Developing a method to estimate the water use of South African natural vegetation using remote sensing.
The scarcity of water is a growing concern throughout the world. It is essential to accurately determine the quantity and quality of this valuable resource to aid in water resource planning and management. For this purpose a hydrological baseline is required to compare against the water use of other land uses. Currently, the Acocks (1988) Veld Type is the baseline land cover used for hydrological studies. However, there are several shortcomings associated with this baseline land cover that may be overcome by using the recently released natural land cover map produced by South African National Biodiversity Institute (SANBI) 2012. A barrier to the use of the SANBI (2012) vegetation map is that, the water use parameters have not been determined for the various vegetation units defined. Vegetation water use can be determined by estimating the total evaporation (ET). There are a number of in-situ methods available to estimate ET. However, these methods estimate ET based on point or line averaged measurements which are only representative of local scales and cannot be extended to large areas because of land surface heterogeneity. The application of remote sensing energy balance models has the potential to overcome these limitations. Remote sensing has the ability to produce large spatial scale estimates of ET. It can also provide information at remote sites where it is difficult to install instruments. The focus of this study was to develop a method to estimate ET for natural vegetation of South Africa using remote sensing. The Surface Energy Balance System (SEBS) model in conjunction with Landsat 7 ETM+ and 8 OLI/TRS images was first used to validate point-based ET from various biomes across the country. The results from the study indicate a fair comparison between the in-situ ET data and the evaporation estimates produced using the SEBS model with coefficient of determination value of 0.66 being achieved and a RMSE of 1.74 mm.day-1. The highest RMSE was attained for the Ingeli forest site whilst the lowest belonged to the Nama Karoo site of 2.2 mm.day-1 and 0.5 mm.day-1, respectively. The SEBS model was able to estimate ET which mimics the trend of in-situ ET well. However, the model tends to over-estimate ET in comparison to in-situ ET data. Following the validation of the in-situ and SEBS ET, the SEBS model was applied to model ET for a year. For this investigation, cloud free Landsat 8 OLI/TRS images was obtained for each biome for the period between 1 July 2014 to 31 June 2015. The highest ET value of 8.7 mm/day was obtained from the Forest biome on the 12 January 2015 and the lowest ET estimate of 0.09 mm/day was on the 17 January 2015 for the Nama Karoo biome. The Forest biome recorded the highest mean ET value of 4.9 mm/day whilst the lowest mean ET value was 0.71 mm/day attained from the Nama Karoo biome. Satellite derived ET using the SEBS model produced reliable estimates when compared to in- situ ET. The spatial and temporal resolution of ET can be achieved using remote sensing. The ET estimates from SEBS compared well to the in-situ ET measurements and followed the seasonal trend, however an over-estimation of ET was present in some cases. Overall, remote sensing proves a viable option to estimate ET over large areas. This method can be applied to derive the water use which can be used to determine water use parameters.
Master of Science in Hydrology. University of KwaZulu-Natal, Pietermaritzburg 2016.
Plants - Water Requirements., Pllants - Remote Sensing., Theses - Environmental Hydrology.