second largest town of Ethiopia with 330,000 inhabitants, whereas

few, very small settlements are scattered within the Dechatu River catchment. The

physiography of the catchment is rather rugged, especially in the headwaters which are

located on the main escarpment of the Ethiopian Rift. Land cover consists for three

quarters of scrubland, open wood, and bare soil.

The Dechatu River is incised into the bedrock and no flood plain is found as far as the

town of Dire Dawa. Here, the channel is about 60–70 m wide and the streambed gradient

varies from 0.035 to 0.02. Upstream of Dire Dawa, the river channel is totally untouched,

whereas flood retaining walls are present in the urban area. They were constructed after the

2006 flood.

The Dechatu streambed is mainly sandy (70 % sand and 30 % gravel), and it is dry for

the most part of the year. Very large flows occur mainly during the summer rainy season in

response to short and intense downpours which sometimes cause flash flooding.

2.3 The Dechatu River flood of 2006

After a sequence of floods that occurred in 1981, 1994, 2004, and 2005 (Demessie 2007;

Alemu 2009; DDAEPA 2011) that caused significant fatalities and damages to property, on

August 6, 2006, the city of Dire Dawa experienced one of the largest and the most

devastating flood ever. The area inundated during the 2006 flood was about 1 km2 and

86 % of this area was covered by 1–2 m of water (Alemu 2009). Since all of the dwellings

in the inundated area are one story houses, this depth of inundation was big enough to

cause several casualties and property damage. This flood affected more than 117,000

people (i.e., one-third of the town population) and officially resulted in the loss of 256

human lives and 244 missing. It caused also the worst property damages to housing and

infrastructures in the town history with an estimated total damage of 10 million USD

(Alemu 2009).

Almost 90 % of the rainfall that generated the flood occurred in the day before (August

5) and the following rainfall depths in 24 h were recorded in the meteo-stations within or

close to the catchment perimeter: Kulubi, 100 mm; Kersa, 159 mm; Alemaya, 118 mm;

Dire Dawa, 37 mm (Fig. 3).

3 Data and methods

3.1 Rainfall intensity

Though Ethiopia has presently a modern and dense network of meteo-stations, daily rain

time series show commonly large gaps ranging from months to entire years and, unfor-

tunately, no station has a long (more than 50 years) and uninterrupted data records.

Moreover, given the vastness of the country and the variety of elevation, landscapes and

climate, data gap filling through correlation with neighboring meteo-stations is commonly

unreliable and poorly significant. The paper of Jury and Funk (2013) reports about spatial

variability of rainfall over Ethiopia, but no information is given about rainfall intensity.

These authors base their research on gridded data and interpolation. Given the low density

of meteo-stations in Ethiopia and its rugged topography, this procedure does not seem to

Nat Hazards (2015) 76:1373–1394 1379

123

provide a realistic pattern of rainfall distribution. Moreover, the procedure used by these

authors includes a five-station correlation to fill data gaps and a correlation coefficient as

low as 0.2 is accepted as significant. In the worst case, the data gap is filled with the mean

values of the time series. This method is considered here as questionable, as it does not

overtake the data limitations and introduces arbitrary conditions that are not necessarily

making the data analysis stronger or more reliable than the simple use of row data. Spatial

distribution of rainfall, considering 14 climate zones is investigated also by Viste et al.

(2013). These authors have used raw data obtained by the Ethiopian Meteorological

Agency, like in the present study, but their study focuses mainly on monthly precipitation.

Finally, a consistent study by Seleshi and Camberlin (2006), dealing with spatial and

seasonal (bega, belg, and kiremt) rainfall extremes, is available, but this paper is focused

on 5-day consecutive rain and no data on daily rainfall is reported.

Notwithstanding the awareness of the data limitations, the approach of this paper

attempts to fill an existing gap of knowledge about rainfall intensity in Ethiopia and in the

Dechatu River catchment and complement the analysis of seasonal and spatial variability

reported by previous papers (e.g., Segele and Lamb 2005; Seleshi and Camberlin 2006;

Korecha and Barnston 2007; Viste et al. 2013).

For this study, 19 meteo-stations with relatively long and as much as possible contin-

uous records were selected to investigate the variability of rainfall intensity in 24 h across

Ethiopia. These 19 meteo-stations are evenly distributed across the country (Fig. 4), but

their daily records encompass different time spans ranging from 1953 to 2010. For these

stations, the 1964–2009 time interval was selected as it has a higher degree of uniformity.

In the area of the Dechatu River, there are only four rain gauges. Three of them have a

daily data record spanning from 1981 to 2009, whereas the longest time series is recorded

at Dire Dawa and the 1964–2009 time interval was considered in accordance with the other

19 meteo-stations (Table 1).

Though the rainfall time series selected are more continuous as possible, data gaps are

still present and only in a few cases, it was possible to fill them by linear correlation with

the nearest station/s. We are well aware about such a limitation, but, in the alternative of

neglecting this part of the world and preventing it from any investigation on rainfall

intensity variation, we believe that the data processing of these time series, though affected

by the bias due to some gaps, can provide an insight into the medium- to long-term rainfall

intensity trend in Ethiopia and some indications to account for the role of climate change in

increasing frequency of flash flood in the Dechatu River and the town of Dire Dawa can be

detected anyway.

Fig. 4 Distribution acrossEthiopia of the study rainfallgauges. The numbers indicate themeteo-station name as reported inTable 1

1380 Nat Hazards (2015) 76:1373–1394

123

For each meteo-station, the following data were sorted: annual precipitation, mean

monthly precipitation, the highest rainfall intensity in 24 h (I24) recorded in each year and

the month in which it occurred, the mean values of I24 (Im) averaged on the entire time

series, the highest value of I24 ever recorded (Ip), and the return time for an intensity

I24 = 100 mm/24 h was calculated by the Gumbel method.

