Search Results (93)

This study presents a novel application of a hybrid regression–kriging (RK) and machine learning (ML) framework to impute missing tropospheric NO₂ data from the TROPOMI satellite over Taiwan during the winter months of January, February, and December 2022. The proposed approach combines geostatistical interpolation with nonlinear modeling by integrating RK with ML models—specifically comparing gradient boosting regression (GBR), random forest (RF), and K-nearest neighbors (KNN)—to determine the most suitable auxiliary predictor. This structure enables the framework to capture both spatial autocorrelation and complex relationships between NO₂ concentrations and environmental drivers. Model performance was evaluated using the coefficient of determination (r²), computed against observed TROPOMI NO₂ column values filtered by quality assurance criteria. GBR achieved the highest validation r² values of 0.83 for January and February, while RF yielded 0.82 and 0.79 in January and December, respectively. These results demonstrate the model’s robustness in capturing intra-seasonal patterns and nonlinear trends in NO₂ distribution. In contrast, models using only static land cover inputs performed poorly (r² < 0.58), emphasizing the limited predictive capacity of such variables in isolation. Interpretability analysis using the SHapley Additive exPlanations (SHAP) method revealed temperature as the most influential meteorological driver of NO₂ variation, particularly during winter, while forest cover consistently emerged as a key land-use factor mitigating NO₂ levels through dry deposition. By integrating dynamic meteorological variables and static land cover features, the hybrid RK–ML framework enhances the spatial and temporal completeness of satellite-derived air quality datasets. As the first RK–ML application for TROPOMI data in Taiwan, this study establishes a regional benchmark and offers a transferable methodology for satellite data imputation. Future research should explore ensemble-based RK variants, incorporate real-time auxiliary data, and assess transferability across diverse geographic and climatological contexts. Full article

Reanalysis air temperature data, characterized by temporal continuity but limited spatial resolution, are commonly downscaled to achieve higher spatial resolution to meet the demands of regional climatological studies and related research fields. However, when large spatial scale differences are involved, the adaptability of statistical downscaling models across different scales warrants further investigation. In this study, a stepwise downscaling method is proposed, employing multiple linear regression (MLR), Cubist regression tree, random forest (RF), and extreme gradient boosting (XGBoost) models to downscale the 3-hourly ERA5-Land reanalysis air temperature data at the resolution of 0.1° to that of 30 m. A comparative analysis was performed to evaluate the accuracy of downscaled ERA5-Land air temperature results obtained from the stepwise and the direct downscaling methods, based on observed air temperatures at meteorological stations and the spatial distribution of air temperature estimated by a remote sensing method. In addition, variations in the importance of driving factors across different spatial scales were examined. The results indicate that the stepwise downscaling method exhibits higher accuracy than the direct downscaling method, with a more pronounced performance improvement in winter. Compared with the direct downscaling method, the RMSE value of the MLR, Cubist, RF, and XGBoost models under the stepwise downscaling method were reduced by 0.48 K, 0.38 K, 0.48 K, and 0.50 K, respectively, at meteorological station locations. In terms of spatial distribution, the stepwise downscaling results demonstrate greater consistency with the estimated spatial distribution of air temperature, and it can capture air temperature variations across different land surface types more accurately. Furthermore, the stepwise downscaling method is capable of effectively capturing changes in the importance of driving factors across different spatial scales. These results generally suggest that the stepwise downscaling method can significantly improve the accuracy of air temperature downscaled from reanalysis data by adopting multiple resolutions as the intermediate downscaling process. Full article

The air temperature near the Earth’s surface is one of the most important meteorological and climatological parameters. Yet, accurate and timely readings are not available in significant parts of the world. The development and first validation of a methodology for the estimation of the near-surface air temperature (NSAT) is presented here. Machine learning and satellite products are at the core of the developed model. Land Surface Analysis Satellite Application Facility (LSA SAF) products related to Earth’s surface radiation, temperature and humidity budgets, albedo and land cover, along with static topography parameters and weather station measurements, are used in the analysis. A series of experiments showed that the Random Forest regression with 20 selected satellite and topography predictors was the optimum selection for the estimation of the NSAT. The mean absolute error (MAE) of the NSAT estimation model was 0.96 °C, while the mean biased error (MBE) was −0.01 °C and the R² was 0.976. Limited seasonality was present in the efficiency of the model, while an increase in errors was noted during the first morning and afternoon hours. The topography influence in the model efficiency was rather limited. Cloud-free conditions were associated to only marginally smaller errors, supporting the applicability of the model under both cloud-free and cloudy conditions. Full article

