Multivariate statistical methods including cluster analysis (CA), discriminant analysis (DA) and component analysis/factor analysis (PCA/FA), were applied to explore the surface water quality datasets including 14 parameters at 28 sites of the Eastern Poyang Lake Basin, Jiangxi Province of China, from January 2012 to April 2015, characterize spatiotemporal variation in pollution and identify potential pollution sources. The 28 sampling stations were divided into two periods (wet season and dry season) and two regions (low pollution and high pollution), respectively, using hierarchical CA method. Four parameters (temperature, pH, ammonia-nitrogen (NH₄-N), and total nitrogen (TN)) were identified using DA to distinguish temporal groups with close to 97.86% correct assignations. Again using DA, five parameters (pH, chemical oxygen demand (COD), TN, Fluoride (F), and Sulphide (S)) led to 93.75% correct assignations for distinguishing spatial groups. Five potential pollution sources including nutrients pollution, oxygen consuming organic pollution, fluorine chemical pollution, heavy metals pollution and natural pollution, were identified using PCA/FA techniques for both the low pollution region and the high pollution region. Heavy metals (Cuprum (Cu), chromium (Cr) and Zinc (Zn)), fluoride and sulfide are of particular concern in the study region because of many open-pit copper mines such as Dexing Copper Mine. Results obtained from this study offer a reasonable classification scheme for low-cost monitoring networks. The results also inform understanding of spatio-temporal variation in water quality as these topics relate to water resources management.

Modeling of hydrological time series is essential for sustainable development and management of lake water resources. This study aims to develop an efficient model for forecasting lake water level variations, exemplified by the Poyang Lake (China) case study. A random forests (RF) model was first applied and compared with artificial neural networks, support vector regression, and a linear model. Three scenarios were adopted to investigate the effect of time lag and previous water levels as model inputs for real-time forecasting. Variable importance was then analyzed to evaluate the influence of each predictor for water level variations. Results indicated that the RF model exhibits the best performance for daily forecasting in terms of root mean square error (RMSE) and coefficient of determination (R2). Moreover, the highest accuracy was achieved using discharge series at 4-day-ahead and the average water level over the previous week as model inputs, with an average RMSE of 0.25 m for five stations within the lake. In addition, the previous water level was the most efficient predictor for water level forecasting, followed by discharge from the Yangtze River. Based on the performance of the soft computing methods, RF can be calibrated to provide information or simulation scenarios for water management and decision-making.