Hydrological performance evaluation of multiple satellite precipitation products in the upper Blue Nile basin, Ethiopia

Hydrological performance evaluation of multiple satellite precipitation products in the upper Blue Nile basin, Ethiopia

Journal of Hydrology: Regional Studies 27 (2020) 100664 Contents lists available at ScienceDirect Journal of Hydrology: Regional Studies journal hom...

3MB Sizes 2 Downloads 87 Views

Journal of Hydrology: Regional Studies 27 (2020) 100664

Contents lists available at ScienceDirect

Journal of Hydrology: Regional Studies journal homepage: www.elsevier.com/locate/ejrh

Hydrological performance evaluation of multiple satellite precipitation products in the upper Blue Nile basin, Ethiopia

T

Haileyesus Belay Lakewa,*, Semu Ayalew Mogesb, Dereje Hailu Asfawb a b

Ethiopian Space Science and Technology Institute, Addis Ababa, Ethiopia School of Civil and Environmental Engineering, Addis Ababa University Institute of Technology, P.O. Box 385, Addis Ababa, Ethiopia

A R T IC LE I N F O

ABS TRA CT

Keywords: CREST Blue Nile Hydrologic modeling Satellite and reanalysis Precipitation

Study region: Ethiopia, upper Blue Nile. Study focus: This study evaluates hydrological performance of multiple globally available precipitation products in the data scarce region of the upper Blue Nile basin over multi-scales (1656–199,812 km2) focusing on multi-year (2000–2012) for daily simulation. Grid based, fully distributed hydrological model Coupled Routing and Excess STorage (CREST) is calibrated using in-situ observed rainfall data. The evaluation compares five precipitation products of gauge adjusted Climate Prediction Center Morphing Technique (CMORPH), Tropical Rainfall Measuring Mission (TRMM) Multi-Satellite Precipitation Analysis 3B42 version 7 (TMPA), ERA-Interim (ERAI), Global Precipitation Climatology Centre (GPCC) and Multi-Source Weighted Ensemble Precipitation (MSWEP). New hydrological insights for the region: Performance results indicate that the MSWEP precipitation shows consistently better performance than from the rest of precipitation products. This global precipitation product can be a priori alternative source of data for various water resources applications in the upper Blue Nile basin for its predictive ability and the length of data availability which the product has an advantage over the satellite products. However, the product needs improvement to estimate the total annual flow volume of the observed flow for the three nested watersheds to perform close to the gauged rainfall. This by no means undermines the potential benefits of the rest products for different watersheds and time scale.

1. Introduction Hydrometeorological datasets have significant importance on the sustainable water resources management, as well as for different socio-economic activities. However, it is noted that acquiring accurate and consistent set of precipitation data is a challenging task throughout the world especially in Africa. The data assessment report of Africa-African Climate Policy Centre (ACPC, 2011) summarized as having poor spatial coverage with many stations became either non-operational or report poor quality data with missing information (Lakew et al., 2017). It is indicated that in some cases even the existing stations neither provide accurate observations nor quality controlled and transmitted to the database system. Hence, the importance of improved global data remains one of the future hopes for the design and management of water resources in data scarce parts of the world. The advent of global data sets such as satellite and reanalysis products play a great role in building a consistent set of hydrometeorological data that can be utilized for various purposes of water resources development and management in Africa and globally.



Corresponding author. E-mail addresses: [email protected] (H.B. Lakew), [email protected] (S.A. Moges), [email protected] (D.H. Asfaw).

https://doi.org/10.1016/j.ejrh.2020.100664 Received 22 August 2018; Received in revised form 3 January 2020; Accepted 9 January 2020 2214-5818/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