Floods are generated by a number of factors, the most important of which are rainfall

amount, catchment geomorphology (namely slope steepness), land use, and antecedent soil

moisture.

These conditions vary widely across Ethiopia, and, unfortunately, no information is

available about any average daily rainfall intensity threshold value capable to trigger a

flash flood.

In the USA, the River Forecast Centers produce rainfall–runoff curves on a regular basis

for each modeled basin. Changes in soil moisture due to recent rain or snowmelt are

included as well in the models to produce these curves. Therefore, when soil conditions

change, the rainfall–runoff relationship will also change.

In their study on flash floods in Europe, Marchi et al. (2010) have studied 25 extreme

events which were generated by storms with duration ranging from 7 to 22 h and rainfall

amounts between 100 and 300 mm. Rainfall intensities from 100 to 250 mm are reported by

Rusjan et al. (2009) to be responsible for flash floods in Slovenia and similar values were

measured also by Tantanee and Prakarnrat (2006) for flash floods in northern Thailand.

About 100 mm/24 h is the mean rainfall intensity recorded in the Dechatu River

catchment the day before the devastating flood of Dire Dawa in 2006. Moreover, all the 23

meteo-stations considered in this study have recorded a peak intensity of almost 100 mm/

24 h or more. Since no factual information about rainfall intensities capable to initiate

flooding in other regions of Ethiopia is available, this value was taken as a possible

reference threshold associated with potential conditions of flash flood risk generation,

especially on steep catchment across Ethiopia.

3.2 Flow discharge of the Dechatu River

Unfortunately, there is no information about the Dechatu River discharge since only flow

level data are available from March 2003 to September 2010 and no rating curve has ever

been constructed. In order to reconstruct the flood history and to calculate the peak dis-

charge of the main floods occurred in such interval, field measurements of cross section,

streambed gradient and bed material grain size were made in the reach with the flow gauge

in order to introduce these data into a simple uniform flow equation such as the Chezy

equation. The river bed at the flow gauge site is rather regular as it is straight, has a

rectangular cross-sectional geometry and is confined between bedrock slope sides (only on

the right banks there is a small alluvial accumulation). Given the reach geomorphology, the

variability through time of the monitoring site cross-section width can be considered as

negligible. As regards the variation of the streambed elevation, very scarce information is

reported in the literature, especially for ephemeral streams with a sand bed and hyper-

concentrated flow. Billi (2011), in his study on field measurement of bedload transport of

the Gereb Oda, a sand bed ephemeral stream in Tigray, Ethiopia, found that in case of

hyperconcentrated flow, streambed scouring is limited. This finding is supported also by

Powell et al. (2005) in their study using scour chains in Arizona. These latter authors

conclude that the streambeds experience little, if any, bed activity during an event. Fol-

lowing these considerations, the streambed elevation at the Dechatu flow gauge site was

considered as stable.

Nat Hazards (2015) 76:1373–1394 1381

123

The cross-sectional and streambed gradient were measured by a theodolite, whereas the

grain size frequency curve of bed material (Fig. 5) was obtained by the transect line,

frequency by number sampling method (Leopold 1970). Since bed material includes also a

non-negligible proportion of sand, the size of the sandy particles was identified by means

of a visual comparator with specimens of all the sand fractions, arranged on � phi scale,

stuck on a wooden tablet. The modal class grain of the sand in a 1 9 1 cm area near the

meter dent is considered and visually compared to the reference sieve specimens to assign

it to a specific phi class (Billi, unpublished).

Mean flow velocity (v) was calculated by the Chezy uniform flow equation

v ¼ CðRSÞ0:5 ð1Þ

in which R is the hydraulic radius, S the streambed gradient, and C the roughness

coefficient.

C ¼ ð8gÞ0:5=f 0:5 ð2Þ

in which g is gravity and 1/f 0.5 is the Darcy–Weisbach friction factor.

To calculate 1/f 0.5, the following equations were used: Leopold and Wolman (1957)

1=f 0:5 ¼ 1þ 2� log (h=D84Þ ð3Þ

in which h is mean depth and D84 is the grain size for which 84 % of the distribution is

finer, Limerinos (1970) and Knighton (1998), respectively:

1=f 0:5 ¼ 1:16þ 2� log(R=D84Þ ð4Þ

1=f 0:5 ¼ 0:82� ln(4:35� R=D84Þ ð5Þ

Equations (3) and (4) were selected because they are among the most used in the

literature and were found to be suitable for sandy gravel rivers. Equation (5) was derived

by Knighton (1998) from a very large set of field data measured on fine gravel to sandy

rivers and reported in the literature.

In addition, the regime theory equation of Lacey (1946) was used to calculate flow

velocity as:

0

20

40

60

80

100

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0

% fi

ner

D (Φ units)

Fig. 5 Grain size frequency distribution of bed material at the flow gauge reach

1382 Nat Hazards (2015) 76:1373–1394

123

v ¼ 10:8R2=3S1=3 ð6Þ

and was derived for fine-grained mobile bed rivers.

Finally, Grant’s equation (1997) was selected as well since it was developed on the base

of critical flow condition considerations for mainly sand bed streams as:

v ¼ ðghSÞ0:54:8ðh=D50Þ0:11 ð7Þ

in which D50 is the grain size for which 50 % of the distribution is finer.

