Modelling the stream flow change in a poorly gauged mountainous watershed, southern Tianshan Mountain, using multi-source remote sensing data

Abstract

Hydrological predictions in ungauged or poorly gauged basin are crucial for sustainable water management and environmental changes study induced by climate change. Application of remote sensing technology has retrieved lots of spatio-temporal dataset during the past decades for references. In this study, TRMM/PR and MODIS LST data were introduced to get spatial patterns of precipitation and temperature changes by Empirical Orthogonal Function (EOF) technique in a mountainous watershed, southern Tianshan. An input variable group was attempted to be constructed for the Artificial Neural Networks (ANN) to model the stream flow change based on the patterns achieved above. The results indicate that the spatial variability patterns of meteorology can be well recognized from the remote sensing data by EOF analysis. The stream flow process can be satisfyingly simulated with input variables captured from the leading modes during the study period. While, since the probabilistic model was not based on full physical mechanisms, and often times, also limited by the amount of input data, uncertainties often implicated in the output. As an example, it is discussed through the rapidly glaciers melting phenomena induced by climate warming, which is expected to cause change in the flow generation mechanism.