In the context of global warming, climate strongly affects forest fires. With long-term and strict fire prevention policies, China has become a unique test arena for comprehending the role of climatic variables in affecting forest fires. Here, using GIS spatial analysis, Pearson correlation, and geographical detector, the climate drivers of forest fires in China are revealed using the 2003–2022 active fire data from the MODIS C6 and climate products from CHELSA (Climatologies at high resolution for the Earth’s land surface areas). The main conclusions are as follows: (1) In total, 82% of forest fires were prevalent in the southern and southwestern forest regions (SR and SWR) in China, especially in winter and spring. (2) Forest fires were mainly distributed in areas with a mean annual temperature and annual precipitation of 14~22 °C (subtropical) and 800~2000 mm (humid zone), respectively. (3) Incidences of forest fires were positively correlated with temperature, potential evapotranspiration, surface downwelling shortwave flux, and near-surface wind speed but negatively correlated with precipitation and near-surface relative humidity. (4) Temperature and potential evapotranspiration dominated the roles in determining spatial variations of China’s forest fires, while the combination of climate variables complicated the spatial variation. This paper not only provides new insights on the impact of climate drives on forest fires, but also offers helpful guidance for fire management, prevention, and forecasting. Full article

Based on statistical analyses applied to official data from the Digital Atlas of Disasters in Brazil over the last 25 years, we evidenced a consistent intensification in the annual occurrence of natural disasters in the state of Pará, located in the eastern Brazilian Amazon. The quantitative comparison between the averages of the most intense period of disasters (2017 to 2023) and the earlier years (1999 to 2016) revealed a remarkable percentage increase of 473%. Approximately 81% of the state’s municipalities were affected, as indicated by disaster mapping. A clear seasonal pattern was observed, with Hydrological disasters (Inundations, Flash floods, and Heavy rainfall) peaking between February and May, while Climatological disasters (Droughts and Forest fires) were most frequent from August to October. The catastrophic impacts on people and the economy were documented, showing a significant rise in the number of homeless individuals and those directly affected, alongside considerable material damage and economic losses for both the public and private sectors. Furthermore, we conducted a comprehensive composite analysis on the tropical ocean–atmosphere dynamic structure that elucidated the various triggering mechanisms of disasters arising from Inundations, Droughts, and Forest fires (on seasonal scale), and Flash floods and Heavy rainfall (on sub-monthly scale) in Pará. The detailed characterization of disasters on a municipal scale is relevant in terms of the scientific contribution applied to the strategic decision-making, planning, and implementation of public policies aimed at early risk management (rather than post-disaster response), which is critical for safeguarding human well-being and strengthening the resilience of Amazonian communities vulnerable to climate change. Full article

Despite significant progress in recent decades, air pollution remains the leading environmental cause of premature death in Europe. Urban populations are particularly exposed to high concentrations of air pollutants, such as particulate matter smaller than 10 µm (PM₁₀). Understanding the spatiotemporal variations of PM₁₀ is essential for developing effective control strategies. This study aimed to enhance PM₁₀ prediction models by integrating landscape metrics as ecological indicators into our previous models, assessing their significance in monthly average PM₁₀ concentrations, and analyzing their correlations with PM₁₀ air pollution across European urban landscapes during heating (cold) and non-heating (warm) seasons. In our previous research, we only calculated the proportion of land uses (PLANDs), but according to our current research hypothesis, landscape metrics have a significant impact on PM₁₀ air quality. Therefore, we expanded our independent variables by incorporating landscape metrics that capture compositional heterogeneity, including the Shannon diversity index (SHDI), as well as metrics that reflect configurational heterogeneity in urban landscapes, such as the Mean Patch Area (MPA) and Shape Index (SHI). Considering data from 1216 European air quality (AQ) stations, we applied the Random Forest model using cross-validation to discover patterns and complex relationships. Climatological factors, such as monthly average temperature, wind speed, precipitation, and mean sea level air pressure, emerged as key predictors, particularly during the heating season when the impact of temperature on PM₁₀ prediction increased from 5.80% to 22.46% at 3 km. Landscape metrics, including the SHDI, MPA, and SHI, were significantly related to the monthly average PM₁₀ concentration. The SHDI was negatively correlated with PM₁₀ levels, suggesting that heterogeneous landscapes could help mitigate pollution. Our enhanced model achieved an R² of 0.58 in the 1000 m buffer zone and 0.66 in the 3000 m buffer zone, underscoring the utility of these variables in improving PM₁₀ predictions. Our findings suggest that increased urban landscape complexity, smaller patch sizes, and more fragmented land uses associated with PM₁₀ sources such as built-up areas, along with larger and more evenly distributed green spaces, can contribute to the control and reduction of PM₁₀ pollution. Full article

Freezing precipitation and the resultant ice glaze can have catastrophic impacts on urban infrastructure, the environment, forests, and various industries, including transportation, energy, and agriculture. In this study, we develop and evaluate regional algorithms for detecting freezing precipitations in the Far East, utilizing the ERA5 reanalysis dataset from the European Centre for Medium-Range Weather Forecasts, along with standard meteorological observations for 20 cold seasons (September–May) from 2004 to 2024. We propose modified diagnostic algorithms based on vertical atmospheric temperature and humidity profiles, as well as near-surface characteristics. Additionally, we apply a majority voting ensemble (MVE) technique to integrate outputs from multiple algorithms, thereby enhancing classification accuracy. Evaluation of detection skills shows significant improvements over the original method developed at the Finnish Meteorological Institute and the ERA5 precipitation-type product. The MVE-based method demonstrates optimal verification statistics. Furthermore, the modified algorithms validly reproduce the spatially averaged inter-annual variability of freezing precipitation activity in both continental (mean correlation of 0.93) and island (correlation of 0.54) regions. Overall, our findings offer a more effective and valuable tool for operational activities and climatological assessments in the Far East. Full article