However, these globally available datasets have the limitations to estimate the observed rainfall and should be evaluated at the local basin scale for water resources applications. Evaluation of the amount of different reliable precipitation products and of its spatiotemporal distribution has significant role for scientific hydrological applications (Beck et al., 2017; Derin et al., 2019; Ehsan et al., 2019; Komma et al., 2007; Kucera et al., 2013; Lettenmaier et al., 2015). Different hydrological performance researches are conducted on the global satellite precipitation products at the upper Blue Nile at specific case studies in the basin. Bitew and Gebremichael (2011) evaluated the hydrological performance of CMORPH, satellite-only product TRMM TMPA 3B42RT, satellitegauge product TRMM TMPA 3B42 and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) for small size watersheds of Gilgel Abbay (1656 km2) and Koga (299 km2) in Abbay basin, Ethiopia. Applying a semi distributed hydrological model Soil and Water Assessment Tool (SWAT) calibrated from 2003 to 2004 and validated from 2006 to 2007. Granting to the subject area, the performance of the results of the simulations obtained from the 3B42RT and CMORPH products improves with the size of watershed area increases. Habib et al. (2014) also showed the improvements of the bias corrected CMORPH satellite data on the hydrological performance of Gilgel Abbay (1656 km2) watershed compared to the uncorrected CMORPH product. The products were simulated using semi distributed Hydrologiska Byråns Vattenbalansavdelning (HBV) hydrological model from 2003 to 2004. The above studies that were conducted in the upper Blue Nile basin, Ethiopia were carried out in small watersheds and for short simulation periods and they didn’t reflect the effect of the size of nested watersheds with different spatial scales for different global precipitation products on multi-year. This research mainly aims to evaluate multiple global precipitation products for their hydrological performance and their potentials for water resources applications in data scarce basins. The study is conducted in the upper Blue Nile basin over three nested watersheds (1656–199,812 km2) and multi-year (2000–2012) using a daily temporal scale. The multiple precipitation products included in the study are the gauge adjusted CMORPH, TRMM TMPA 3B42v7, ERAI, GPCC and MSWEP. The CREST fully distributed hydrological model (Shen and Hong, 2014) is used for the evaluation and comparision. 2. Data and method 2.1. The study area This study is conducted in the upper Blue Nile basin at three nested watersheds, Gilgel Abbay (small), Abbay at Kessie station (middle upper Blue Nile) and the upper Blue Nile at Eldiem (at the border of Ethiopia and Sudan). The descriptions of the watersheds are provided below. 2.1.1. Small scale (Gilgel abbay watershed) Gilgel Abbay watershed which has a drainage area of 1656 km2 is located in the upper part of the Abbay basin between 36°42′E–39°48′E and 9°15′N–12°45′N. The streamflow station is located upstream of Lake Tana before it joins the Lake (Fig. 1). The landscape is complex topography with elevation ranging from 1880 m to 3530 m. The watershed gets high amount of rainfall relative to the two medium and large scale case studies. Hence, this small scale watershed is located at the central part of the basin of the highlands, relatively high amount of rainfall is observed. 2.1.2. Medium scale (Kessie Station) Abbay at Kessie gauging station is located at the mid of the Abbay basin (including Gilgel Abbay watershed and Lake Tana) which has a drainage area of 65,784 km2, in the upper Blue Nile basin between 36°42′E–39°48′E and 9°15′N–12°45′N (Fig. 1). The landscape is complex topography with elevation range between 1102 m–4176 m. The medium scale covers more area of the lowlands and arid regions with small amount of rainfall, relative to the small scale of Gilgel Abbay. 2.1.3. Large scale (eldiem) Eldiem station is located at the outlet of the Abbay basin (boarder of Ethiopia and Sudan) and has a drainage area of 199,812 km2 (including Gilgel Abbay and Kessie) is the main part of the Ethiopian highlands between 34°24′E–39°48′E and 7°42′N–12°30′N. The landscape is complex topography together with plain (flatter) area on the North-West part and highly elevated complex mountains on the other parts of the basin with elevation range between 503 m–4176 m. The large scale consists more are coverage of the low lands relative to the small and medium scale case studies. High amount of rainfall at the central part of the basin and insignificant amount of rainfall is observed in wider area of the North-West part of the basin with low elevation and arid area coverage. The location of the three spatial scales and the rain gauge distribution is shown in Fig. 1 (Lakew et al., 2019). 2.2. Hydrometeorological data The daily stream flow of the three case studies and the daily rainfall historical data are collected from Ethiopia Ministry of Water, Irrigation and Electricity, and Ethiopian Meteorological Agency respectively. A total of 89 in-situ rainfall measuring stations are used for the three case studies. From the total, 5 stations are used for Gilgel Abbay, 51 stations for Kessie and 89 stations for Abbay at Eldiem station. As shown in Fig. 1, most of the available stations are located in the central and northern highlands, that have a relatively higher amount of rainfall records. The western part of the basin (the lowlands) has relatively lower rain gauge station coverage as shown in Fig. 1. 2

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

Fig. 1. Location of the three spatial scales and rain gauge distribution in the upper Blue Nile (Abbay) basin (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

The grid based (0.5°) with daily temporal scale of potential Evapotranspiration (PET, mm/day) that is derived from PenmanMonteith method (Allen et al., 1998) is used as an input to setup CREST model for the hydrologic simulations. The potential evapotranspiration of a given crop is defined as soil evaporation and plant transpiration under unlimited soil water supply and actual

