Student Theses
Several students have completed or are completing their theses as part of this particular project. Below if a list of the work completed thus far or that is in progress
William Kristian Kalum (2024)
Estimation of recharge using WTFM and random forest in Hout-Sand catchment, South Africa
Groundwater recharge estimations are of crucial importance for the management of groundwater resources. Recharge for 105 wells was calculated using the Water table fluctuation method in the Hout-Sand catchment in South Africa. In total, 1721 recharge values were calculated. The rechange value ranges from being between the hydrology years of 1955 to 2023. Multiple
random forest (RF) models were set up for the catchment. A final RF model was created where the data had been aggregated to avoid overfitting. For the final RF model, recharge at every individual year, and the long-term average was predicted. The prediction was possible by including the time-dependent data discharge and precipitation. The final RF model had an R2 of
0.16 with RMSE at 726 mm/year for OOB samples. Using all the samples, R2 was 0.67 with a RMSE of 456 mm/year.
​
The average recharge for the final RF model was approximately 200 mm/year between 1955 to 2023. On average, the recharge to precipitation was 46%, with a standard deviation of 14%. The Pearson correlation between the predicted recharge and precipitation was 72%. It is believed that the relatively high amount of recharge is likely a result of bias in the dataset.
​
Change in focus and diffuse recharge over time was analyzed. The recharge of both focused and diffuse recharge varied around an average value between 1995 and 2006. From 2006 to today, there was less diffuse recharge and total recharge. Focus recharge from 2006 shows a significant trend in its relative contributions to total recharge.
Improvements are possible to make to the RF model. Suggestions include adding diffuse data points with a constant recharge/precipitation at 3.4%. A further improvement could be to add urban land use as a feature. A third improvement could be giving samples representing diffuse recharge a higher weight. A fourth improvement is increasing the consistency between the river network and the alluvial dataset. A fifth improvement could include landuse data from more years and dynamically updating the land use.
Thage, K.V (2023)
Fully integrated hydrological modelling using regression-based interpolation merging raingauge- and satellite data for the Hout-Sand catchment, South Africa. MSc thesis University of Copenhagen.
Assessment of the water balance in semi-arid and arid regions is essential for proper understanding of variation in groundwater level and it requires robust precipitation data to estimate correct spatial and temporal fluxes. However, the precise spatial distribution of precipitation in the Hout-Sand catchment, South Africa, a semi-arid area in the Limpopo river basin is not known. An existing fully integrated hydrological model for the area is enhanced by replacing the original precipitation product of coarse Thiessen polygons with a new product derived from rain gauge data and the satellite precipitation product IMERG using the regression-based interpolation (RBI) method followed by an auto-calibration. In the pre-analysis for the RBI, we found IMERG to perform better in a point-to-pixel- and covariate analysis than the satellite products ERA5 and TAMSAT for the period 2008-2019.
Here we show that the new precipitation product simulates a more realistic spatial distribution of the actual evapotranspiration (ET) than the Thiessen polygons, and that the new product is more adapted to the catchment than IMERG both spatially and seasonally. This affects all water fluxes resulting in precipitation events correlating more with changes in observed head elevation in the saturated zone (SZ) than the Thiessen polygons for locations far from the rain gauge stations. We found actual ET to be 93.5% of mean annual precipitation (MAP) and recharge to the SZ to be 2.9%MAP. The ET is slightly higher than previous studies estimate, while the recharge is comparable. The model proved unable to simulate correct discharge and head elevation in SZ due to falsely simulated exchange of water from the SZ to the river, an absence of overland flow and
no variation in the head elevation. These results demonstrate how hydrological models of sparse catchments, like the Hout-Sand, are sensitive to how the hydrogeological characteristics are implemented in the conceptual model as well as the impact spatiotemporal information from satellite products have on components otherwise lacking data in a hydrological model.
Aaes, A.H. (2023)
Fully integrated hydrological modelling using NDVI as a proxy for spatiotemporal agricultural field variation in the Hout-Sand catchment, South Africa. MSc thesis University of Copenhagen.
Groundwater is a fundamental component of the global hydroclimatic system and plays a central role in sustaining water supplies and livelihoods. The Hout and Sand catchments, South Africa, have a long history of commercial potato cultivation heavily dependent on intensive groundwaterfed center pivot irrigation.
However, the temporal variability of the cultivated agricultural practices are poorly documented. Based on Landsat remote sensing data, NDVI time series were computed for the fields in 2008-2019. An NDVI-threshold value of 0.7 was used to determine whether each field was active a given year. An existing fully integrated hydrological model for the area was improved with the data and subsequently object for an OSTRICH-autocalibration. The modified irrigation input showed that irrigation is higher than earlier anticipated. The absolute change was too small to measure in the water balance of the whole catchment area. Nevertheless, the cultivated agricultural field activity showed moderate spatial autocorrelation with agricultural fields being more frequently active in the southern part of the catchment. The model was calibrated with RMSE for the simulated groundwater head elevations, and the Pearson correlation coefficient, high flow event frequency, bias term, and flow variability error for discharge. The model’s simulated high flow event discharge improved, however a constant inflow from the saturated zone to the riverbed caused the baseflow to be higher than observations. A confining aquifer lens implemented around
the river did not completely limit the inflow to the river. The model proved unable to simulate accurate head elevations in the saturated zone. The results emphasize the inbound parameter sensitivity when setting up hydrological models in data-sparse environments, and that improved data inputs as well as a change in the conceptualization of the model framework is required.
Vejby J (2021)
Remote sensing pilot study for Waterpoort, South Africa
Correlation of SAR and NDVI, BSc Thesis, University of Copenhagen
Vejby J (2021)
Analyzing Sentinel data at the outlet of Sand River
BSc Thesis, University of Copenhagen.
Andersen ME (2021)
Fully integrated hydrological Mike SHE model of the Hout/Sand river
catchment in South Africa
MSc Thesis, University of Copenhagen.
Lindholm NHR, Jensen M (2021)
Rainfall-runoff modelling for two catchments in the Limpopo basin
BSc Thesis, University of Copenhagen.
Mkali A (2020)
Developing a hydrogeological conceptual model for subterranean groundwater control areas using remote sensing and geophysical techniques, Hout catchment, Limpopo, South Africa
MSc Thesis, University of Western Cape.
Vinqui L (2020)
Using groundwater process model to assess aquifer-river interaction for groundwater dependent communities; Hout River Basin, Limpopo, South Africa
MSc Thesis, University of Western Cape.
Frantsen FB (2019)
Catchment Investigation of Letsitele, Limpopo Basin, South Africa: Groundwater-surface water estimation and catchment modelling
MSc thesis, University of Copenhagen, 67 p.
Lindle J (2019)
Inferring groundwater recharge associations to ephemeral rivers, land use and climate using multi-decadal groundwater level observations from the semi-arid Limpopo basin of South Africa
MSc thesis, University of Copenhagen, 100 p.
Theagarajah A (2019)
Impacts on groundwater quality of managed aquifer recharge using discharges from the Polokwane wastewater treatment works, South Africa
MSc thesis, University of Oxford, UK, 74 p.
Wendt D (2019)
Water balance assessment for the Olifant Sand Water Scheme in Polokwane, South Africa, Internship Report
University of Birmingham, UK, 11 p.