No specific equation developed for fine-grained, ephemeral streams is available in the

literature; however, Eqs. (3) to (7) that result from different approaches are based on a

rather large set of field data and proved to work satisfactorily for gravelly sand rivers;

hence, they represent the best alternative to be used on a dryland river such as the Dechatu.

The river discharge was calculated for a wide range of flow levels recorded by the flow

gauge using all the criteria described above and the results averaged. By these results, a

stage/discharge rating curve was constructed (R2 = 0.95) and the discharge associated with

all the flow level data recorded was calculated. This enabled to reconstruct the flood history

of the Dechatu from March 2003 to September 2010 that is summarized in Fig. 6, in which

only the discharges higher than 300 m3 s-1 are reported. Though August 6, 2006 flood was

the most destructive, Fig. 6 shows that it was not the largest. The reasons of that will be

discussed later in the paper.

3.3 Land use/cover change

The following Landsat images were used to identify the soil use in the Dechatu River

catchment in the years 1985 and 2006: TM-09/03/1985 and ETM? 12/04/2006. Though

both images were taken during the belg (small rains) interval, they correspond to a phase of

little vegetative response following 30 days of negligible or no rainfall at all.

To identify the land use/cover, we adopted the NDVI index (Rouse et al. 1974), an

index widely used to evaluate vegetation and soil cover changes (e.g., Sohl 1999; Kindu

et al. 2013). Positive NDVI values indicate increasing amounts of green vegetation, NDVI

values near zero and negative values represent non-vegetated surfaces such as bare rock

and soil, water, snow, ice, and clouds.

Vegetation typically ranges between 0.2 and 0.8 with higher index values associated

with higher levels of healthy vegetation cover.

Fig. 6 Occurrence of floods with a discharge higher than 300 m3 s-1 observed after 2003

Nat Hazards (2015) 76:1373–1394 1383

123

The NDVI index is calculated following different phases: (1) Landsat 5 TM data are

converted to the equivalent Landsat 7 ETM ? data (DN) (Vogelmann et al. 2001), (2) DN

data (digital number 0–255) are converted to true radiance, then to reflectance (Chander

et al. 2009), (3) the NDVI index is calculated as NDVI = ((IR - R)/(IR ? R)) in which

IR and R are band 4 and 3, that is, near infrared and red band, respectively.

The NDVI maps were classified by using a supervised classification (MLC—maximum

likelihood classification). Four land use types were identified on the base of Google Earth�

high-definition images of the same period and the analysis of false color of the same

Landsat images: (1) bare soil, (2) scrubland, (3) open wood and (4) cultivated lands/mix

scrubland. The result of the classification, after considering the overlaps in spectral sig-

nature between the four soil uses, is reported in Fig. 7. The different land uses, expressed

as percentage of the Dechatu catchment area, and their variation between 1985 and 2006

are reported in Table 2.

Fig. 7 Land use/cover in the Dechatu catchment in a 1985 and b 2006

1384 Nat Hazards (2015) 76:1373–1394

123

4 Results

4.1 Rainfall intensity

The values of the highest rainfall intensity in 24 h ever recorded in the study meteo-

stations (Ip) is highly variable and range between 90.9 and 174.0 mm/24 h at Fitche and

Gore, respectively (Table 1). The highest annual rainfall intensity in 24 h (I24) may occur

in every month, but the frequency of both I24 and Ip follow a bimodal distribution with the

two modal classes coinciding with the belg and kiremt rain seasons (Fig. 8). This is not

surprising since the majority of Ethiopia is subjected to the bimodal rainfall regime.

The variability of both I24 and Ip is not affected by either elevation or annual rainfall as

no significant correlation was found between these parameters. On the other hand, the areas

with the higher values of average rainfall intensity (Im) experience also the highest peak

intensities as shown by the good correlation (R2 = 0.79) between Im and Ip (Fig. 9).

The return time interval for a rainfall intensity I24 = 100 mm/24 h was calculated using

the Gumbel method for the meteo-stations with at least 30 years of data record. The results

of this analysis show that for 50 % of the meteo-stations, the probability of a rainfall

intensity of 100 mm/24 h is less than one in 20 years and only at five stations the return

time is higher than 40 years (Fig. 10).

In order to investigate the rainfall intensity variability across Ethiopia during the last

three–five decades, 14 meteo-stations with the most uniform time series were selected and

the angular coefficient, m, of the trend lines was calculated for the 1964–2009 and

1981–2009 intervals (Table 3). The latter interval was included for a comparison with the

data of the Dechatu River watershed which, with the exception of Dire Dawa, cover only

the last three decades (Fig. 11). The data in Table 3 show a very irregular pattern of trends,

especially for the 1981–2009 interval, with m values ranging from -1.3 to 1.5. Though

Table 2 Land use/cover varia-tion between 1985 and 2006 inthe Dechatu River catchment

Land use/cover % in 1985 % in 2006 % variation

Bare soil 9.5 11.2 1.7

Scrubland 36.3 44.0 7.7

Open wood 25.9 20.3 -5.6

Cultivated/mix scrub 28.3 24.5 -3.9

0

5

10

15

20

25

J F M A M J J A S O N D

Freq

uenc

y (%

)

All sta�ons

I24 Ip

Fig. 8 Histogram of thefrequency of the months inwhich the yearly maximaand the absolute maximumof rainfall intensity wererecorded at each meteo-station

Nat Hazards (2015) 76:1373–1394 1385

123

more than one-third of the stations considered show a change of sign from negative to

positive (i.e., from decreasing to increasing trend), the mean value of m for the two

intervals remains close to zero, indicating no spatial correlation and no detectable change

at country scale. By contrast, the trend lines obtained for all the meteo-stations in the

