This study presents the first comprehensive evaluation in West Africa of four imputation methods, Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), and Ordinary Kriging (OK), applied to six core meteorological variables across Senegal over a ten-year period (2015-2024). By simulating realistic missing data scenarios informed by field conditions (e.g., power outages, observer absences, sensor failures), it establishes a robust benchmark for climate data reconstruction using machine learning in resource-constrained settings. The findings highlight the clear superiority of ensemble learning approaches. XGB consistently outperformed all methods across variables and scenarios, achieving the highest average predictive accuracy with R2 values up to [95 % CI: 0.82-0.88], along with lower Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). RF yielded comparable performance, especially for maximum and minimum temperature (TMAX, TMIN), maintaining strong stability even at 20 % missingness. In contrast, DT performance declined sharply with increased data loss, and OK was constrained by the sparse spatial distribution of meteorological stations, notably impairing its ability to impute precipitation (PRCP) and wind speed (WDSP). This work contributes a multivariable imputation framework specifically adapted to West African climatic and infrastructural realities. It also integrates block bootstrap methods to quantify uncertainty and derive 95 % confidence intervals for all error metrics. Results confirm that imputation effectiveness is highly variabledependent: continuous and temporally autocorrelated variables (TMAX, TMIN, dew point temperature - DEWP) are well reconstructed, whereas discontinuous or noisy variables (WDSP and PRCP) remain challenging. Although ensemble models offer clear advantages, their computational demands and need for hyperparameter tuning may limit real-time implementation in low-resource national meteorological services. Furthermore, the exclusion of satellite or reanalysis inputs may constrain model generalizability. Ultimately, this study reinforces the role of advanced machine learning methods in improving climate data completeness and reliability in Africa. Although not a substitute for direct observations, imputation emerges as a critical complementary tool to support robust and resilient climate information systems essential for agriculture, public health, and disaster risk management under intensifying climate variability.

AbstractAs extreme precipitation intensifies under climate change, traditional risk models based on the ‘100-year return period’ concept are becoming inadequate in assessing real-world risks. In response, this nationwide study explores shifting extremes under non-stationary warming using high-resolution data across the contiguous United States. Results reveal pronounced variability in 100-year return levels, with Coastal and Southern regions displaying the highest baseline projections, and future spikes are anticipated in the Northeast, Ohio Valley, Northwest, and California. Exposure analysis indicates approximately 53 million residents currently reside in high-risk zones, potentially almost doubling and tripling under 2 °C and 4 °C warming. Drought frequency also rises, with over 37% of major farmland vulnerable to multi-year droughts, raising agricultural risks. Record 2023 sea surface temperature anomalies suggest an impending extreme El Niño event, demonstrating the need to account for natural climate variability. The insights gained aim to inform decision-makers in shaping adaptation strategies and enhancing the resilience of communities in response to evolving extremes.

Water resource managers require accurate estimates of the 7-day, 10-year low flow (7Q10) of streams for many reasons, including protecting aquatic species, designing wastewater treatment plants, and calculating municipal water availability. StreamStats, a publicly available web application developed by the United States Geologic Survey that is commonly used by resource managers for estimating the 7Q10 in states where it is available, utilizes state-by-state, locally calibrated regression equations for estimation. This paper expands StreamStats’ methodology and improves 7Q10 estimation by developing a more regionally applicable and generalized methodology for 7Q10 estimation. In addition to classical methodologies, namely multiple linear regression (MLR) and multiple linear regression in log space (LTLR), three promising machine learning algorithms, random forest (RF) decision trees, neural networks (NN), and generalized additive models (GAM), are tested to determine if more advanced statistical methods offer improved estimation. For illustrative purposes, this methodology is applied to and verified for the full range of unimpaired, gaged basins in both the northeast and mid-Atlantic hydrologic regions of the United States (with basin sizes ranging from 2–1419 mi2) using leave-one-out cross-validation (LOOCV). Pearson’s correlation coefficient (R2), root mean square error (RMSE), Kling–Gupta Efficiency (KGE), and Nash–Sutcliffe Efficiency (NSE) are used to evaluate the performance of each method. Results suggest that each method provides varying results based on basin size, with RF displaying the smallest average RMSE (5.85) across all ranges of basin sizes.