3

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

meteorological conditions. It is an important agrometeorological parameter for climatological and hydrological studies, as well as for irrigation planning and management (Labedzki et al., 2011). The grid PET data is accessed from the eartH2Observe data portal (https://wci.earth2observe.eu/portal/). 2.3. Global gridded precipitation products In this study multiple precipitation products with daily temporal scale are considered for hydrological evaluation of daily simulation flow for different basin scale in the upper Blue Nile basin. The products are described below. Climate Prediction Centre's morphing technique (CMORPH) products are derived using the morphing technique that uses precipitation estimates derived from passive microwave (MW) observations and propagates these features in space using motion vectors derived from half-hourly geostationary satellite infrared data (Joyce et al., 2004). This is one of the satellite precipitation products available with different temporal and spatial scales. For this study the gauge adjusted with 0.25° grid resolution at daily temporal scales is considered Tropical Rainfall Measuring Mission (TRMM) Multi Satellite Precipitation Analysis (TMPA) post-real time research version product 3B42v7 uses MW data to calibrate the infrared (IR)-derived estimates and creates estimates that contain MW-derived rainfall estimates when and where MW data are available and the calibrated IR estimates where MW data are not available (Huffman et al., 2007). The same with the satellite product of CMORPH, the daily temporal scale with 0.25° spatial scale of TRMM is considered for this study. Like satellite precipitation products, reanalysis precipitation products are considered in the hydrological evaluation in this study. Reanalysis data provide a reconstruction of the historical global atmospheric circulation patterns through assimilation of available observation data into Global Climate Model (GCM) predictions. Thus, reanalysis data are widely considered to be a robust proxy of historic atmospheric observations that reproduce observed climate dynamics and are therefore useful for climate application studies where there are inadequate meteorological input data (Hwang et al., 2013). Reanalysis studies, that merge climate model predictions and observation data, have been undertaken to develop long-term historical spatially and temporally distributed climatic variables such as precipitation, temperature, solar radiation, humidity and wind speed. ERA-Interim (ERAI) is reanalysis precipitation product with 0.25° spatial scales. ERA-Interim/Land is a global land surface reanalysis data set covering the period 1979–2010. It describes the evolution of soil moisture, soil temperature and snowpack. ERAInterim/Land is the result of a single 32-year simulation with the latest ECMWF (European Centre for Medium-Range Weather Forecasts) land surface model driven by meteorological forcing from the ERA-Interim atmospheric reanalysis and precipitation adjustments based on monthly GPCP v2.1 (Global Precipitation Climatology Project). The horizontal resolution is about 80 km and the time frequency is 3 -hly (Balsamo et al., 2015; Dee et al., 2011). For this study the 3 -hly precipitation data are aggregated into daily for the daily flow computation. Global Precipitation Climatology Centre (GPCC) has been established in1989 on request of the World Meteorological Organization (WMO) and it is new global precipitation climatology version 1.1 and available from 1979 to 2015 (Beck et al., 2017; Schneider et al. (2015)). The product is available at different spatial scales (2.5° x 2.5°, 1.0° x 1.0°, 0.5° x 0.5°, and 0.25° x 0.25° resolution) with daily based estimates. For this study the finer with 0.25° spatial and daily temporal scale is considered. Multi Source Weighted Ensemble Precipitation (MSWEP) retrospective is a new fully global precipitation dataset (1979–2015) with a high 3 -hly temporal and 0.25° spatial resolutions (Beck et al., 2017). The dataset is unique in that it takes advantage of a wide range of data sources, including gauges, satellites, and atmospheric reanalysis models, to obtain the best possible precipitation estimates at global scale. For this study the 3 -hly precipitation data are aggregated into a daily temporal scale for the simulation of daily flow. 2.4. CREST fully distributed hydrological model The Coupled Routing and Excess Storage model (Shen and Hong, 2014) is a distributed hydrological model developed to simulate the spatial and temporal variation of land surface, and subsurface water fluxes and storages by cell-to-cell simulation. CREST’s distinguishing characteristics include: (1) distributed rainfall–runoff generation and cell-to-cell routing; (2) coupled runoff generation and routing; (3) representation of sub-grid cell variability of soil moisture storage capacity and sub-grid cell routing (via linear reservoirs). The coupling between the runoff generation and routing mechanisms allows detailed and realistic treatment of hydrological variables such as soil moisture. Furthermore, the representation of soil moisture variability and routing processes at the sub-grid scale enables the CREST model to be readily scalable to multi-scale modelling research. In this study, CREST v2.1 which involves an updated routing scheme based on fully distributed linear reservoir (FDLRR) scheme (Shen and Hong, 2014) has been implemented. 2.4.1. Model setup The requirements of data to setup the CREST hydrological model is discussed below. Two types of datasets are required i) the geospatial input data which are static in nature and ii) the hydrometeorological input data which are dynamic time series in nature. The geospatial data with the resolution of 1 km by 1 km resolution Digital Elevation Model (DEM) is extracted from the United States Geological Survey (USGS) National Elevation Dataset (NED). The DEM is used to derive the geospatial data for CREST model such as slope, flow direction and flow accumulation and drainage area of the case studies. The second input consists of the dynamic 4

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

Table 1 Estimated parameter values of independent calibration for different case studies. Parameter

Description

RainFact Ksat MW B IM KE coeM expM coeR

The The The The The The The The The The The The The The

coeS KS KI

Unit

multiplier on the precipitation field Soil saturate hydraulic conductivity Mean Water Capacity exponent of the variable infiltration Curve impervious area ratio factor to convert the PET to local actual overland runoff velocity coefficient overland flow speed exponent multiplier used to convert overland flow speed to channel flow speed multiplier used to convert overland flow speed to interflow flow speed overland reservoir Discharge Parameter interflow Reservoir Discharge Parameter