Dechatu catchment show an increase in rainfall intensity in this area.

y = 0.6014x1.2951

R² = 0.79

0

50

100

150

200

250

0 20 40 60 80 100

I p(m

m/2

4h)

Im (mm/24h)

All sta�ons

Fig. 9 Correlation between the mean yearly maximum and the absolute maximum rainfall intensity foreach meteo-station

0 10 20 30 40 50 60

ZikwalaRobe BaleNekemte

NegelleMoyale

MeteharaMekele

KulubiKonsoKobo

Kebri DeharJimma

HaykGore

GonderGode

Gina AgerFitche

Dire DawaDese

DengecoAlemayaAlamata

Return Time (years)

I24 = 100 mmFig. 10 Return time of a rainfallintensity of 100 mm/24 h for thestudy meteo-stations

1386 Nat Hazards (2015) 76:1373–1394

123

These results indicate that (1) conditions of intense rainfall are ubiquitous all across

Ethiopia; (2) in the Dechatu River, rainfall intensities of the same order of magnitude of

those that generated the devastating flood of August 2006 have a high probability to occur

given the short return time ranging from 14 to 21 years calculated for Dire Dawa and the

neighboring station; (3) no significant change in rainfall intensity occurred across Ethiopia

Table 3 Angular coefficient, m,of the I

24trend lines

a Dechatu River catchmentmeteo-stations

Meteo-station 1964–2009 1981–2009

Fitche -0.081

Gina Ager 0.369 -0.977

Gode 0.520 1.514

Gonder -0.079 0.123

Gore -0.498 0.085

Hayk -0.037 0.371

Jimma 0.026 -0.104

Kebri Dehar 0.407 1.463

Mekele -0.023 0.493

Metehara -0.488 0.354

Moyale 0.511 0.564

Megelle -0.614 -1.310

Nekemte 0.072 -0.241

Robe Bale 0.386 0.115

Mean 0.048 0.048

Standard deviation 0.390 0.785

Dire Dawaa 0.296 0.302

Alemayaa 0.660

Dengegoa 0.201

Kulubia 0.777

Meana 0.486

0

20

40

60

80

100

120

140

1960 1970 1980 1990 2000 2010

I 24(m

m/2

4h)

Dire Dawa 1953-1980

Dire Dawa 1981-2009

Kulubi 1981-2009

Dengego 1982-2009Alemaya 1981-2009

Fig. 11 Time variation ofrainfall intensity in 24 h recordedby the Dechatu River meteo-stations

Nat Hazards (2015) 76:1373–1394 1387

123

during the last four decades, whereas the data of all the Dechatu stations show a marked

increase becoming more prominent after the mid-1990s.

As introduced in Sect. 3.1, whether or not a rainfall intensity of 100 mm in 24 h is

capable to generate a flash flood in the study area may depend on many factors, with the

three most important being: (1) the pattern of rainfall distribution, (2) the spatial distri-

bution, and (3) the initial soil moisture conditions. Since of these three main parameters

only the third one can be inferred from the available data for the Dechatu River, in order to

point out the role of and land use/cover change in the increased frequency of flash floods in

Dire Dawa during the last decades, the runoff generated by a rainfall of 100 mm in 24 h

was calculated by the curve number method for the land use/cover referred to 1985 and

2006. According to this method, the change in land use/cover observed in 2006 is

responsible for a runoff increase of about 4.5 %.

4.2 Flood flows

The floods larger than 300 m3 s-1, calculated by the methods described in Sect. 3.2, are

reported in Fig. 6. The flood of August 6, 2006 was the most devastating and killed 256

people. Its peak discharge was calculated to be around 1,508 m3 s-1 (2.28 m3 s-1 km-2).

Alemu (2009), using the Soil Conservation Service Curve Number method (USDA 1986),

obtained a peak discharge of 1,400 m3 s-1, i.e., a value very close to that calculated by the

simple uniform flow equation. Though the SCS-CN method is empirical and developed for

small catchments in the mid-western USA, it was found to be particularly suited for

streams with negligible base flow, i.e., rivers for which the ratio of direct runoff to total

runoff is close to one, as it is commonly observed in ephemeral streams of arid and

semiarid regions (Ponce and Hawkins 1996). This is also the case of the Dechatu that is dry

for most of the time and has some water flowing only in response to individual, intense

rainstorms. Therefore, it is not surprising that the hydraulic and hydrologic approaches

used, though conceptually different, produced a very similar result.

In the last decade, the largest flood had a peak discharge of about 2,338 m3 s-1

(3.54 m3 s-1 km-2) and occurred on May 20, 2005. Other six large floods, with peak

discharge (Qp) higher than 1,000 m3 s-1 occurred on April 14, 2004 (Qp = 1,095 m3 s-1),

March 20, 2005 (Qp = 1,456 m3 s-1), March 25, 2006 (Qp = 1,269 m3 s-1), April 6,

2006 (Qp = 1,080 m3 s-1), April 12, 2007 (Qp = 1,508 m3 s-1), and March 8, 2010

(Qp = 1,118 m3 s-1).

