PhD Defence Moiteela Lekula

multiple data sources and integrated hydrological modeling for groundwater assessment in the central kalahari basin

Moiteela Lekula is a PhD student in the department of Water Resources. His supervisor is M.W. Lubczynski from the faculty of Geo-Information Science and Earth Observation.  

Groundwater resources in arid and semi-arid regions is often the only, but vulnerable source of potable water; therefore its reliable evaluation and management is critically important. As such, it is typically done through integrated hydrological models (IHM), considered nowadays as most reliable simulators because they dynamically integrate surface and groundwater fluxes. Once properly calibrated, they allow to predict aquifer behaviour in response to groundwater abstraction, climatic and/or land use changes etc. However, the reliability of the IHMs, as of any other models, is constrained by availability and quality of model input data, by realism of hydrogeological conceptual model and by adequate numerical model setup and calibration.

Data scarcity, particularly in arid to semi-arid Developing Countries, has always been hampering development of IHMs. Advancement in IHMs, made remote sensing (RS) data an alternative to the scarce, ground-based data sets. Integration of these RS-data, together with other available data sets in IHMs, is thus vital in water resource management in arid and semi-arid regions, hence, also in the Central Kalahari Basin (CKB), hosting the most productive and exploitable transboundary Karoo System Aquifer in Botswana and Namibia. As such, the CKB became an interesting and strategic study area considering its promising hydrogeological characteristics and importance of groundwater resources. The main objective of this PhD research was therefore the assessment of groundwater resources in the CKB using multiple data sources and integrated hydrological modelling.

The CKB study area, with its geological and hydrogeological characteristics were described in details in Chapter 2. These characteristics are particularly important because they were used as the basis to define hydrogeological conceptual model of the CKB.

The development of the hydrogeological conceptual model (HCM) is presented in Chapter 3. The HCM was developed with help of 3-D geological modelling code called Rockworks, operated in iterative combination with ArcGIS. As a result, a six-layer model was defined. A characteristic feature of the CKB and of the model, is its first, thick Kalahari Sand layer (~60 m), that restricts the erratic groundwater recharge to <1 mm yr-1 in the centre of CKB and to about 5-10 mm yr-1 in the eastern fringe. That layer is the only one in the system, which covers the entire study area. The other five layers are non-uniformly distributed, having variable thickness. Among them there are three aquifers (locally divided by semi-permeable layers), all having similar, radially-concentric regional groundwater flow patterns, all directed towards the same discharge area of Makgadikgadi Pans. That similarity is likely due to various layer-to-layer, hydraulic interconnections. The iterative use of RockWorks code with ArcGIS turned out to be relatively simple and efficient solution for  developing HCM of a large and complex multi-layered aquifer system such as the CKB and also for interfacing it with the numerical model. The developed HCM was the basis for the setup of the numerical model of the CKB.

Rainfall, addressed in Chapter 4, is the main driving force of IHMs. Most of IHMs require spatial data coverage of daily rainfall. It used to be defined by interpolation of rain gauge data, which usually have limited spatial coverage, especially in arid to semi-arid areas of developing countries such as the CKB area. An alternative source of rainfall data for IHMs, are satellite-based rainfall estimates (SREs). A complication is that different SREs, perform differently in different parts of the world, hence at each study area, various SREs need to be investigated to find the best performing one and optimize it to be used in hydrological studies. The Chapter 4 presents detailed evaluation of daily rainfall detection capabilities of four SREs in the CKB, i.e. FEWSNET RFE 2.0 with ~11 km spatial resolution, TRMM-3B42 v7 with ~27 km spatial resolution, CMORPH v1 with 8 km spatial resolution and CMORPH v1with ~27 km spatial resolution. The results showed that FEWSNET RFE 2.0 had best daily rainfall detection capability and therefore was further used as input of the IHM. The analysis of the four SREs highlighted also the high spatial and temporal variability of rainfall in the CKB. The overall rainfall assessment confirmed usefulness of using SREs in areas where rain gauges are scarce, such as the CKB.

A six-layer IHM of the CKB, calibrated in transient on daily basis, throughout a 13-year period (2002-2014) was developed and is presented in Chapter 5. For that purpose, MODFLOW-NWT code with active UZF1 package capable to simulate variably saturated flow, was used. In that model setup, various RS products coupled with long term, in-situ monitoring data were used as inputs. The calibrated model showed a good match between the simulated and measured temporal head patterns, with low MAE ranging from 0.02 to 2.70 m and low RMSE from 0.02 to 3.13 m. The 13-year model simulation confirmed steep water table declines in the major wellfields and gentler declines in the areas outside wellfields’ influences. The declines outside wellfields’ influences, were due to the generally lower gross recharge (Rg) than groundwater discharge, the latter consisting mainly of groundwater evapotranspiration (ETg) and implying generally negative, yearly net recharge (Rn). The large ETg was due to groundwater uptake by deep-rooted and ‘thirsty’ acacia trees and possibly also due to direct evaporation from the water table. The positive Rn occurred only twice in the 13 year simulation period, i.e. in 2006 and in 2014, when rainfall was 664 and 606 mm yr-1, resulting in Rn of 3.4 and 1.0 mm yr-1, respectively. The IHM results showed also large spatial variability of the Rg, ETg and Rn, all attributed mainly to: i) large spatio-temporal variability of rainfall; ii) presence of local surface morphological features such as dry river channels or local pans; iii) variable thickness of the unsaturated zone; and iv) variable vegetation density and species diversity. The proposed MODFLOW-NWT model solution with active UZF1 package and remote sensing data integrated with ground-based data as model input, showed to be optimal modelling solution for the study area such as the CKB.

Chapter 6 presents the synthesis of the PhD thesis while in chapter 7 recommendations are formulated.