Multi-Scale case studies Gilgel Abbay

Kessie

Eldiem

—— mm/day mm —— —— —— —— —— ——

1.4 143.3 260.9 0.75 0.016 1.19 25.36 0.270 1.94

0.76 1777.5 470.6 1.09 0.167 0.82 135.75 0.46 2.93

0.67 1514.9 360.6 0.87 0.067 1.02 6.63 0.11 2.12

——

0.54

0.029

0.74

—— ——

0.31 0.18

0.37 0.27

0.58 0.17

hydrometeorological datasets. The meteorological datasets (rainfall & PET) are geo-referenced with the geographic location of the upper Blue Nile and gridded to the model grid resolution that is consistent with the satellite and reanalysis resolution of 0.25°. The process of geo-referencing and gridding is carried out using matlab. Daily time series observed streamflow (hydrological) dataset is used at the outlet of the watersheds to calibrate model parameters. 2.4.2. Calibration and parameters Many of the parameters in the CREST model are physical in nature and can in principle be estimated from field surveyed data, such as soil texture, land cover maps and vegetation coverage. However, in many large basins worldwide and particularly in the African context, it is much more difficult to estimate model parameters from ancillary data. Models need to be trained through calibrated either manually, automatically, or using combined approaches given observations of rainfall and streamflow (Wang et al., 2011). CREST has 12 parameters and the model is calibrated for the three case studies using daily in-situ gauge observed rainfall, potential evapotranspiration (PET), and observed streamflow data. Calibration is done from 2000 to 2007 and validated from 2008 to 2012. Two step approach is used to calibrate the CREST model. Manual calibration of multi-parameter model is time consuming and less practical, First, automatic calibration has been implemented to estimate the model parameters. CREST uses, Shuffle Complex Evolution University of Arizona (SCE-UA) automatic kernel algorithm for calibration of the model using in-situ gauged rainfall. CREST directly uses the screen output of SCE-UA codes in matlab written by (Duan et al., 1992). The objective function value is the NSCE of each simulation. The automatic calibration results are shown in Table 1. Second, sensitivity analysis is done to identify the most sensitive (or dominant) parameter. In our previous paper (Lakew et al., 2017), the RainFact parameter which is the multiplier on the precipitation field is identified as the most dominant parameter controlling the accuracy of the model performance. The parameters that are calibrated using the gauged rainfall are used for the simulation forcing the global precipitation products. The ‘RainFact’ parameter is tweaked for each global precipitation product for the performance improvement of each watershed and precipitation products. 2.5. Performance in discharge simulation The performance of the model for the various precipitation products and catchments is evaluated using statistical comparisons between observed flow (O) and simulated flow (C) on the daily temporal resolution scale. The Q-Q plot is also used as additional visual comparison tool. First, for statistical goodness of fit of simulated flows, we utilized the Nash-Sutcliffe coefficient of efficiency (Nash and Sutcliffe, 1970):

NSCE = 1 −

∑ (Qi, o − Qi, c )2 ∑ (Qi, o − Q¯o )2

(1)

Where Qi, o is the observed discharge of the ith day; Qi, c is the simulated discharge of ith day; and Q¯O the average of all the daily observed discharge values. If NSCE ≤ 0, then the model provides no skill in relation to using the observed mean as a predictor and higher values indicating better agreement.

CC =

∑ (Qi, o − Q¯o )(Qi, c − Q¯c ) ∑ (Qi, o − Q¯o )2 (Qi, c − Q¯c )2

(2)

Second, the Pearson Correlation Coefficient (CC) is used to assess the agreement between simulated and observed discharge as follows:Where Q¯C the average of all daily simulated discharge values. 5

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

Fig. 2. Mean annual rainfall of (a) Gilgel Abbay (b) Kessie and (c) Eldiem case studies.

Third, relative bias ratio assesses the systematic bias of the simulated discharge:

Bias =

∑ Qi, c − ∑ Qi, o *100% ∑ Qi, o

(3)

The best skill occurs with NSCE = 1, CC = 1, and Bias = 0 %.

3. Result and discussion 3.1. Rainfall analysis The areal average global precipitation products used in this study are compared with the in-situ gauged rainfall data (Fig. 2). The areal mean annual observed rainfall of Gilgel Abbay, Kessie and Eldiem watersheds for the study period (2000–2012) is 1750, 1100 and 1400 mmyr−1 respectively. In comparison to the in-situ gauged rainfall, most of the global precipitation products exhibit good agreement with gauged data except the ERAI product which overestimates the observed rainfall for the three case studies. It is also noted that the accuracy of the areal average precipitation (except ERAI) slightly increases as a function of the size of the watershed and the magnitude of the areal mean annual observed rainfall decreases as the basin scale increases. This is due to; as the basin size increases it has the tendency to incorporate more lowlands that have insignificant amount of observed rainfall compared with the highlands and vice versa.

6

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

Fig. 3. Performance of various precipitation products of the Gilgel Abbay watershed.

