Search

Climate change is one of the leading issues affecting river basins due to its direct impacts on the cryosphere and hydrosphere. General circulation models (GCMs) are widely applied tools to assess climate change but the coarse spatial resolution of GCMs limit their direct application for local studies. This study selected five CMIP5 GCMs (CCSM4, HadCM3, GFDL-CM3, MRI-CGCM3 and CanESM2) for performance evaluation ranked by Nash–Sutcliffe coefficient (NSE) and Kling–Gupta Efficiency (KGE). CCSM4 and HadCM3 large-scale predictors were favored based on ranks (0.71 and 0.68, respectively) for statistical downscaling techniques to downscale the climatic indicators Tmax, Tmin and precipitation. The performance of two downscaling techniques, Statistical Downscaling Methods (SDSM) and Long Ashton Research Station Weather Generator (LARS-WG), were examined using the Mean Absolute Error (MAE), Root Mean Square Error (RMSE), bias, NSE and KGE with weights (Wi) for the validation period. The results of statistical measures proved SDSM more efficient (0.67) in comparison to the LARS-WG (0.51) for the validation time for the Jhelum River basin. The findings revealed that the SDSM simulation for Tmax and Tmin are more comparable to the reference data for the validation period except simulation of extreme events by precipitation. The 21st century climatic projections exhibited a significant rise in Tmax (2.37–4.66 °C), Tmin (2.47–4.52 °C) and precipitation (7.4–11.54%) for RCP-4.5 and RCP-8.5, respectively. Overall, the results depicted that winter and pre-monsoon seasons were potentially most affected in terms of warming and precipitation, which has the potential to alter the cryosphere and runoff of the Jhelum River basin. [ABSTRACT FROM AUTHOR]

The potential of several predictive models including multiple model-artificial neural network (MM-ANN), multivariate adaptive regression spline (MARS), support vector machine (SVM), multi-gene genetic programming (MGGP), and 'M5Tree' were assessed to simulate the pan evaporation in monthly scale (EPm) at two stations (e.g. Ranichauri and Pantnagar) in India. Monthly climatological information were used for simulating the pan evaporation. The utmost effective input-variables for the MM-ANN, MGGP, MARS, SVM, and M5Tree were determined using the Gamma test (GT). The predictive models were compared to each other using several statistical criteria (e.g. mean absolute percentage error (MAPE), Willmott's Index of agreement (WI), root mean squared error (RMSE), Nash-Sutcliffe efficiency (NSE), and Legate and McCabe's Index (LM)) and visual inspection. The results showed that the MM-ANN-1 and MGGP-1 models (NSE, WI, LM, RMSE, MAPE are 0.954, 0.988, 0.801, 0.536 mm/month, 9.988% at Pantnagar station, and 0.911, 0.975, 0.724, and 0.364 mm/month, 12.297% at Ranichauri station, respectively) with input variables equal to six were more successful than the other techniques during testing period to simulate the monthly pan evaporation at both Ranichauri and Pantnagar stations. Thus, the results of proposed MM-ANN-1 and MGGP-1 models will help to the local stakeholders in terms of water resources management. [ABSTRACT FROM AUTHOR]

The Himalayan region, a rugged mountain zone is among the most susceptible zones to the landslide hazard due to its terrain, geography, and active tectonics. Machine learning (ML) techniques are most advanced and precise methods to develop landslide susceptibility model (LSM). The current study was designed to analyze and assess the landslide susceptibility using ML approaches for District Hattian Bala, NW Himalayas, Pakistan. The historical satellite imageries are used to generate spatiotemporal landslide inventories of year 2005, 2007 and 2012. A spatial database was created pertaining to topographic, environmental, geologic, and anthropogenic factors including slope, aspect, elevation, curvature, plane curvature, profile curvature, topographic wetness index (TWI), lithology, distance to faults, distance to streams, distance to roads, normalized difference vegetation index (NDVI) and land use/ land cover (LULC). These LCFs were selected to analyze periodic landslide susceptibility in the region. The experimental design utilized 349, 393, and 735 landslide inventory of 2005, 2007, and 2012 respectively. Two ML models, i.e., Random Forest (RF) and Logistic Regression (LR) were applied to assess landslide susceptibility determine by thirteen landslide causative factors (LCFs). The spatiotemporal landslide inventory was partitioned into training (70%) and testing (30%) landslides for respective years to check the prediction accuracies of selected ML models. Comparative analysis of different LSMs was performed by the Receiver Operator Curves – Area Under Curves (ROC-AUC). The resultant accuracy, MAE, RMSE, Kappa, Precision, Recall, F1 indicated that RF outperformed the LR model. The study aims to minimize losses to lives and potential economic damage linked with recurrent slope instabilities in the region. It is anticipated that use of ML algorithms would support concerned authorities and organizations to effectively plan and manage landslide hazard in the region. [ABSTRACT FROM AUTHOR]