The August 6, 2006 flood was by far the worst in terms of human lives lost because it

was a typical, very flashy flood and occurred in the night, before dawn, when most of the

people were sleeping. According to eyewitness reports, in fact, the flood wave had a high

velocity of propagation and peak discharge followed shortly after the flood onset. The

largest flood of May 20, 2005 instead occurred during the day and people had time to move

to safe places. After the August 6, 2006 flood, retaining walls were constructed throughout

Table 4 Discharge in m3 s-1 of the Dechatu River at Dire Dawa for floods with different return time

Method Q5 Q10 Q20 Q50 Q100

Gumbel EV 1,611 1,981 2,336 2,795 3,139

Semilog 1,227 1,636 2,045 2,586 2,995

Q5–100 indicate floods with return time from 5 to 100 years

1388 Nat Hazards (2015) 76:1373–1394

123

the river reach within the town of Dire Dawa and the subsequent floods, though a couple of

them were of the same order of magnitude as the previous ones, had limited impact.

In order to quantify the magnitude of the August 6, 2006 flood and the largest flood of

May 20, 2005 (Demessie 2007; Alemu 2009), a flood frequency curve was constructed

with the discharge data calculated on the basis of the water level measured by the flow

gauge. The discharges with the return times of 5, 10, 20, 50, and 100 years were calculated

using the Gumbel EV method and by a simple semilog interpolation (Table 4). The results

in Table 4 indicate that the return interval of the largest flood of May 20, 2005 is 20 years,

whereas that of August 6, 2006 flood is about 4.5 years. The data used to calculate the

flood frequency curve, though from one side may reflect the most recent flood trend, from

the other cover a very short time interval (March 2003 to September 2010) that can make

the results obtained questionable. Given the limitation of the data available, in an attempt

to, at least, indirectly confirm the results obtained, the most renown empirical equations to

calculate the maximum flood (i.e., Q100) reported in the literature were applied to the

Dechatu. They are the equations of Pagliaro (1936, in Maione 1977); Rodier and Roche

(1984); and Griffiths and McKerchar (2008) which returned the following discharges for

Q100, 2,552, 8,154 and 4,536 m3 s-1, respectively. None of these criteria were specifically

developed for ephemeral streams of arid and semiarid regions. In fact, the data of Pagliaro

refers to Italian rivers data of the early twentieth century, that is, with less human impact

and less forest cover than the present day; Rodier and Roche’s equation is very general

since it was derived from data of the world catalog of very large floods including rivers

from all over the world; and Griffiths and McKerchar data refer to rivers in the Westland of

New Zealand South Island, where a temperate climate with high rainfall prevails. A subset

of data of maximum flood recorded in rivers of arid and semiarid environments, with

catchment area less than 2,500 km2, was extracted from Rodier and Roche (1984) and data

of catchment area (A) were plotted versus maximum recorded discharge (Qp).

The best fitting curve (R2 = 0.70) is the following:

Qp ¼ 92:625A0:5902 ð8Þ

with Qp in m3 s-1 and A in km2, and explains 70 % of the maximum discharge variability.

According to Rodier and Roche, Qp is very close to Q100. By applying Eq. (8) to the

Dechatu, we obtain Q100 = 4,257 m3 s-1. This value is about 25 % larger than that cal-

culated by Gumbel EV method with the flow data recordings available but, given the

brevity of the time series, which may result in an approximation for defect, a discharge

between 3,100 and 4,200 m3 s-1 can be considered as a possible actual value for the

Dechatu’s Q100. Following this result, the return time intervals of the largest flood of May

20, 2005 and of the devastating flood of August 6, 2006 are expected to be larger. The

return intervals of I24 = 100 mm/24 h calculated for the four meteo-stations within or

close to the Dechatu catchment (Dengego, Dire Dawa, Alemaya and Kulubi—Fig. 3) range

from 14 to 21 years and seem to confirm a larger return time for these high floods.

5 Discussion

A large number of studies have examined potential trends of river discharge during the

twentieth century. Many of them, however, have found no trends or have been unable to

separate the effects of variations in temperature and precipitation from the effects of human

interventions in the catchment (Di Baldassarre et al. 2010). In the case of the Dechatu

Nat Hazards (2015) 76:1373–1394 1389

123

River, though flow data have been recorded only during the last decade, local authorities

(DDAEPA 2011) and local witnesses have confirmed the increased frequency of high flash

floods. In the recent past, in fact, only four, minor floods affected Dire Dawa in 1945, 1977,

1981, and 1994 (Alemu 2009). No flow data are available for these floods and are classified

as ‘‘minor’’ because they are not reported to have caused fatalities and property damage.

The comparison between the flood frequency in the 1945–1994 and the 2003–2010 (Fig. 6)

intervals confirms that the flood hazard in Dire Dawa has markedly increased in the second

interval.

The analysis of I24 time series produced contrasting results for the 14 meteo-stations

selected to represent the variability of extreme rainfall at country scale. Eight out of 14

stations are characterized by an increasing trend over the 1964–2009 interval, whereas the

proportion of decreasing trend lines is 8 out of 15 stations for the 1981–2009 period.

Though the number of data with negative values of m is a little higher in the last three

decades, it is worth noticing that only two stations maintain the same sign and the change

from positive to negative is observed only in three stations. At country scale, it can be

concluded that there is a high variability of extreme rainfalls in time and space, but no clear

overall trend can be recognized (mean m = 0.048—Table 3). By contrast, in the Dechatu

stations, the trend lines are all positive (mean m = 0.486—Table 3) and indicate an

average rainfall intensity increase of 32 % from 1981 to 2009 (Fig. 11).

In his 2009 paper, Morin investigated the minimal detectable absolute trends in annual

precipitation and concluded that similar analyses may be applied to other related variables.