3.2. Simulated runoff analysis Runoff simulations are carried out for multiple global precipitation products of gauge-adjusted CMORPH, (TRMM) TMPA 3B42v7, ERAI, MSWEP and GPCC described in global gridded precipitation products section. The model is calibrated using in-situ gauge observed daily rainfall data for 8 years period (2000–2007) and validated for 5 years (2008–2012). The final list of parameters is shown in Table 1. For illustration the focus is on the RainFact parameter (sensitive) which is the multiplier on the precipitation field and the adjustment factor of the precipitation either due to canopy interception or bias. Increasing the value of the RainFact result in increasing runoff and vice versa, and has value of greater than zero. If the parameter value is greater than one, it shows that the average areal rainfall amount that is taken for that watershed underestimate the actual average areal rainfall of the region to capture the observed flow and vice versa. The effect of the parameter shows that as the scale of the watershed increases the RainFact parameter value decreases from 1.4 to 0.67. As the basin scale increases, it has the tendency to incorporate more lowlands and, in this study more gauging station are taken from highlands that have a high amount of rainfall. This makes, as the basin scale increases, the tendency to overestimate the actual rainfall of the region increases. To correct this effect CREST uses the RainFact value less than one of 0.67 for Eldiem (large scale) and 0.76 for Kessie (medium scale) that have relatively large and medium lowland coverage respectively. Gilgel Abbay watershed has well distributed rain-gauges that are set up in the central highlands of the basin, the RainFact parameter is expected to be close to 1. However, the parameter is 1.4 which is more than the expected value. This indicates that there is a high groundwater contribution in the streamflow of Gilgel Abbay watershed that makes the RainFact parameter value greater than 1. Melesse and Abtew (2015) also investigated that the hydrological process governing runoff generation in Gilgel Abbay is dominated by saturation access and base flow contribution comes from the groundwater. The performance of the simulated flow forcing the multiple global precipitation for Gilgel Abbay watershed is shown in Fig. 3. Gilgel Abbay watershed that has relatively very good rain gauge coverage, the simulated flows from different products perform very well in the NSCE evaluation (from 0.70 to 0.81) at calibration period. Especially MSWEP shows outstanding performance of NSCE = 0.81, and GPCC also shows very good performance of NSCE = 0.80. The satellite precipitation product of CMORH (0.75) and TRMM (0.70) shows also encouraging performance with slightly reduced performance from MSWEP and GPCC products as shown in Fig. 3. This shows that the newly released products of MSWEP and GPCC show significant improvement over the satellite products to fit with 7

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

the observed runoff of Gilgel Abbay watershed that has relatively good rainfall coverage with high amount of records in the calibration period. In the validation period, the global precipitation products perform poorly relative to the calibration period together with the gauged rainfall. This may be ascribable to the gauging station located near the bridge along the main road from Bahir Dar to Addis Ababa is shifted hundreds meters downstream from the original site due to the road construction after 2006 and the rating curve is not updated yet. Based on the bias (%) performance evaluation in the Gilgel Abbay watershed, the simulated flows from all products including satellite and reanalysis products underestimates the annual volume of the observed flow. Relatively CMORPH shows the minimum bias value of (-4.8 %) in the calibration period from the rest products. This indicates that the simulated flow of satellite precipitation product of CMORPH outperforms the rest global products including MSWEP (-18.9 %) and GPCC (-18.6 %) based on bias performance evaluation in the Gilgel Abbay watershed. The validation period bias evaluation of the global products shows poor performance, but still CMORPH shows relatively better performance from the rest products. This indicate that the newly released products (MSWEP & GPCC) have the limitation to capture the total annual flow volume for the small scale of the basin (Gilgel Abbay) relative to the satellite products. The Correlation Coefficient (CC) evaluation manifests that each product shows very good coefficient performance above 0.84 for both calibration and validation periods. Especially, MSWEP shows the highest of 0.92 in the calibration period. These all show that based on (NSCE & CC) statistical metrics performance measurement, MSWEP product shows high performance over the rest precipitation products, and it can be an alternative data source for water resources applications in the Gilgel Abbay watershed. However, based on bias evaluation, it still requires improvement to capture the total observed annual flow volume of the Gilgel Abbay watershed relative to the gauged rainfall and even from the satellite product of CMORPH. The medium spatial scale of this study (Kessie) has the effect of Lake Tana. The lake is located at some upstream distance from the streamflow gauging station of Kessie. The daily performance of NSCE for the simulated flows for the entire products show less performance (NCSE < 0.73) at Kessie relative to the Gilgel Abbay watershed in the calibration period. This is imputable to the effect of Lake Tana that has a reservoir surface area ∼3500 square km located at the upstream of Kessie streamflow gauging station. The fully distributed hydrological model CREST has the limitation to capture watersheds with the effect of reservoirs and doesn’t incorporate the simulation of the reservoir in to the cell to cell routing. For medium scale basin (Kessie, to the effect of Lake Tana), similar to Gilgel Abbay the simulated flow of the recently released products MSWEP (0.72) and GPCC (0.71) show better NSCE performance over the satellite products of CMORPH (0.62) and TRMM