Two fundamental problems have inhibited progress in the simulation of river water quality under climate (and other) uncertainty: 1) insufficient data, and 2) the inability of existing models to account for the complexity of factors (e.g., hydro-climatic, basin characteristics, land use features) affecting river water quality. To address these concerns this study presents a technique for augmenting limited ground-based observations of water quality variables with remote-sensed surface reflectance data by leveraging a machine learning model capable of accommodating the multidimensionality of water quality influences. Total Suspended Solids (TSS) can serve as a surrogate for chemical and biological pollutants of concern in surface water bodies. Historically, TSS data collection in the United States has been limited to the location of water treatment plants where state or federal agencies conduct regularly-scheduled water sampling. Mathematical models relating riverine TSS concentration to the explanatory factors have therefore been limited and the relationships between climate extremes and water contamination events have not been effectively diagnosed. This paper presents a method to identify these issues by utilizing a Long Short-Term Memory Network (LSTM) model trained on Moderate Resolution Imaging Spectroradiometer (MODIS) satellite reflectance data, which is calibrated to TSS data collected by the Ohio River Valley Water Sanitation Commission (ORSANCO). The methodology developed enables a thorough empirical analysis and data-driven algorithms able to account for spatial variability within the watershed and provide effective water quality prediction under uncertainty.

Bangladesh is one of the climate risk-prone countries in South Asia facing tremendous challenges to combat sea-level rise and its associated coastal morphological changes. This study aimed to determine the interaction of the sea-level rise and morphological changes, particularly at Cox’s Bazar and Kutubdia Island along the eastern coast of Bangladesh. Available hourly tide gauge data, daily temperature, daily rainfall data, and 15 LANDSAT satellite images for the period of 1983–2016 were analyzed to examine the sea level shore morphological change and associated climate change phenomenon. First, we identified the historical nonlinear sea-level trend using Hilbert-Huang Transformation (HHT) based on the complete ensemble empirical mode decomposition (CEEMD) technique. We divided the study period into three distinct sea-level change periods of 1983–1993, 1993–2003, and 2003–2014 based on nonlinear sea-level trend analysis. The study revealed that the sea level on the east coast of Bangladesh had a moderate rising trend during 1983–1993, slight decrease during 1993–2003, and steep rising trend during 2003–2014. We also observed that a sea-level change within a particular period impacted the shore morphological change after approximately two years, such that the average sea-level change during the period of 1993–2003 might have affected the shore morphology for 1996–2005. Alarming shore erosion was found for the period of 2005–2016 compared to the previous periods of 1989–1996 and 1996–2005 for both Cox’s Bazar and Kutubdia Island. The shore morphology of some segments was also substantially affected due to the geometric shape of the land, significant waves, and shore protection works. This study encourages policymakers to minimize the threats of sea-level rise and ensure sustainable coastal management strategies are introduced to sustain the vital eastern coast of Bangladesh.

