Interating in-situ data with satelllite-derived products to assess surface-groundwater interactions and sustainability of groundwater resources in semi-arid environment
Mewcha Gebremedhin is a PhD student in the department of Water Resources. (Co)Promotors are dr. M.W. Lubczynski and dr. B.H.P. Maathuis from the faculty ITC and dr. D. Teka from Mekelle University, Ethiopia.
In arid and semi-arid regions, water scarcity is nowadays a primary challenge, because of continuously increasing spatio-temporal rainfall variability and high evapotranspiration, both implying a decline of freshwater resources. Moreover, it is expected that this problem will worsen with ongoing climate change. Therefore, in the Zamra Catchment (ZC), Northern Ethiopia, the recurrent droughts have resulted in a plan to utilize groundwater resources to support agricultural activities to ensure food security. Implementation of such policy requires in-depth investigations of the spatio-temporal surface-groundwater (SW-GW) interactions and sustainability of groundwater resources, which can be optimally realized through application of an integrated hydrological model (IHM). However, the poor coverage of ground-based hydro-meteorological gauging stations in the ZC- challenged that application. Therefore, this study proposes a novel approach to integrate satellite-derived products with in-situ measurements, providing RS-based input data for the ZC IHM focussed on assessment of spatio-temporal SW-GW interactions and on sustainability of groundwater resources. The research approach consists of four specific objectives (SO): 1) integrate daily satellite rainfall with in-situ rainfall measurements; 2) estimate rainfall interception loss (EI) at catchment scale from earth observation; 3) derive potential evapotranspiration (PET) from satellite based reference evapotranspiration (ETo) and land use land cover (LULC) factor (Kc); and 4) develop and calibrate ZC IHM using semi-continuous (gridded) soil moisture treated as additional, non-standard calibration state variable and assess the spatio-temporal variability of surface-groundwater interactions and sustainability of groundwater resources in the topographically and hydrologically complex system of ZC.
The first SO focusses on integration of daily satellite rainfall with in-situ rainfall. The CHIRPS and MPEG satellite rainfall products were downscaled to the grid size of the IHM and validated against in-situ observations. The downscaled products were then merged with in-situ observations applying Geographically Weighted Regression (GWR) considering local explanatory variable (topography), with the intention to improve the accuracy of the two rainfall products and also to evaluate them to select the better performing one. The descriptive statistics, categorical statistics and bias decomposition methods, including novel protocol proposed in this study with new bias indicators, were applied to evaluate the biases of the two rainfall products before and after bias-correction. The evaluation showed large biases in both downscaled products prior the bias-correction, but the GWR bias-correction approach, substantially improved the accuracy of the two products, with slightly better final accuracy of MPEG than of CHIRPS. The study demonstrated that the GWR approach, could substantially reduce the daily biases of the two satellite rainfall products (MPEG and CHIRPS) in topographically complex areas. Further validation and improvement of rainfall in the ZC, can be achieved by increasing in-situ gauge network and eventually considering more accuracy-effective explanatory variables.
The second SO focusses on the spatio-temporal estimate of EI at the ZC, characterized by various land use types. For that purpose, a RS solution of the Revised Gash Model (RGM) was applied on daily basis. As input data for RGM, hydrometeorological variables from earth observation satellite and in-situ, weather stations were used, while biophysical variables, exclusively from earth observation satellite. Before calculating the final EI, the RGM assumption permitting the use of rainfall as one storm per rainy day was tested. As that assumption turned to be valid, the daily, bias-corrected rainfall was used as the gross daily rainfall input of the RGM. The overall estimated EI demonstrated high spatial and temporal variability, ranging on annual basis from zero at bare lands to 30% in forested areas. This study emphasized the benefit of using RS solution of RGM, facilitated by in-situ data, as an optimal way of spatio-temporal quantification of EI, particularly suitable for large and data scarce-areas like the ZC. Further validation of the RS solution of the RGM can be done by installing more climatic stations and through simultaneous measurements of rainfall, throughfall and stemflow at different LULC types.
The third SO focusses on derivation of PET as the product of RS-based, FAO-Penman-Monteith ETo and Kc. Considering the ETo, the RS-based Daily Reference Evapotranspiration (DMETREF-ETos) was first validated using four ground based weather stations. The evaluation revealed significant underestimation of the DMETREF-ETos, especially in warm conditions. That underestimation was attributed to advection, which was corrected as function of near-surface air temperature obtained from ERA5-Land product, merged with in-situ air temperature. That bias-correction approach, substantially improved the ETo. Subsequently, the bias-corrected DMETREF-ETos was converted into PET by multiplying it by the spatio-temporally variable LULC factor (Kc), defined as linearly dependent on the NDVI obtained from Sentinel 2 high resolution Multi Spectral Instrument (MSI). The NDVI-based Kc demonstrated high spatio-temporal variability, ranging from ~0.15 (bare land) to ~1.4 (forest), which resulted in higher PET values than the bias-corrected DMETREF-ETos in locations where Kc >1 and vice versa (i.e. in locations where Kc <1). This part of study, underlined substantial difference between ETo and PET.
The fourth SO focusses on IHM assessment of the spatio-temporal variability of SW-GW interactions and on sustainability of groundwater resources in the hydrologically complex ZC. RS approaches to estimate rainfall, EI and PET were applied as the driving force inputs of the MODFLOW 6 IHM (MOD6-IHM) of the ZC. The 6 layer model was set up in transient, using daily stress periods, employing MODFLOW 6 ‘advanced packages’. The calibration was carried out using PEST++ throughout five hydrological years. Ground-based measurements of groundwater levels, streamflow and soil moisture were used as in-situ calibration state variables. Besides, a RS-based Global Surface Soil Moisture (GSSM), with semi-continuous (gridded) spatial coverage, was used as additional calibration state variable to better constrain the model. The calibration results presented good fit between simulated and observed state variables. Such well-calibrated MOD6-IHM allowed for the assessment of SW-GW interactions and sustainability of groundwater resources in the complex and dynamic hydrological system of the ZC.
The calibrated ZC model, showed high spatio-temporal water fluxes variability in the ZC, largely influenced by high spatio-temporal rainfall variability, in which the rainfall (P) was the only source of water input. The P over the five hydrological years investigated (01 October 2015 to 30 September 2020) averaged to 640 mm yr-1. Throughout the MOD6-IHM solution, that P was partitioned into the two dominant sinks, evapotranspiration (ET=53.2% of P) and stream outflow (q= 46% of P). Considering the ET, its largest part was lost by the unsaturated zone (ETu=62% of ET), then was followed by ET loss from saturated zone (ETg=26% of ET), while the remaining 12% was lost at the ground surface by interception of plants. Considering the q, the rejected infiltration (RI) contributed the largest part (80% of q), because of ZC steep hillslopes, dense drainage stream networks and erratic rainfall characteristics, while the groundwater exfiltration (Exfgw) accounted for only about 19% of q. The ETu and RI played the key role in reducing the gross recharge (Rg = 24.1% of P). The large contribution of ETg (56.9% of Rg) and modest of Exfgw (34.2% of Rg) resulted in the relatively small net groundwater recharge (Rnet= 6.5% of Rg). As the Rnet constrains groundwater resources sustainability, the issue of sustainability of ZC groundwater resources is crucial considering their future utilization for agricultural purposes.