Fig. 4. Performance of various precipitation products of Kessie. 8

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

(0.67) in the calibration period as shown in Fig. 4. For validation period except ERAI, the global products show competitive NSCE performance ranges from 0.60 to 0.64. The product ERAI shows the worst skill performance of NSCE with a negative value for both calibration and validation periods. Relative to the small scale case study, the medium scale of Kessie NSCE performance of the entire global products show relatively reduced performance that is due to the lake effect on the performance of the distributed model CREST. The bias performance of Kessie is similar with Gilgel Abbay watershed. The satellite driven simulation flow from CMORPH manifests the minimum bias value of -16 (%) in calibration and -22 (%) in validation periods. However, attributable to the effect of lake Tana, the bias value in calibration period shows high value relative to Gilgel Abbay. Unlike the calibration period, the validation period bias performance for Kessie shows better performance relative to Gilgel Abbay watersheds. As we discussed in the Gilgel Abbay performance, the validation period shows poor performance due to the gauging station was moved from the original position onwards late of 2006. The medium scale of Kessie station has the same Correlation Coefficient (CC) efficiency with the Gilgel Abbay. It has very good performance of above 0.77 for the entire products for both calibration and validation periods. The product from MSWEP shows the best correlation coefficient of 0.87 and 0.80 for calibration and validation periods respectively. The product form MSWEP shows outstanding performance from the rest precipitation products and also yields consistent performance with the small scale watershed of Gilgel Abbay based on (NSCE and CC) statistical metrics. This indicates that MSWEP is also applicable for water resources application for medium scale of data scarce region in the upper Blue Nile basin, Ethiopia. However, similar to the small scale of Gilgel Abbay, the medium scale of Kessie the MSWEP bias performance needs improvement to capture the total volume annual flow compared with the gauged rainfall and even from the satellite product of CMORPH. Both satellite and reanalysis (Except ERAI) precipitation products show excellent performance (NSCE > 0.75) in large scale (Eldiem) compared with small and medium scale of the upper Blue Nile basin in both calibration and validation periods. This is because that (i) the distributed hydrological model CREST performance increases as the watershed area increases. As distributed hydrological model, the model error decreases with the increase of the drainage area. (ii) Eldiem station is located far from the lake Tana reservoir effect in which the reservoir has insignificant effect for the measured flow. (iii) it has consistent flow data with better quality records for both calibration and validation periods unlike Gilgel Abbay. These reasons make the global precipitation product perform well from the small and medium scales. Based on NCSE performance evaluation, the MSWEP shows the highest performance of NSCE = 0.89 for both calibration and validation periods for the large scale of Eldiem as shown in Fig. 5. ERAI shows encouraging skill to capture the observed flow with NSCE value of 0.57 and 0.81 for calibration and validation

Fig. 5. Performance of various precipitation products of Eldiem. 9

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

periods respectively. Resembling the other case studies of small and medium scale basin, minimum bias value is observed in satellite driven simulated flow of TRMM (-9.9 %) and CMORPH (-11.2 %) over the rest products in large basin scale (Eldiem) in the calibration period. The same with the previous small and medium scale case studies, the bias statistic metric indicate that MSWEP product needs improvement to capture the total annual flow volume of the large basin scale (Eldiem) in the upper Blue Nile. High CC metric performance is observed at Eldiem for both calibration and validation periods, especially MSWEP shows the best CC performance of 0.96 for both calibration and validation periods. The same with the small and medium spatial scale case studies, MSWEP shows high performance metrics of NCSE and CC. This shows, the product of MSWEP is a priori alternative data sources for water resources applications for the large basin scale of Eldiem, if precipitation adjustment is introduced to reduce the bias and to perform like the gauged rainfall. The result of this study shows that the simulated flow forcing with both satellite and reanalysis products perform well to capture the observed flow. However, the product from ERAI shows poor skill and needs improvement to perform like the other products. The newly released product of MSWEP shows consistent and very good performance based on NSCE and CC performance evaluation for all three case studies for both calibration and validation periods in the upper Blue Nile basin. The bias values reveal that the simulated flows have negative values, indicates that all products have limitation to capture the peaks and underestimate the total volume of observed flows. Satellite products show minimum bias value and better performance to capture the observed flow over the rest for both calibration and validation periods. The newly released products (MSWEP and GPCC) don't show any improvement regarding on bias performance to estimate the total volume of the runoff over the satellite products for different spatial scale case studies in the upper Blue Nile basin. Generally, the MSWEP product perform very well and shows outstanding performance from the rest products including satellite and reanalysis precipitation products. The rest global precipitation (except MSWEP) datasets do not take full advantage of the complementary nature of satellite and reanalysis data. The product of MSWEP specifically designed for hydrological modeling. The design philosophy of MSWEP was to optimally merge the highest quality precipitation data sources available as a function of timescale and location. 3.2.1. Quantile-Quantile evaluation The Quantile-Quantile (Q-Q) plot is a graphical technique for determining if two data sets come from populations with a common distribution. A 45-degree reference line is also plotted. If the two sets come from a population with the same distribution, the points should fall approximately along this reference line (Wilk and Gnanadesikan, 1968). The greater the departure from this reference line, the greater the evidence for the conclusion that the two data sets come from populations with different distributions or with the same distribution but biased or shifted with the mean. Fig. 6 shows the Q-Q plots of the simulated flow of multiple satellite and reanalysis precipitation products with a reference line of the observed flow for the three case studies in upper Blue Nile basin. The result of the Q-Q plot of the simulation mode for small scale basin study of Gilgel Abbay watershed shows that the global precipitation products try to capture very well the observed flow including ERAI, especially the satellite product of CMORPH shows good agreement with the reference line next to the simulated flow from gauged rainfall as shown in Fig. 6(a). For medium spatial case study (Kessie), with the exception of ERAI, the whole global datasets of including the reanalysis products show good performance to capture the reference line as shown in Fig. 6(b). Especially, MSWEP and GPCC products show high agreement with the reference line even from the gauged rainfall and the rest products to estimate the quantiles of the observed flow. The Q-Q plot of the simulation mode for the large spatial scale of Eldiem manifests that the same with the Kessie station, the whole global precipitation products (except ERAI) perform very well to estimate the quantiles, as well the products from MSWEP and GPCC shows the better agreement with reference line. The result of Q-Q plots of this study confirms that the global precipitation products MSWEP and GPCC products perform well. ERAI shows poor performance for all case studies to estimate the quantiles of observed flow and it needs some improvement to perform like the other global precipitation products that are considered in this study. Generally, MSWEP and GPCC show consistent and higher performance to estimate the quantile observed flow of each case study over the other products. Especially, MSWEP can be an alternative source of data to estimate the quantile flow for water resources planning and management in data scarce areas of Africa. 4. Conclusion The main purpose of this study is comparison and evaluation of globally available different precipitation products over a multiscale (1656–199,812 km2) and multi-year (2000–2012) over three nested watersheds in the upper Blue Nile basin (Abbay) using a fully distributed hydrological model CREST. The watersheds included are Gilgel Abbay (1656 km2), Abbay at Kessie (65,784 km2) and Abbay at Eldiem (199,812 km2). Two satellite precipitation products of gauge adjusted CMORPH, and TRMM TMPA 3B42v7, as well three different precipitation products of ERAI, GPCC and MSWEP with 0.25° spatial resolution are evaluated for daily simulated flow. The comparison is made between the simulated flow using rain gauge calibrated parameters forcing to the global precipitation products. Only tweaking the sensitive parameter of ‘RainFact’ is carried out, which is the rainfall adjustment factor on the field to improve the performance of the product for water resources application. The overall performance evaluation of the statistical metrics NSCE, CC and the graphical technique of Q-Q plot show that the products from MSWEP manifests consistent and higher performance for the nested watersheds with high improvement from the satellite products of TRMM and CMORPH as well from the rest precipitation products. We can conclude that MSWEP reanalysis precipitation product can be a prior alternative source of data for various water resources applications in the Blue Nile basin for its 10