Salinity causes a hostile environmental impact throughout the year in the coastal region of Bangladesh, and its severity increases day by day. Because of upstream freshwater flow reduction and massive groundwater extraction, salinity has increased substantially over the last three decades. Moreover, arsenic contamination in shallow groundwater makes the groundwater unsuitable for potable use. Consequently, the coastal area suffers from acute storage of safe water supply. Salinity also negatively impacts human activities, livelihood, agricultural production, and the aquatic ecosystem. Though the shallow aquifer contains high salinity and a small amount of Arsenic (As), the very shallow aquifer (within 3m to 8m) contains fresh water in many areas in the rainy season due to the direct recharge of rainwater. However, rainfall recharge varies significantly depending on the geological and hydrogeological settings. Specifically, up to 50% of annual rainfall is stored in shallow aquifers of Quaternary sands through direct infiltration. The research’s principal objective is to identify the safe and sustainable drinking water source in the arsenic and saline-prone coastal region. Groundwater samples were collected from the different locations of the study area during both dry and wet seasons and examined seasonal variations in groundwater table and salinity levels. The chemical analyses and Physico-chemical parameters indicate that the groundwater samples are suitable for drinking. Except for some groundwater samples from the wet season, the salinity of all samples was under the allowable limit for Bangladesh (<2000 µS/cm), and the targeted aquifer was almost arsenic (50 µg/l) free. Therefore, a comprehensive analysis has been made to accomplish the study goals. Particularly, the groundwater’s electrical conductivity (EC) values of most samples were measured within the limit of fresh or brackish water (<2000 μS/cm). Overall, the results indicate the prospect of a very shallow aquifer as a source of freshwater for drinking purposes throughout the year, considering both arsenic and salinity, which effectively solve the freshwater shortage, especially in the saline-arsenic prone area.

Urbanization leads to the construction of various urban infrastructures in the city area for residency, transportation, industry, and other purposes, which causes major land use change. Consequently, it substantially affects Land Surface Temperature (LST) by unbalancing the surface energy budget. Higher LST in city areas decreases human thermal comfort for the city dwellers and affects the urban environment and ecosystem. Therefore, a comprehensive investigation is needed to evaluate the impact of land use change on the LST. Remote Sensing (RS) and Geographic Information System (GIS) techniques were used for the detailed investigation. RS data for the years 1993, 2007 and 2020 during summer (March–May) in Dhaka city were used to prepare land cover maps, analyze LST, generate hazard maps and relate the land cover change with LST by using GIS. The results show that the built-up area in Dhaka city increased by 67% from 1993 to 2020 by replacing lowland mainly, followed by vegetation, bare soil and water bodies. LSTs found in the study area were ranged from 23.26 to 39.94 °C, 23.69 to 43.35 °C and 24.44 to 44.58 °C for the years 1993, 2007 and 2020, respectively. The increases of spatially distributed maximum and mean LST were found 4.62 °C and 6.43 °C, respectively, for the study period of 27 years while the change in minimum LST was not substantial. LST increased by around 0.24 °C per year and human thermal discomfort shifted from moderate to strong heat stress for the total study period due to the increase of built-up and bare lands. This study also shows that normalized difference vegetation index (NDVI) and normalized difference water index (NDWI) were negatively correlated with LST while normalized difference built-up Index (NDBI) and normalized difference built-up Index (NDBAI) were positively correlated with LST. The methodology developed in this study can be adapted to other cities around the globe.

The representation of precipitation variability over mountainous regions by ground-based sensors is an open problem in hydrometeorological applications that necessitates the use of satellite-based precipitation products (SPPs). An extended network of ground-based X-band radar (GR) deployments over complex terrain areas, including the northeastern Italian Alps, North Carolina, Olympic Mountain, and the southern tip of Vancouver Island, is used in this study as a benchmark rainfall data set for error characterization and modeling of Level 2 PMW retrievals (Goddard profiling (GPROF) V05 algorithm) for the different sensors: the Microwave Humidity Sounder (MHS), the Special Sensor Microwave Imager/Sounder (SSMIS), the Global Precipitation Measurement Microwave Imager (GMI), and the Advanced Microwave Scanning Radiometer 2 (AMSR2). Matchups of Level 2 PMW/GR rainfall are extracted based on a matching methodology that identifies GR volume scans with PMW overpasses, and scales GR parameters to the satellite products' nominal spatial resolution. The error model is the nonparametric machine learning tree-based quantile regression forest (QRF), which we developed using matchups of PMW/GR rainfall data from the different study areas. Validation of the error model is conducted using three cross-validation techniques: the k-fold, leave-one region out, and enforced. All validations showed that the error model-based corrections can significantly reduce both the mean relative error and the random component of PMW products. Moreover, the error reduction demonstrated with the leave-one region out cross-validation technique indicated that the error model is transferable among complex terrain regions. Algorithm developers may find this error model useful to integrate in the Level 3 products.