The daily rainfall intensity trends calculated in this study show rates of change that are

within the limits reported by this author for the study areas to be considered as minimal

detectable absolute trend. The reasons why different areas of Ethiopia are experiencing

different patterns of extreme rainfall variability are complex and beyond the scope of this

paper; however, it is rather evident that the Dechatu River has been subjected to a marked

increase in extreme rains throughout the last three decades (Fig. 11). Furthermore, if the

data of Dire Dawa are considered, it appears that rainfall intensities start to increase in

1970 and proceed at an almost constant rate as far as 2009 (Fig. 11); therefore, this factor

alone cannot be invoked to account for the concentration of high floods observed in the last

decade (Fig. 6) and other factors are to be considered.

Land use/cover changed significantly from 1985 to 2006 and a nick point can be

identified in the first years following the change of Government in Ethiopia that formally

occurred in 1992. Some problems were encountered in defining with accuracy the

boundaries between some cultivated areas and the neighboring scrubland. In the study area,

traditionally, people were not used to cultivate perennial crops and that resulted in very

serious soil erosion problems. From the late period of the DERG government up to the

present, construction of terraces and planting eucalyptus trees have been the physical and

biological measures implemented in the Dechatu catchment, particularly in the down-

stream portion, to contrast erosion. Late land management plans of the DERG forced the

farmers to set up cultivations also on the steep slopes of the Dechatu headwater and,

though soil countermeasures such as terracing and eucalyptus tree plantations were

implemented through the food for work program, the local farmers considered the natural

constraints and limitations too hard to cope with and making such cultivations scarcely

remunerative and rewarding. After the fall of the DERG regime in 1991, these cultivations

were abandoned and grown trees were cut and sold for daily consumption. The abandoned

cultivations, though partly obliterated by the spontaneous re-growth of scrub vegetation,

still maintain traces of their original structure, and for these reasons, they are not easily

1390 Nat Hazards (2015) 76:1373–1394

123

distinguished by the NDVI analysis. Presently, only little cultivation is taking place due to

the limitations of soil degradation severity.

This important land management transformation, consisting substantially in crop and

conservation practices abandonment, which occurred after 1991, has left the slopes,

especially in the river headwaters, devoid of any efficient soil maintenance and water

conservations practices, de facto giving rise to conditions that favor higher than before

overland flow volumes

In order to assess the relative influence of land use/cover change, the runoff generated

by a rainfall of 100 mm in 24 h was calculated by the curve number method and an

increase of about 4.5 % was found for the land use/cover conditions of 2006 with respect to

1985.

The average rainfall intensity calculated from the data of the Dechatu meteo-stations is

52.8 mm/24 h for the 1981–1991 interval and becomes 60.1 mm/24 h for the period

1992–2009, i.e., after agriculture abandonment, with an increment of 14 %.

These results suggest that in the Dechatu catchment, notwithstanding the marked

change in land use/cover and conservation practices that occurred in the last three decades,

the recent increase in rainfall intensity seems to be a key concurrent factor, presumably

more important than poor land management in contributing to rise the risk level of flash

floods in the town of Dire Dawa. In the very recent years, however, the local adminis-

trations have constructed high flood retaining walls that have contained the last large floods

of 2010 with peak discharges around 1,000 m3 s-1 (Fig. 6), preventing the town from

further inundation, and started a new program of flash flood risk reduction by expanding

and renovating the old soil and water conservation measures on the slopes of the Dechatu

catchment headwater.

6 Conclusions

The analysis of daily rainfall data over a time span ranging from 50 to 30 years from 19

meteo-stations scattered across Ethiopia has shown that almost all the country is subjected

to high rainfall intensities ranging from 90 to 170 mm/24 h. Though the highest values of

daily rainfall may be recorded in every month, i.e., also in the dry interval, their distri-

butions follow the typical monthly rainfall pattern characterized by two almost equivalent

modal classes in April and August.

On the base of the rainfall data associated with the devastating flood of Dire Dawa on

August 6, 2006, a rainfall intensity of 100 mm/24 h was taken as a reference value capable

to set conditions for a flash flood. In about 60 % of the meteo-stations considered, a rainfall

intensity of 100 mm/24 h has a probability to occur as once in less than 20 years, and in all

but five meteo-stations such intensity has a return time \40 years.

The Dechatu River is an ephemeral stream which is dry for most of the time. During the

last decade (2003–2010), however, 23 flash floods were recorded. These floods range from

335 to 2,338 m3 s-1, equivalent to 0.51 and 3.54 m3 s-1 km-2, respectively, and seven of

them had a discharge higher than 1,000 m3 s-1 (1.52 m3 s-1 km-1). The return time of the

largest flood recorded on May 20, 2005 and of the most devastating flood of August 6,

2006 is calculated as 20 and 4.5 years, respectively. These values were obtained from a

very short time series and a comparison with the maximum possible flood (Q100), derived

by regression analysis between measured maximum flood and catchment area of rivers in

arid and semiarid areas (subset of data from Rodier and Roche 1984), and the maximum

rainfall intensity in 24 h, recorded at the Dechatu catchment meteo-stations, indicates that

Nat Hazards (2015) 76:1373–1394 1391

123

the higher floods return interval is likely longer than that calculated by the Gumbel flood

frequency analysis.

The maximum rainfall intensity (I24) recorded in the study meteo-stations shows a

contrasting pattern of decrease and increase trends for different stations across Ethiopia,

but the long-term mean of the trend lines angular coefficient does not show any evidence of

change. By contrast, a marked increase is observed in the Dechatu catchment. Here, in the

last decades after 1991 (i.e., after the general abandonment of agriculture and land man-

agement practices, following the fall of the DERG Government, especially in the Dechatu

River headwaters), extreme rainfall intensity has increased by 14 %.

