A statistical downscaling approach for improving extreme rainfall simulation was proposed to predict the daily rainfalls at Shih-Men Reservoir catchment in northern Taiwan. The structure of the proposed downscaling approach is composed of two parts: the rainfall-state classification and the regression for rainfall-amount prediction. Predictors of classification and regression methods were selected from the large-scale climate variables of the NCEP reanalysis data based on statistical tests. The data during 1964–1999 and 2000–2013 were used for calibration and validation, respectively. Three classification methods, including linear discriminant analysis (LDA), random forest (RF), and support vector classification (SVC), were adopted for rainfall-state classification and their performances were compared. After rainfall-state classification, the least square support vector regression (LS-SVR) was used for rainfall-amount prediction for different rainfall states. Two rainfall states (i.e., dry day and wet day) and three rainfall states (dry day, non-extreme-rainfall day, and extreme-rainfall day) were defined and compared for judging their downscaling performances. The results show that RF outperforms LDA and SVC for rainfall-state classification. Using RF for three-rainfall-states classification and LS-SVR for rainfall-amount prediction can improve the extreme rainfall downscaling.

This work provides a design for two types of sensors, based on the thermal dissipation and heat ratio methods of sap flow calculation, for moderate to large scale deployments for the purpose of monitoring tree transpiration. These designs include a procedure for making these sensors, a quality control method for the final products, and a complete list of components with vendors and pricing information. Both sensor designs were field tested alongside a commercial sap flow sensor to assess their performance and show the importance for quality controlling the sensor outputs. Results show that for roughly 2% of the cost of commercial sensors, self-made sap flow sensors can provide acceptable estimates of the sap flow measurements compared to the commercial sensors.

Monthly rainfall forecasts can be translated into monthly runoff predictions that could support water resources planning and management activities. Therefore, development of monthly rainfall forecasting models in reservoir watersheds is essential for generating future rainfall amounts as an input to a water-resources-system simulation model to predict water shortage conditions. This research aims to examine the reliability of linking a data preprocessing method (singular spectrum analysis, SSA) with machine learning, least-squares support vector regression (LS-SVR), and random forest (RF), for monthly rainfall forecasting in two reservoir watersheds (Deji and Shihmen reservoir watersheds) located in Taiwan. Merging SSA with LS-SVR and RF, the hybrid models (SSA-LSSVR and SSA-RF) were developed and compared with the standard models (LS-SVR and RF). The proposed models were calibrated and validated using the watersheds’ observed areal monthly rainfalls separated into 70 percent of data for calibration and 30 percent of data for validation. Model performances were evaluated using two accuracy measures, root mean square error (RMSE) and Nash–Sutcliffe efficiency (NSE). Results show that the hybrid models could efficiently forecast monthly rainfalls. Nonetheless, the performances of the hybrid models vary in both watersheds which suggests that prior knowledge about the watershed’s hydrological behavior would be helpful to implement the appropriate model. Overall, the hybrid models significantly surpass the standard models for the two studied watersheds, which indicates that the proposed models are a prudent modeling approach that could be employed in the current research regions for monthly rainfall forecasting.