Suburban growth and its impacts on surface runoff were investigated using the soil conservation service curve number (SCS-CN) model, compared with the integrated advanced remote sensing and geographic information system (GIS)-based integrated approach, over South Kingston, Rhode Island, USA. This study analyzed and employed the supervised classification method on four Landsat images from 1994, 2004, 2014, and 2020 to detect land-use pattern changes through remote sensing applications. Results showed that 68.6% urban land expansion was reported from 1994 to 2020 in this suburban area. After land-use change detection, a GIS-based SCS-CN model was developed to examine suburban growth and surface runoff estimation. The developed model demonstrated the spatial distribution of runoff for each of the studied years. The results showed an increasing spatial pattern of 2% to 10% of runoff from 1994 to 2020. The correlation between runoff co-efficient and rainfall indicated the significant impact of suburban growth in surface runoff over the last 36 years in South Kingstown, RI, USA, showing a slight change of forest (8.2% area of the total area) and agricultural land (4.8% area of the total area). Suburban growth began after 2000, and within 16 years this land-use change started to show its substantial impact on surface runoff. We concluded that the proposed integrated approach could classify land-use and land cover information to understand suburban growth and its potential impact on the area.

Very high spatial resolution commercial satellite imagery can inform observation, mapping, and documentation of micro-topographic transitions across large tundra regions. The bridging of fine-scale field studies with pan-Arctic system assessments has until now been constrained by a lack of overlap in spatial resolution and geographical coverage. This likely introduced biases in climate impacts on, and feedback from the Arctic region to the global climate system. The central objective of this exploratory study is to develop an object-based image analysis workflow to automatically extract ice-wedge polygon troughs from very high spatial resolution commercial satellite imagery. We employed a systematic experiment to understand the degree of interoperability of knowledge-based workflows across distinct tundra vegetation units-sedge tundra and tussock tundra-focusing on the same semantic class. In our multi-scale trough modelling workflow, we coupled mathematical morphological filtering with a segmentation process to enhance the quality of image object candidates and classification accuracies. Employment of the master ruleset on sedge tundra reported classification accuracies of correctness of 0.99, completeness of 0.87, and F1 score of 0.92. When the master ruleset was applied to tussock tundra without any adaptations, classification accuracies remained promising while reporting correctness of 0.87, completeness of 0.77, and an F1 score of 0.81. Overall, results suggest that the object-based image analysis-based trough modelling workflow exhibits substantial interoperability across the terrain while producing promising classification accuracies. From an Arctic earth science perspective, the mapped troughs combined with the ArcticDEM can allow hydrological assessments of lateral connectivity of the rapidly changing Arctic tundra landscape, and repeated mapping can allow us to track fine-scale changes across large regions and that has potentially major implications on larger riverine systems.

This study presents a comprehensive investigation of multiple Artificial Intelligence (AI) techniques—decision tree, random forest, gradient boosting, and neural network—to generate improved precipitation estimates over the Upper Blue Nile Basin. All the AI methods merged multiple satellite and atmospheric reanalysis precipitation datasets to generate error-corrected precipitation estimates. The accuracy of the model predictions was evaluated using 13 years (2000–2012) of ground-based precipitation data derived from local rain gauge networks in the Upper Blue Nile Basin region. The results indicate that merging multiple sources of precipitation substantially reduced the systematic and random error statistics in the Upper Blue Nile Basin. The proposed methods have great potential in predicting precipitation over the complex terrain region.