The land use/cover change in the Dechatu catchment was analyzed by means of 1985

and 2006 satellite images interpretation. This analysis indicates that the land transforma-

tions occurred between 1985 and 2006 resulted in an increase in all the factors that favor

larger volumes of overland flow. The use of the curve number method, refereed to the 1985

and 2006 land use/cover situations shows a runoff increase of 4.5 %. This value, however,

is modest if compared with the three times larger increase in rainfall intensity observed

after 1991. The increase in extreme rains, paired by a marked change in land use/cover and

management practices, is considered the main factor responsible for the increased fre-

quency of high flash floods in the town of Dire Dawa during the last decade, though the

increase in rainfall intensity is likely playing a more relevant role.

Acknowledgments The authors are indebted to two anonymous referees who posed key points, stimulateddiscussion, and greatly improved the general quality of the manuscript. This paper was funded by theNational Geography Grant 8400-08, University of Ferrara and University of Dire Dawa funds. The authorsare indebted to Girma Moges and to the technical staff of the Geography Department of Dire DawaUniversity for their assistance during the field measurements on the Dechatu River. The Department ofHydrology of the Ethiopian Ministry of Water Resources is acknowledged for providing the flow level dataof the Dechatu River.

References

Adhikari P, Hong Y, Douglas KR, Kirschbaum DB, Gourley J, Adler R, Brakenridge GR (2010) A digitizedglobal flood inventory (1998–2008): compilation and preliminary results. Nat Hazards 55:405–422.doi:10.1007/s11069-010-9537-2

Alemu YT (2009) Socio-economic impact of flooding in Dire Dawa, Ethiopia. Unpublished MSc. Thesis,International Center for Water Hazard and Risk Management, Tsukuba, Japan

Alexander LV, Zhang X, Peterson TC, Caesar J, Gleason B, Klein Tank AMG, Haylock M, Collins D,Trewin B, Rahimzadeh F, Tagipour A, Kumar KR, Revadekar J, Griffiths G, Vincent L, StephensonDB, Burn J, Aguilar E, Brunet M, Taylor M, New M, Zhai P, Rusticucci M, Vazquez-Aguirre JL(2006) Global observed changes in daily climate extremes of temperature and precipitation. J GeophysRes Atmos 111:D05109

Billi P (2011) Flash flood sediment transport in a steep sand-bed ephemeral stream. Int Jour. Sediment Res26(2):193–209

Chander G, Markham BL, Helder DL (2009) Summary of current radiometric calibration coefficients forLandsat MSS, TM, ETM?, and EO-1 ALI sensors. Remote Sens Environ 113:893–903

Conway D, Schipper ELF (2011) Adaptation to climate change in Africa: Challenges and opportunitiesidentified from Ethiopia. Glob Environ Change 21:227–237

Conway D, Persechino A, Ardoin-Bardin S, Hamandawana H, Dieulin C, Mahe G (2009) Rainfall and waterresources variability in sub-Saharan Africa during the twentieth century. J Hydrometeorol 10(1):41–59

DDAEPA (Dire Dawa Administration Environmental Protection Authority) (2011) Dire Dawa Adminis-tration, program of adaptation to climate change. DDEPA, Dire Dawa

Demessie DA (2007) Assessment of flood risk in Dire Dawa town, eastern Ethiopia, using gis. MSc Thesis,Addis Ababa University

1392 Nat Hazards (2015) 76:1373–1394

123

Di Baldassarre G, Montanari A, Lins H, Koutsoyiannis D, Brandimarte L, Bloschl G (2010) Flood fatalitiesin Africa: from diagnosis to mitigation. Geophys Res Lett 37:L22402

Easterling DR, Evans JL, Groisman PY, Karl TR, Kunkel KE, Ambenje P (2000) Observed variability andtrends in extreme climate events. Bull Am Meteorol Soc 81:417–425

EM-DAT (2014) The OFDA/CRED International Disaster Database—www.emdat.be, Universite Catho-lique de Louvain, Brussels (Belgium)

Grant GE (1997) Critical flow constraints flow hydraulic in mobile-bed streams: a new hypothesis. WaterResour Res 33(2):349–358

Griffiths GA, McKerchar AI (2008) Dependence of flood peak magnitude on catchment area. J Hydrol (N Z)47(2):123–131

Griffiths GM, Salinger MJ, Leleu I (2003) Trends in extreme daily rainfall across the South Pacific andrelationship to the South Pacific Convergence Zone. Int J Clim 23(8):847–869

Herath S, Ratnayake U (2004) Monitoring rainfall trends to predict adverse impacts—a case study from SriLanka (1964–1993). Glob Environ Change 14(1):71–79

Hulme M (2003) Climate change within the period of meteorological records. In: Adams WM, Goudie WS,Orme AR (eds) The physical geography of Africa. Oxford University Press, Oxford, pp 88–102

IPCC (2008) Technical Paper VI. In: Bates BC, Kundzewicz ZW, Wu S, Palutikof JP (eds) IPCC Secre-tariat, Geneva, 210 pp

IPCC (2012) Managing the risks of extreme events and disasters to advance climate change adaptation.Cambridge University Press, Cambridge

Jury MR, Funk C (2013) Climatic trends over Ethiopia: regional signals and drivers. Int J Clim33(8):1924–1935

Kindu M, Schneider T, Teketay D, Knoke T (2013) Land use/land cover change analysis using object-basedclassification approach in Munessa-Shashemene Landscape of the Ethiopian Highlands. Remote Sens5:2411–2435

Knighton AD (1998) Fluvial forms and processes: a new perspective. Arnold, LondonKorecha D, Barnston AG (2007) Predictability of June–September rainfall in Ethiopia. Mon Weather Rev

135(2):628–650Lacey G (1946) A theory of flow in alluvium. J Inst. Civil Eng 57, Paper No. 5518, 16-47Leopold LB (1970) An improved method for size distribution of stream bed gravel. Water Resour Res

6(5):1357–1366Leopold LB, Wolman MG (1957) River channel pattern: braided meandering and straight. U.S. Geol. Surv.