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

Fig. 6. Quantile-Quantile plots of (a) Gilgel Abbay (b) Kessie & (c) Eldiem stations.

predictive ability and the length of data availability which the product has advantage over the satellite products. However, based on the Bias (%) statistical metrics evaluation, the product needs improvement to estimate the total annual flow volume of the observed flow for the three nested watersheds to perform close to the gauged rainfall and even from the satellite precipitation products of CMORPH and TRMM. This study has considered only three different sized watersheds for the analysis. Further investigation is desirable to consider more watersheds on the basin with high quality of datasets. Declaration of Competing Interest The authors declare no conflict of interest. Acknowledgements The authors would like to thank the Ethiopian Ministry of Water, Irrigation and Electricity for the updated observed river runoff data. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.ejrh.2020. 100664. References (ACPC), 2011. Climate Change and Water in Africa: Analysis of Knowledge Gaps and Needs. African Clim. Policy Cent., Addis Ababa, Ethiopa. Balsamo, G., Albergel, C., Beljaars, A., Boussetta, S., Brun, E., Cloke, H., Dee, D., Dutra, E., Muñoz-Sabater, J., Pappenberger, F., de Rosnay, P., Stockdale, T., Vitart, F., 2015. ERA-Interim/Land: a global land surface reanalysis data set. Hydrol. Earth Syst. Sci. Discuss. 19, 389–407. https://doi.org/10.5194/hess-19-389-2015.

11

Journal of Hydrology: Regional Studies 27 (2020) 100664

H.B. Lakew, et al.

