Many developing countries experience widespread groundwater declination. Sustainable management actions include generation of an accurate groundwater distribution based on an extensive groundwater monitoring network which is often cost prohibiting in the context of a developing country such as Bangladesh. Further, such knowledge is lacking for the Sylhet region where groundwater was documented to be under tremendous pressure. Specifically, the gap in the current literature exists regarding groundwater trends and its areal extent for this region. This paper bridges the gap in research by focusing on trends and spatial and temporal variation of groundwater level changes for this area. This study addresses this problem by creating groundwater level predictions at the ungauged areas using geostatistical methods applied to a detailed set of data. In this study, the spatial variability of annual-average depth to the water table at 46 observation wells in the Sylhet division in Bangladesh is analyzed for 2000, 2005, 2010, and 2015. The geostatistical analysis applies the ordinary kriging method with cross-validation to create the water table maps for the study area. The results indicate a substantial increase in groundwater depths during the studied period from 2000 to 2015 in some locations in the study area. Importantly, this work identifies the vulnerable zones in the area due to the groundwater lowering trend. The study adds to the groundwater management research in developing countries and focuses on the spatial and temporal groundwater variation. The findings from the modeling exercise contribute to identification of the vulnerable areas and therefore help policymakers in making informed decisions to manage groundwater resources in this sensitive region sustainably.
Current knowledge about the potential impacts of microplastics (MP) on vadose zone hydrology is scarce. The primary goal of this study was to address some of the limitations of previous research by developing more reliable and conclusive statistical evidence to better understand whether MP pollution can potentially cause hydrological impacts. We examined the effects of MP shape (type), as well as the magnitude of pollution (MP/soil mass ratio, λ) on water holding capacity (WHC) and bare soil water evaporation (ER) of fine sand, under controlled laboratory conditions. Three different shapes (types) of MP—fiber (polyacrylic), strand (polymethyl methacrylate), and pellet (acrylonitrile butadiene styrene), with six environmentally relevant MP concentration levels (MP/soil mass ratio), all ≤1.5%, were studied. Statistical regressions and non-parametric analyses of variance (i.e., Kruskal–Wallis analysis) indicate that MP pollution has a substantial potential to change WHC and late-stage evaporation, even at relatively low MP concentrations, but has minimal impacts on early stage evaporation of the studied fine sand. The magnitude of the impacts depends on individual MP shape (type) and concentration, connoting those MP impact mechanisms are complex. These findings suggest that the global issue of growing soil–MP pollution should be regarded as a concerning environmental and water resources stressor that could potentially cause widespread environmental change by altering soil-water dynamics at the watershed scale.
In the last two decades, artificial neural networks (ANNs) have offered a valid alternative to physical modeling for river flow forecasting based on rainfall input, as testified by the numerous successful models in the literature. While there is noticeable heterogeneity in the range of applications, e.g., basin size, temporal resolution of the data, and neural network architecture, little effort has been made so far to systematically compare alternative modeling procedures on the same hydrological system, and among basins of different characteristics. The former could be helpful to establish a list of “best practices” when building data-driven ANN models for river flow forecasting; the latter could allow for a deeper understanding of the underlying hydrological phenomena, by detecting which information most improves model performance in forecasting flows and, therefore, is most relevant in the rainfall-runoff conversion process. By training and testing multiple ANN models for a range of basins in the state of Georgia (USA), with surface areas from tens to a few thousands of square miles, we systematically study the effects of (1) the spatial heterogeneity of past precipitation; (2) the length of the time-window of predictors and lead time for the forecasts; and (3) the basin size.