Within water resources management, surface water area (SWA) variation plays a vital role in hydrological processes as well as in agriculture, environmental ecosystems, and ecological processes. The monitoring of long-term spatiotemporal SWA changes is even more critical within highly populated regions that have an arid or semi-arid climate, such as Iran. This paper examined variations in SWA in Iran from 1990 to 2021 using about 18,000 Landsat 5, 7, and 8 satellite images through the Google Earth Engine (GEE) cloud processing platform. To this end, the performance of twelve water mapping rules (WMRs) within remotely-sensed imagery was also evaluated. Our findings revealed that (1) methods which provide a higher separation (derived from transformed divergence (TD) and Jefferies–Matusita (JM) distances) between the two target classes (water and non-water) result in higher classification accuracy (overall accuracy (OA) and user accuracy (UA) of each class). (2) Near-infrared (NIR)-based WMRs are more accurate than short-wave infrared (SWIR)-based methods for arid regions. (3) The SWA in Iran has an overall downward trend (observed by linear regression (LR) and sequential Mann–Kendall (SQMK) tests). (4) Of the five major water basins, only the Persian Gulf Basin had an upward trend. (5) While temperature has trended upward, the precipitation and normalized difference vegetation index (NDVI), a measure of the country’s greenness, have experienced a downward trend. (6) Precipitation showed the highest correlation with changes in SWA (r = 0.69). (7) Long-term changes in SWA were highly correlated (r = 0.98) with variations in the JRC world water map.

Continuous monitoring of Water Quality Parameters (WQPs) is crucial due to the global degradation of water quality, primarily caused by climate change and population growth. Typically, Machine Learning (ML) models are employed to retrieve WQPs, but they require a large amount of training samples to accurately capture the data relationships. Even with sufficient training data, discrepancies still exist between values of predicted and in-situ WQPs. This study proposes a Fuzzy Similarity Analysis (FSA) technique to enhance ML estimates of WQPs by using the prediction errors in Effective Training Samples (ETS). The method was successfully applied to retrieve Turbidity (Turb) and Specific Conductance (SC) in Lake Houston, USA, using Sentinel-2 remote sensing data. Three ML algorithms, namely Mixture Density Networks, Support Vector Regression, and Partial Least Squares Regression, were tested to evaluate the method's effectiveness. The results showed that FSA significantly improved the accuracy of all ML predictions. This improvement resulted in up to a 9.15% reduction in Mean Absolute Percentage Error (MAPE) and a 12% increase in R2 for Turb, while for SC, the improvements were 5.47% in MAPE and 7% in R2. The adaptability of the proposed method to other WQPs, various satellite data, and different ML models is promising for monitoring water quality in inland waters.

Remote Sensing (RS) data provide large-scale observations to measure Water Quality Parameters (WQPs) like Turbidity (Turb), which indicates the haziness of the water. Accurately estimating these parameters solely from RS data is inherently complex due to various factors, necessitating the use of advanced models capable of capturing the intricate relationships between RS spectral bands as input and the target parameter as output. One promising approach is the use of ensemble Machine Learning (ML) models, which construct more complex models by leveraging the complementary strengths of multiple base models. In this paper, a novel method known as Pruned Bagged Mixture Density Network (PBMDN) was proposed. First, using a bootstrap-based bagging approach, a pool of base MDN models was generated. Then, a forward pruning scheme was utilized to find an optimal subset of the base models for final ensemble aggregation. Coincident in-situ measurements of Turb and multispectral Sentinel-2 (S2) and Landsat-8 (L8) RS data for three water bodies in the USA were used to evaluate the performance of PBMDN. Results showed that PBMDN could achieve lower estimation errors (Mean Absolute Percentage Error (MAPE) of 25.25% for S2 and 35.71% for L8) compared to the single MDN model (MAPE of 40.53% for S2 and 47.92% for L8). PBMDN also performed significantly better than other widely used Turb estimation techniques, including other ML models and semi-empirical algorithms, indicating its strong potential in the estimation of WQPs using RS data.

Flash droughts are characterized by rapid onset and intensification, with severe impacts on agriculture and ecosystems. They often begin as meteorological droughts and, if conditions worsen, evolve into agricultural droughts. While precipitation deficit is often the primary driver, atmospheric and hydrological anomalies can exacerbate flash drought development. This study characterizes flash droughts across the Contiguous United States using remote sensing data from 2003 to 2020. A combination of satellite‐derived meteorological, agricultural, and ecological variables are used to investigate large‐scale flash drought development. Events are defined using root‐zone soil moisture. We used the Aridity Index to assess how background aridity influences agricultural and ecological impacts. Cross‐correlation and Cross Wavelet analyses are applied to examine the propagation of flash droughts from meteorological to agricultural and ecological stages. Results show that flash drought characteristics—including frequency, duration, and onset/recovery rates—are significantly influenced by landscape aridity characteristics. Precipitation is identified as the main driver across all climate regimes while Relative Humidity (RH) and Vapor Pressure Deficit (VPD) also indicate early signals. Time lags between meteorological variables and Soil Moisture (SM), as well as between soil moisture and ecological variables, vary across climates. Generally, results show that ecosystems respond to flash drought after soil moisture. Solar Induced Fluorescence (SIF), a measure of ecological stress, detects flash drought onset earlier than SM, highlighting its potential for early detection and monitoring.

Forecasting Water Quality Parameters (WQPs) in rivers and lakes is a challenging time series problem. Long Short-Term Memory (LSTM) networks are widely used for this purpose; however, prior studies showed that LSTM-based architectures struggle with the vanishing gradient problem due to their reliance on a single previous hidden state. As a result, LSTM-based models are unable to fully capture nonlinear dynamics over long temporal sequences. This study proposes a novel LSTM variant, referred to as NLSTM, which employs a multi-state memory structure with a child-parent architecture. In NLSTM, each parent node aggregates information from multiple prior hidden states through child-specific forget gates. The proposed model is evaluated for forecasting key WQPs, including pH, turbidity, and conductivity, under both temporal and spatiotemporal validation frameworks across three hydrologically and ecologically distinct study sites in the United States with varying data availability conditions, comprising two riverine systems and one inland lake. Results from temporal transferability experiments demonstrate that NLSTM, when trained on a specific temporal period, achieves superior performance compared to other models when forecasting WQPs for hold-out temporal periods within the same region. Spatiotemporal transferability experiments further show that the NLSTM model, when trained on one region, consistently provides more accurate forecasts when applied to unseen regions compared to other models. These improvements are attributed to (1) superior capture of long-range temporal dependencies, (2) mitigation of the vanishing gradient problem through enhanced gradient flow, and (3) richer high-level feature extraction, resulting in NLSTM outperforming other well-known models in this field.