Trends in Rainfall and Runoff in the Blue Nile Basin: 1964-2003

Zelalem K. Tesemma 1, Yasir A. Mohamed 2, 3 , Tammo S. Steenhuis 1,4

1 Integrated Watershed Management and Hydrology Master’s Program, Cornell University, Bahir Dar,

Ethiopia.

2 International Water Management Institute, IWMINBEA, PO Box 5689, Addis Ababa, .

3 UNESCOIHE Institute for Water Education, P.O. Box 3015, 2601DA Delft, Netherlands.

4 Biological and Environmental Engineering, Cornell University, Ithaca, NY 14853, USA.

Abstract

Most Nile water originates in Ethiopia but there is no agreement on how land degradation or climate change affects the future flow in downstream countries. The objective of this paper is to improve understanding of future conditions by analyzing historical trends. During the period 1963 to 2003, average monthly basin wide precipitation and monthly discharge data were collected and analyzed statistically for two stations in the upper 30% of Blue Nile Basin and one station at the Ethiopia border. A rainfall runoff model examined the causes for observed trends. The results show that while there was no significant trend in the seasonal and annual basinwide average rainfall, significant increases in discharge during the long rainy season (June to September) at all three stations were observed. In the upper Blue Nile the short rainy season flow (March to May), increased while the dry season flow (October to February) stayed the same. At the Sudan border the dry season flow decreased significantly with no change in the short rainy season flow. The difference in response was likely due to weir construction in the nineties at the Lake

Tana outlet that affected significantly the upper Blue Nile discharge but only affected less than 10% of the discharge at the Sudan border. The rainfall runoff model reproduced the observed trends, assuming that an additional ten percent of the hillsides were eroded in the 40 year time span and generated overland flow instead of interflow and base flow. Models concerning future trends in the Nile cannot assume that the landscape runoff processes will remain static.

Key words : Climate change, Watershed hydrology, Model, RainfallRunoff models, Blue Nile.

1 Introduction be an effective method. (Yilma and Demarce,

1995;; Kim, 2008; Conway, 2000) especially if

The Nile basin is one of the most waterlimited these trends can be related to changes in land use basins in the world. Without the Nile major and rainfall. portions of Sudan and Egypt would run out of water. There is a growing anxiety about climate Previous studies employed simple linear induced changes of the river’s discharge, regressions over time to detect trends in annual especially because Ethiopia, which generates runoff and rainfall series without removing the

85% of the annual Main Nile flow (Sutcliffe and seasonal effects or trying to predict seasonal

Parks, 1999), is actively planning major differences in discharge (Conway, 2000; hydropower and irrigation development. To Sutcliffe and Parks, 1999). The objective of this develop appropriate adaptation strategies to relay research is therefore to improve on these these concerns, longterm trends in stream flow predictions by using both the MannKendall and should be investigated (Conway, 2000; Conway Sen’s T test to detect trends in both seasonal and and Hulme, 1993 1996; Yilma and Demarce, annual runoff and rainfall and then using a semi

1995; Kim et al., 2008), which requires a better distributed rainfall runoff model to both confirm understanding of the basin’s hydrology and that the rainfall runoff relationship is changing embedded longterm variability over to forty year period and to find the

underlying physical conditions that explains the

The literature shows an increasing number of observed runoff trends. climate change studies in the Nile basin (e.g.,

Conway and Hulme 1993, 1993; Conway, 2000; The Blue Nile Basin

Elshamy et al., 2009; Strzepek et al., 1996; Kim et al., 2008). Impact of climate change on Blue The Upper Blue Nile River (named Abbay in

Nile discharge was highly variable in these Ethiopia) starts at Lake Tana and ends at the studies. One of the reasons is that the Global EthiopiaSudan border. The topography of the

Circulation Models cannot even agree on the Blue Nile is composed of highlands, hills, sign(s) of change (Elshamy et al., 2009). valleys and occasional rock peaks. Most of the

Therefore, predicting future scenarios by streams feeding the Blue Nile are perennial. The studying past trends of rainfall and discharge can average annual rainfall varies between 1200 and

2 1800 mm/yr (Figure 1 a), ranging from an May; and long rainy period (Kiremt) from June

average of about 1000 mm/year near the to September, with the greatest rainfall occurring

Ethiopia/Sudan border,to1400 mm/yr in the in July and August. The year to year variation in

upper part of the basin, and in excess of 1800 monthly rainfall is most pronounced in the dry

mm/yr in the south within Dedessa subbasin season, with the lowest annual variation

(Conway 2000; Sutcliffe and Parks, 1999). occurring in the rainy season. Interannual

Locally the climatic seasons are defined as: dry variability of rainfall in the basin is 10% (Table

season (Bega) from October to the end of 1).

February; short rain period (Belg) from March to

Table 1: The seasonal MannKendall and Sen’s T tests statistics for Upper Blue Nile basin hydroclimatology record from1963 to 2003.