Beck, H.E., van Dijk, A.I.J.M., Levizzani, V., Schellekens, J., Miralles, D.G., Martens, B., de Roo, A., 2017. MSWEP: 3-hourly 0.25° global gridded precipitation (1979–2015) by merging gauge, satellite, and reanalysis data. Hydrol. Earth Syst. Sci. Discuss. 21, 589–615. https://doi.org/10.5194/hess-21-589-2017. Bitew, M.M., Gebremichael, M., 2011. Assessment of satellite rainfall products for streamflow simulation in medium watersheds of the Ethiopian highlands. Hydrol. Earth Syst. Sci. Discuss. 15, 1147–1155. https://doi.org/10.5194/hess-15-1147-2011. Dee, D.P., Uppala, S.M., Simmons, A.J., Berrisford, P., Poli, P., Kobayashi, S., Andrae, U., Balmaseda, M.A., Balsamo, G., Bauer, P., Bechtold, P., Beljaars, A.C.M., van de Berg, L., Bidlot, J., Bormann, N., Delsol, C., Dragani, R., Fuentes, M., Geer, A.J., Haimberger, L., Healy, S.B., Hersbach, H., Hólm, E.V., Isaksen, L., Kållberg, P., Köhler, M., Matricardi, M., McNally, A.P., Monge-Sanz, B.M., Morcrette, J.-J., Park, B.-K., Peubey, C., de Rosnay, P., Tavolato, C., Thépaut, J.-N., Vitart, F., 2011. The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc. 137, 553–597. https://doi.org/10.1002/qj. 828. Derin, Y., Anagnostou, E., Berne, A., Borga, M., Boudevillain, B., Buytaert, W., Chang, C.-H., Chen, H., Delrieu, G., Hsu, C.Y., Lavado-Casimiro, W., Manz, B., Moges, S., Nikolopoulos, I.E., Sahlu, D., Salerno, F., Rodríguez-Sánchez, J.-P., Vergara, J.H., Yilmaz, K.K., 2019. Evaluation of GPM-era global satellite precipitation products over multiple complex terrain regions. Remote Sens. 11 (24), 2936. https://doi.org/10.3390/rs11242936. Duan, Q., Sorooshian, S., Gupta, V., 1992. Effective and efficient global optimization for conceptual rainfall-runoff models. Water Resour. Res. 28, 1015–1031. https:// doi.org/10.1029/91WR02985. Ehsan, M.A., Nikolopoulos, E.I., Anagnostou, E.N., 2019. Machine learning–based blending of satellite and reanalysis precipitation datasets: a multiregional tropical complex terrain evaluation. J. Hydrometeorol. 20, 2147–2161. https://doi.org/10.1175/JHM-D-19-0073.1. Habib, E., Haile, A., Sazib, N., Zhang, Y., Rientjes, T., 2014. Effect of Bias correction of satellite-rainfall estimates on runoff simulations at the source of the upper blue nile. Remote Sens. (Basel) 6, 6688. https://doi.org/10.3390/rs6076688. Huffman, G.J., Bolvin, D.T., Nelkin, E.J., Wolff, D.B., Adler, R.F., Gu, G., Hong, Y., Bowman, K.P., Stocker, E.F., 2007. The TRMM multisatellite precipitation analysis (TMPA): quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeorol. 8, 38–55. https://doi.org/10.1175/jhm560.1. Hwang, S., Graham, W.D., Adams, A., Geurink, J., 2013. Assessment of the utility of dynamically-downscaled regional reanalysis data to predict streamflow in west central Florida using an integrated hydrologic model. Reg. Environ. Chang. 13, 69–80. https://doi.org/10.1007/s10113-013-0406-x. Joyce, R.J., Janowiak, J.E., Arkin, P.A., Xie, P., 2004. CMORPH: a method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeorol. 5, 487–503. https://doi.org/10.1175/1525-7541(2004)005<0487:camtpg>2.0.co;2. Komma, J., Reszler, C., Blöschl, G., Haiden, T., 2007. Ensemble prediction of floods – catchment non-linearity and forecast probabilities. Nat. Hazards Earth Syst. Sci. Discuss. 7, 431–444. https://doi.org/10.5194/nhess-7-431-2007. Kucera, P.A., Ebert, E.E., Turk, F.J., Levizzani, V., Kirschbaum, D., Tapiador, F.J., Loew, A., Borsche, M., 2013. Precipitation from space: advancing earth system science. Bull. Am. Meteorol. Soc. 94, 365–375. https://doi.org/10.1175/bams-d-11-00171.1. Labedzki, L., Kanecka-Geszke, E., Bąk, B., Słowińska, S., 2011. Estimation of Reference Evapotranspiration Using the FAO Penman-monteith Method for Climatic Conditions of Poland. https://doi.org/10.13140/2.1.5080.4802. Lakew, H.B., Moges, S.A., Anagnostou, E.N., Nikolopoulos, E.I., Asfaw, D.H., 2019. Evaluation of global water resources reanalysis runoff products for local water resources applications: case study-upper blue nile basin of Ethiopia. Water Resour. Manag. https://doi.org/10.1007/s11269-019-2190-y. Lakew, H.B., Moges, S.A., Asfaw, D.H., 2017. Hydrological evaluation of satellite and reanalysis precipitation products in the Upper Blue Nile Basin: a case study of gilgel abbay. Hydrology 4, 39. https://doi.org/10.3390/hydrology4030039. Lettenmaier, D.P., Alsdorf, D., Dozier, J., Huffman, G.J., Pan, M., Wood, E.F., 2015. Inroads of remote sensing into hydrologic science during the WRR era. Water Resour. Res. 51, 7309–7342. https://doi.org/10.1002/2015WR017616. Melesse, A.M., Abtew, W., 2015. Landscape Dynamics, Soils and Hydrological Processes in Varied Climates. Springer International Publishinghttps://doi.org/10.1007/ 978-3-319-18787-7. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I — a discussion of principles. J. Hydrol. (Amst) 10, 282–290. https://doi.org/ 10.1016/0022-1694(70)90255-6. Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Evapotranspiration - Guidelines for Computing Crop Water Requirements - FAO Irrigation and Drainage Paper 56. FAO Irrig. Drain. Pap. 56. Rome, Italy Food Agric. Organ, United Nations ISBN 92-5-104219-5. Schneider, U., Ziese, M., Becker, A., Meyer-Christoffer, A., Finger, P., 2015. Global Precipitation Analysis Products of the GPCC. Glob. Precip. Climatol. Cent, Dtsch. Wetterdienst, Offenbach a. M., Ger. Wang, J., Hong, Y., Li, L., Gourley, J.J., Khan, S.I., Yilmaz, K.K., Adler, R.F., Policelli, F.S., Habib, S., Irwn, D., Limaye, A.S., Korme, T., Okello, L., 2011. The coupled routing and excess storage (CREST) distributed hydrological model. Hydrol. Sci. J. Des Sci. Hydrol. 56, 84–98. https://doi.org/10.1080/02626667.2010.543087. Wilk, M.B., Gnanadesikan, R., 1968. Probability plotting methods for the analysis for the analysis of data. Biometrika 55, 1–17. https://doi.org/10.1093/biomet/55. 1.1. Shen, Xinyi, Hong, Y., 2014. CREST Coupled Routing and Excess STorage CREST User Manual v2.1. Univ. Oklahoma HyDROS Lab http//hydro.ou.edu.

12