A growing number of electricity utilities use machine learning-based outage prediction models (OPMs) to predict the impact of storms on their networks for sustainable management. The accuracy of OPM predictions is sensitive to sample size and event severity representativeness in the training dataset, the extent of which has not yet been quantified. This study devised a randomized and out-of-sample validation experiment to quantify an OPM's prediction uncertainty to different training sample sizes and event severity representativeness. The study showed random error decreasing by more than 100% for sample sizes ranging from 10 to 80 extratropical events, and by 32% for sample sizes from 10 to 40 thunderstorms. This study quantified the minimum number of sample size for the OPM attaining an acceptable prediction performance. The results demonstrated that conditioning the training of the OPM to a subset of events representative of the predicted event's severity reduced the underestimation bias exhibited in high-impact events and the overestimation bias in low-impact ones. We used cross entropy (CE) to quantify the relatedness of weather variable distribution between the training dataset and the forecasted event.

The Integrated Multisatellite Retrievals for Global Precipitation Measurement (GPM) (IMERG) Level 3 estimates rainfall from passive microwave sensors onboard satellites that are associated with several uncertainty sources such as sensor calibration, retrieval errors, and orographic effects. This study aims to provide a comprehensive investigation of multiple machine learning (ML) techniques (Random Forest, and Neural Networks), to stochastically generate an error-corrected improved IMERG precipitation product at a daily time scale and 0.1°-degree spatial resolution over the Brahmaputra river basin. In this study, we used the operational IMERG-Late Run version 06 product along with several meteorological and land surface parameters (elevation, soil type, land type, soil moisture, and daily maximum and minimum temperature) to produce an improved precipitation product in the Brahmaputra basin. We trained, tested, and optimized ML algorithms using 4 years (from 2015 through 2019) of reference rainfall data derived from the rain gauge. The ML generated precipitation product exhibited improved systematic and random error statistics for the study area, which is a strong indication for using the proposed algorithms in retrieving precipitation across the globe. We conclude that the proposed ML-based ensemble framework has the potential to quantify and correct the error sources for improving and promoting the use of satellite-based precipitation estimates for water resources applications.

Permafrost thaw has been observed at several locations across the Arctic tundra in recent decades; however, the pan-Arctic extent and spatiotemporal dynamics of thaw remains poorly explained. Thaw-induced differential ground subsidence and dramatic microtopographic transitions, such as transformation of low-centered ice-wedge polygons (IWPs) into high-centered IWPs can be characterized using very high spatial resolution (VHSR) commercial satellite imagery. Arctic researchers demand for an accurate estimate of the distribution of IWPs and their status across the tundra domain. The entire Arctic has been imaged in 0.5 m resolution by commercial satellite sensors; however, mapping efforts are yet limited to small scales and confined to manual or semi-automated methods. Knowledge discovery through artificial intelligence (AI), big imagery, and high performance computing (HPC) resources is just starting to be realized in Arctic science. Large-scale deployment of VHSR imagery resources requires sophisticated computational approaches to automated image interpretation coupled with efficient use of HPC resources. We are in the process of developing an automated Mapping Application for Permafrost Land Environment (MAPLE) by combining big imagery, AI, and HPC resources. The MAPLE uses deep learning (DL) convolutional neural nets (CNNs) algorithms on HPCs to automatically map IWPs from VHSR commercial satellite imagery across large geographic domains. We trained and tasked a DLCNN semantic object instance segmentation algorithm to automatically classify IWPs from VHSR satellite imagery. Overall, our findings demonstrate the robust performances of IWP mapping algorithm in diverse tundra landscapes and lay a firm foundation for its operational-level application in repeated documentation of circumpolar permafrost disturbances.