UB Station Seasons Mean 1 CV 2 z Test 3 T test 4 Slope 5 Change 6 Change 7 (%) (10 6 m3 / yr) (Billion m 3) (%) Rainfall ( mm) Areal rainfall Annual 1286 10 0.5 0.5 Dry 151 36 0.6 1.1 Short rainy 218 26 0.6 0.5 Long rainy 916 10 0.2 0.7

Runoff ( ( Billion m 3) Bahir Dar Annual 3.8 36 2.0 3.3 Dry 2.0 33 0.6 0.9 Sho rt rainy 0.3 100 3.2 2.4 2.1 0.08 33 Long rainy 1.5 45 2.7 2.6 9.8 0.39 26

Kessie Annual 16.0 31 2.2 3.8 109.0 4.36 27 Dry 3.2 30 0.8 1.0 Short rainy 0.6 60 3.2 3.2 7.2 0.288 51 Long rainy 12.2 35 3.6 2.7 83.7 3.35 27

El Di em Annual 46.9 20 0.5 0.3 Dry 11.0 32 -2.5 -2.4 28.3 1.13 10 Short rainy 1.3 34 0.8 0.7 Long rainy 34.6 19 3.0 2.0 87.7 3.52 10 Note: Bold figures are significant at 5% significance level. Dry season (OctFeb); Short rainy season (MarchMay); Long rainy season (June – September) 1= Mean of the seasonal total runoff/rainfall (19642003). 2= coefficient of variation (19642003). 3=the MannKendall test statistics. 4=Sen’s T test statistics. 5=Sen’s slope estimator. 6=calculated as slope times years of record (40 years). 7=calculated as change over the respective mean seasonal runoff

3 The longterm (19122003) mean annual 1997). There is uncertainty about how forest discharge of Blue Nile entering Sudan and cover has changed over the last 50 years. Some measured at Roseires/El Diem is 48.9 *10 9 m3/yr report a decrease (USBR, 1964; Mohammed, which is about 60% of the flow of Main Nile 2007) while Bewket (2002) showed that green

(Sutcliffe and Parks, 1999), with flows of cover has increased since 1950 over the 364 km 2

3.9*10 9 m3/yr at Bahir Dar (19592003) and 16.3 Chemoga watershed in the upper Blue Nile

*10 9 m3/yr at Kessie (19532003) respectively basin.

(Figure 1b). The distribution of seasonal Input for statistical analysis and rainfall discharge varies considerably (Figure 1c). The runoff modeling average discharge at El Diem is smallest in April Input Data and greatest in August, about 35 times the April Monthly data were collected for statistical flow. The annual variability of stream flow analysis, and modeling required 10day data. varies Monthly rainfall data for statistical analysis were

downloaded from Global Historical Climatology

Network (NOAA, 2009) and the 10day rainfall

data for the selected stations (shown in Figure 2)

were obtained from the National Meteorological

by less than 20% (Conway and Hulme, 1993;

Conway, 2000; Yilma and Demarce, 1995).

Most of the soil types covering the Blue Nile basin are volcanic vertisols or latosols (Conway, Services Agency of Ethiopia. Monthly stream

4 flow data were obtained from the Hydrology UNESCO/IHP available at

Department of the Ministry of Water Resources http://dss.ucar.edu/datasets/ds553.2/data/. From of Ethiopia, and Ministry of Irrigation and Water the data available, three stream flow gages

Resources of Sudan and the Global Hydro (Figure 2) were selected that had more than 25

Climate Data Network operated by years data, which is sufficiently long to yield statistically valid trends (Burn and Elnur, 2002). decimal digits were fixed. Outliers were

Of these, the gaging station at El Deim at the identified by comparison with upper and a lower

Sudanese Ethiopian border had the longest and boundary limits. Values outside the limits were most reliable record, extending from 1912 to further validated by comparing the data plots of present (Conway, 2000; Sutcliffe and Parks, neighboring stations. The confirmed suspect

1999). The Kessie hydrometric station is located values were removed and replaced by values near the bridge where the main road to Addis derived by a relation curve with neighboring

Ababa from Bahir Dar crosses the Abbay (Blue station(s). Missing data of the rainfall were fitted

Nile) river, with discharge data recorded since using best fit regression with neighboring

1953. Except for the last few years during the stations. bridge construction, the data is fair to good

(Conway, 2000). The third station is downstream Methodology of the outlet of Lake Tana in Bahir Dar. The construction in 1996 of the CharaChara weir for Both statistical analysis and a semidistributed generating hydropower has affected the rainfall runoff model were used to assess trends discharge by storing water in Lake Tana during in the discharge in the Blue Nile basin. The the wet season and releasing it during dry season. statistical analysis of trends in climate and

hydrologic variables uses the MannKendall test

Data validation and completion (Zhang et al, 2001; Huth and Pokorna, 2004;

After the raw rainfall and discharge data were Harry et al, 1999). To gain more confidence in collected, a thorough checking and validation our results, a categorically different and less was performed. First the data were visually common technique, Sen’s T test, was employed screened, and mistyped numbers and misplaced as well (KarabÖrk, 2007). Both tests are non assumptions about the distribution of the parametric approaches and do not require any variables.

5 widely to identify trends in hydroclimatic

Mann-Kendall test variables (see e.g., Kahya and Kalayci, 2004; Xu

The MannKendall (Mann, 1945; Kendall, 1975) et al., 2003; Partal and Kalya, 2006; Yue and test is a rankbased method that has been applied Hashimoto, 2003). Following Burn et al. (2004), we have corrected the data for serial correlation through a modified version of percent chance for error exists in concluding that the Trend Free PreWhitening (TFPW) approach a trend is statistically significant when in fact no developed by Zhang et al. (2001) and Yue et al. trend exists.

(2002). The TFPW approach attempts to separate the serial correlation that arises from a linear trend from the original time series. This involves Rainfall-Runoff modeling estimating a monotonic trend for the series, Statistical tests examine rainfall and discharge removing this trend prior to PreWhitening the separately. Rainfall runoff models can establish, series and finally adding the monotonic trend if the relationship between rainfall and discharge back to the PreWhitened data series to remove has changed over time and may indicate the the serial correlation. underlying physical mechanisms if a c