Global change impact on hydro-biogeochemical processes in tropical Kenyan catchments

点击翻译按钮获取中文摘要

Climate and land use change have a significant impact on hydrobiogeochemical processes in the tropics. For tropical Africa in particular, however, scientific knowledge about the possible impacts of global change is limited. Nevertheless, this knowledge is essential for sustainable management of water resources. In this project, an established monitoring programme is being continued in four catchments with different land uses (tea and tree plantations, smallholder agriculture, natural montane rainforest) in the Mau Forest complex in western Kenya. Since 2014, automatic measuring systems have been recording almost gap-free in 10-minute resolution the water level (converted to discharge via rating curves) as well as the concentrations of NO3, DOC and turbidity (converted to suspended sediments via rating curves) by means of UV spectrometry. In addition, the concentrations of stable isotopes of water are measured weekly. While the measuring systems were established in the first project phase and basic knowledge about the relationship between land use and water quantity/quality was gained, the second project phase aims at an improved understanding of the underlying hydrobiogeochemical process and a projection of water fluxes (quantity and quality) with regard to climate and land use change. Three work packages (WP) are planned for this purpose. In WP1, the measurement programme and the necessary maintenance measures will be carried out. At the end of the project, a 10-year data set of the above parameters will be made available open access. WP2 focuses on process identification by means of statistical methods and analyses of systematic temporal patterns (diurnal variations, seasonal influences) using wavelet functions. Automated analyses of hysteresis loops of concentration-discharge dynamics will help to identify transport and mobilisation processes of water and its solutes. In addition, the established concept of "hydrological signatures" will be transferred to develop "hydro-biogeochemical signatures". This will allow to comparatively characterise the hydrochemistry of streams and describe their hydro-biogeochemical process behaviour. In WP3, data-based models will be developed using Deep Learning to simulate both runoff and water quality parameters. The latest Long Short-Term Memory (LSTM) methods will be used, which also take into account spatial (land use) and temporal (climate time series) predictors. For model validation, the wavelets, hysteresis loops and hydro-biogeochemical signatures calculated in WP2 on the basis of field measurements will be compared with those calculated on the basis of the LSTM models. Finally, using the LSTM models with spatio-temporal predictors, projections of climate and land use change will be made. For this purpose, the latest CORDEX simulations of regional climate models and in-house developed land use scenarios based on multitemporal land use classifications will be used.