Prof. Paper 282B:39–85Limerinos JT (1970) Determination of the Manning coefficient from measured bed roughness in natural

channels. U.S. Geol. Surv., Water Supply Pap 1898-B, 47 ppLin X (ed.) (1999) Flash floods in arid and semi-arid zones. IHP-V Technical Documents in Hydrology, No.

23, UNESCO, Paris, 65 ppMaione U (1977) Appunti di idrologia. Le piene fluviali. Ed. La Goliardica Pavese, Pavia, 224 ppManton MJ, Della-Marta PM, Haylock MR, Hennessy KJ, Nicholls N, Chambers LE, Collins DA, Daw G,

Finet A, Gunawan D, Inape K, Isobe H, Kestin TS, Lefale P, Leyu CH, Lwin T, Maitrepierre L,Ouprasitwong N, Page CM, Pahalad J, Plummer N, Salinger MJ, Suppiah R, Tran VL, Trewin B, TibigI, Yee D (2001) Trends in extreme daily rainfall and temperature in Southeast Asia and the SouthPacific: 1961–1998. Int Jour Clim 21(3):269–284

Marchi L, Borga M, Preciso E, Gaume E (2010) Characterisation of selected extreme flash floods in Europeand implications for flood risk management. J Hydrol 394(1–2):118–133

Merla G, Abbate E, Azzaroli A, Bruni P Canuti P, Mazzuoli M, Sagri M, Tacconi P (1979) A geologicalmap of Ethiopia and Somalia 1:2.000.000 and comment. Consiglio Nazionale delle Ricerche, Uni-versity of Florence, Italy

Mulugeta G, Ayonghe S, Daby D, Dube PO, Gudyanga F, Lucio F, Durrheim R (2007) Natural and human-induced hazards and disasters in sub-Saharan Africa. ICSU (International Council For Science)Regional Office for Africa, Pretoria

NMA (1989) Climatic and agroclimatic resources of Ethiopia. National Meteorological Agency, AddisAbaba

Peterson TC, Taylor MA, Demeritte R, Duncombe DL, Burton S, Thompson F, Porter A, Mercedes M,Villegas E, Fils RS, Klein Tank A, Martis A, Warner R, Joyette A, Mills W, Alexander L, Gleason B(2002) Recent changes in climate extremes in the Caribbean region. J Geophys Res Atmos 107:4601

Ponce VM, Hawkins RH (1996) Runoff Curve Number: has it reached maturity? ASCE J Hydrol Eng1(1):11–19

Powell DM, Brazier RE, Wainwright J, Parsons AJ, Kaduk J (2005) Stream-bed scour and fill in low-orderdryland channels. Water Resour Res 41:W05019. doi:10.1029/2004WR003662

Nat Hazards (2015) 76:1373–1394 1393

123

Rodier JA, Roche MJ (1984) World catalogue of maximum observed floods. IAHS Publ No 143Rosenzweig C, Tubiello FN (2007) Adaptation and mitigation strategies in agriculture: an analysis of

potential synergies. Mitig Adapt Strateg Glob Change 12:855–873Rouse JW, Haas RH, Schell JA, Deering DW (1974) Monitoring vegetation systems in the Great Plains with

ERTS. In: Fraden SC, Marcanti EP, Becker MA (eds) Third ERTS-1 symposium, NASA SP-351.NASA, Washington, DC, pp 309–317

Rusjan S, Kobold M, Mikos M (2009) Characteristics of the extreme rainfall event and consequent flashfloods in W Slovenia in September 2007. Nat Hazards Earth Syst Sci 9:947–956

Segele ZT, Lamb PJ (2005) Characterization and variability of Kiremt rainy season over Ethiopia. MeteorolAtmos Phys 89(1):153–180

Seleshi Y, Camberlin P (2006) Recent changes in dry spell and extreme rainfall events in Ethiopia. TheorAppl Clim 83(1–4):181–191

Sohl TL (1999) Change Analysis in the United Arab Emirates: an Investigation of Techniques. PhotogrammEng Remote Sens 65(4):475–484

Tantanee S, Prakarnrat S (2006) Post flash flood hydrological investigation. Adv Geosci 11:265–282Terefe A, Tsige G, Wondafrash G, Alemu YT (2006) Causes, consequences and management aspects of the

2006 flooding hazards in Awash river basin. In: Kelbessa E (ed) Ethiopia between the sharp scissors:climate change and environmental degradation (Forum for Environment, December 2006), pp 41–42

United States Department of Agriculture (1986) Urban hydrology for small watersheds. Technical Release55 (TR-55) (Second Ed.), Natural Resources Conservation Service, Conservation Engineering Division

Viste E, Korecha D, Sorteberg A (2013) Recent drought and precipitation tendencies in Ethiopia. TheorAppl Climatol. doi:10.1007/s00704-012-0746-3

Vogelmann JE, Howard SM, Yang L, Larson CR, Wylie BK, Van Driel JN (2001) Completion of the 1990’sNational land cover data set for the conterminous United States. Photogramm Eng Remote Sens67:650–662

1394 Nat Hazards (2015) 76:1373–1394

123