This paper evaluates the machine learning-based weather-related crash prediction model in Connecticut. Crash severity prediction has always been the principal focus of safety professionals and emergency responders for appropriate policy making and resource management. Over the years, different statistical methodologies (e.g., random forest, support vector machine) have been explored in various research efforts to develop efficient crash severity prediction models. As technology is advancing and computing has started becoming more efficient, machine learning-based models for crash severity prediction are being brought into light for more accurate data-driven prediction. However, some machine learning methodologies provide increased efficiency and better performance compared to others from the same genre. To explore different machine learning methodologies for crash severity prediction, this study considered two machine learning applications—random forest (RF) and bayesian additive regression trees (BART) for performance comparison. RF model produced higher prediction probabilities for determining crash severity compare to BART. The results were evaluated using various prediction probability analyses obtained from these two machine learning models and showed the model capabilities to generate prediction consistent with the observed data. We found the performance of the RF model to be highly promising with a higher skill score (0.73) than BART (0.61). Overall, our findings demonstrate the robust performances of the RF algorithm in predicting weather-related crashes. The analysis of this study confirms that stakeholders can use weather-related RF model with confidence to obtain a better prediction of crash severity that enables them to facilitate appropriate emergency responses and support essential preemptive measures.

We developed a high-throughput mapping workflow, which centers on deep learning (DL) convolutional neural network (CNN) algorithms on high-performance distributed computing resources, to automatically characterize ice-wedge polygons (IWPs) from sub-meter resolution commercial satellite imagery. We applied a region-based CNN object instance segmentation algorithm, namely the Mask R-CNN, to automatically detect and classify IWPs in North Slope of Alaska. The central goal of our study was to systematically expound the DLCNN model interoperability across varying tundra types (sedge, tussock sedge, and non-tussock sedge) and image scene complexities to refine the understanding of opportunities and challenges for regional-scale mapping applications. We corroborated quantitative error statistics along with detailed visual inspections to gauge the IWP detection accuracies. We found promising model performances (detection accuracies: 89% to 96% and classification accuracies: 94% to 97%) for all candidate image scenes with varying tundra types. The mapping workflow discerned the IWPs by exhibiting low absolute mean relative error (AMRE) values (0.17–0.23). Results further suggest the importance of increasing the variability of training samples when practicing transfer-learning strategy to map IWPs across heterogeneous tundra cover types. Overall, our findings demonstrate the robust performances of IWPs mapping workflow in multiple tundra landscapes.

Deep learning (DL) convolutional neural networks (CNNs) have been rapidly adapted in very high spatial resolution (VHSR) satellite image analysis. DLCNN-based computer visions (CV) applications primarily aim for everyday object detection from standard red, green, blue (RGB) imagery, while earth science remote sensing applications focus on geo object detection and classification from multispectral (MS) imagery. MS imagery includes RGB and narrow spectral channels from near- and/or middle-infrared regions of reflectance spectra. The central objective of this exploratory study is to understand to what degree MS band statistics govern DLCNN model predictions. We scaffold our analysis on a case study that uses Arctic tundra permafrost landform features called ice-wedge polygons (IWPs) as candidate geo objects. We choose Mask RCNN as the DLCNN architecture to detect IWPs from eight-band Worldview-02 VHSR satellite imagery. A systematic experiment was designed to understand the impact on choosing the optimal three-band combination in model prediction. We tasked five cohorts of three-band combinations coupled with statistical measures to gauge the spectral variability of input MS bands. The candidate scenes produced high model detection accuracies for the F1 score, ranging between 0.89 to 0.95, for two different band combinations (coastal blue, blue, green (1,2,3) and green, yellow, red (3,4,5)). The mapping workflow discerned the IWPs by exhibiting low random and systematic error in the order of 0.17–0.19 and 0.20–0.21, respectively, for band combinations (1,2,3). Results suggest that the prediction accuracy of the Mask-RCNN model is significantly influenced by the input MS bands. Overall, our findings accentuate the importance of considering the image statistics of input MS bands and careful selection of optimal bands for DLCNN predictions when DLCNN architectures are restricted to three spectral channels.