Search

Climate change imposes profound influence on regional hydrological cycle and water security in many alpine regions worldwide. Investigating regional climate impacts using watershed scale hydrological models requires a large number of input data such as topography, meteorological and hydrological data. However, data scarcity in alpine regions seriously restricts evaluation of climate change impacts on water cycle using conventional approaches based on global or regional climate models, statistical downscaling methods and hydrological models. Therefore, this study is dedicated to development of a probabilistic model to replace the conventional approaches for streamflow projection. The probabilistic model was built upon an advanced Bayesian Neural Network (BNN) approach directly fed by the large-scale climate predictor variables and tested in a typical data sparse alpine region, the Kaidu River basin in Central Asia. Results show that BNN model performs better than the general methods across a number of statistical measures. The BNN method with flexible model structures by active indicator functions, which reduce the dependence on the initial specification for the input variables and the number of hidden units, can work well in a data limited region. Moreover, it can provide more reliable streamflow projections with a robust generalization ability. Forced by the latest bias-corrected GCM scenarios, streamflow projections for the 21st century under three RCP emission pathways were constructed and analyzed. Briefly, the proposed probabilistic projection approach could improve runoff predictive ability over conventional methods and provide better support to water resources planning and management under data limited conditions as well as enable a facilitated climate change impact analysis on runoff and water resources in alpine regions worldwide. [ABSTRACT FROM AUTHOR]

The Indicators of Hydrologic Alteration (IHA) based on 33 parameters in five dimensions (flow magnitude, timing, duration, frequency and change rate) have been widely used in evaluation of hydrologic alteration in river systems. Yet, inter-correlation seriously exists amongst those parameters, therefore constantly underestimates or overestimates actual hydrological changes. Toward the end, a new method (Representative-IHA, RIHA) is developed by removing repetitions based on Criteria Importance Through Intercriteria Correlation (CRITIC) algorithm. RIHA is testified in evaluating effects of future climate change on hydro-ecology in the Niger River of Africa. Future flows are projected using three watershed hydrological models forced by five general circulation models (GCMs) under three Representative Concentration Pathways (RCPs) scenarios. Results show that: (1) RIHA is able to eliminate self-correlations amongst IHA indicators and identify the dominant characteristics of hydrological alteration in the Upper Niger River, (2) March streamflow, September streamflow, December streamflow, 30-day annual maximum, low pluses duration and fall rates tends to increase over the period 2010–2099, while July streamflow and 90-day annual minimum streamflow shows decrease, (3) the Niger River will undergo moderate flow alteration under RCP8.5 in 2050s and 2080s and low alteration other scenarios, (4) future flow alteration may induce increase water temperatures, reduction dissolved oxygen and food resources. Consequently, aquatic biodiversity and fish community of Upper Niger River would become more vulnerable in the future. The new method enables more scientific evaluation for multi-dimensional hydrologic alteration under the context of climate change. [ABSTRACT FROM AUTHOR]