The Gedeo zone agroforestry systems are the main source of Ethiopia’s coffee beans. However, land-use and suitability analyses are not well documented due to complex topography, heterogeneous agroforestry, and lack of information. This research aimed to map the coffee coverage and identify land suitability for coffee plantations using remote sensing, Geographic Information Systems (GIS), and the Analytical Hierarchy Process (AHP) in the Gedeo zone, Southern Ethiopia. Remote sensing classifiers often confuse agroforestry and plantations like coffee cover with forest cover because of their similar spectral signatures. Mapping shaded coffee in Gedeo agroforestry using optical or multispectral remote sensing is challenging. To address this, the study identified and mapped coffee coverage from Sentinel-1 data with a decibel (dB) value matched to actual coffee coverage. The actual field data were overlaid on Sentinel-1, which was used to extract the raster value. Pre-processing, classification, standardization, and reclassification of thematic layers were performed to find potential areas for coffee plantation. Hierarchy levels of the main criteria were formed based on climatological, edaphological, physiographic, and socioeconomic factors. These criteria were divided into 14 sub-criteria, reclassified based on their impact on coffee growing, with their relative weights derived using AHP. From the total study area of 1356.2 km², the mapped coffee coverage is 583 km². The outcome of the final computed factor weight indicated that average annual temperature and mean annual rainfall are the primary factors, followed by annual mean maximum temperature, elevation, annual mean minimum temperature, soil pH, Land Use/Land Cover (LULC), soil texture, Cation Exchange Capacity (CEC), slope, Soil Organic Matter (SOM), aspect, distance to roads, and distance to water, respectively. The identified coffee plantation potential land suitability reveals unsuitable (413 km²), sub-suitable (596.1 km²), and suitable (347.1 km²) areas. This study provides comprehensive spatial details for Ethiopian cultivators, government officials, and agricultural extension specialists to select optimal coffee farming locations, enhancing food security and economic prosperity. Full article

Characterized as a climatologically sensitive region, the southeastern Tibetan Plateau (STP) is an ideal location for dendrochronological research. Here, five tree-ring width (TRW) chronologies were developed: three for Picea likiangensis along altitudinal gradients from 3600 to 4400 m a.s.l. and two for Sabina saltuaria and Abies squamata from 4200 m a.s.l. Significant differences in the growth rates and age composition of Picea likiangensis were observed at various elevation gradients. The chronology statistics (mean sensitivity, etc.) fluctuated with the elevation gradient. Picea likiangensis showed distinct growth patterns in response to climatic variability along the altitude gradient: the minimum temperature influenced tree growth at lower and middle altitudes, while higher altitudes were affected by precipitation. The radial growth of different tree species growing in the same region is controlled by the same climatic factors. Sabina saltuaria and Abies squamata exhibited similar growth responses to Picea likiangensis. Stand conditions (wind speeds, slope, and elevation) and biotic factors (the depth of root, forest type, tree age, and sensitivity) can partially explain why the ring width–climate relationships change with altitude. Full article

Climatic seasonality has lacked research attention in terms of global tropical forests, where it impacts vegetation productivity, biodiversity, and hydrological cycles. This study employs two methods—climatological anomalous accumulation (CAA) and potential evapotranspiration (PET) threshold—to detect the climatic seasonality of global tropical forests, including the onset and duration of wet seasons. Spatial clustering based on the length of the wet season is used to delineate smaller regions within the tropical forest areas to observe their precipitation patterns. The results show that these methods effectively reveal more homogeneous regions and their respective rainfall patterns. In particular, we found that the wet season in Amazon forests detected by the CAA method is more uniform in space than the PET threshold, but the global tropical forest regions divided by the CAA method on average contain more complex climates than the PET threshold. Moreover, the year-round abundant precipitation in Southeast Asia, which is strongly influenced by monsoons, presents challenges for wet season detection. Overall, this work provides an objective perspective for understanding the climatic seasonality changes in tropical forests and lays a scientific foundation for future forest management and the development of adaptation strategies to global climate change. Full article

The first case of dengue fever (DF) in Saudi Arabia appeared in 1993 but by 2022, DF incidence was 11 per 100,000 people. Climatologic and population factors, such as the annual Hajj, likely contribute to DF’s epidemiology in Saudi Arabia. In this study, we assess the impact of these variables on the DF burden of disease in Saudi Arabia and we attempt to create robust DF predictive models. Using 10 years of DF, weather, and pilgrimage data, we conducted a bivariate analysis investigating the role of weather and pilgrimage variables on DF incidence. We also compared the abilities of three different predictive models. Amongst weather variables, temperature and humidity had the strongest associations with DF incidence, while rainfall showed little to no significant relationship. Pilgrimage variables did not have strong associations with DF incidence. The random forest model had the highest predictive ability (R² = 0.62) when previous DF data were withheld, and the ARIMA model was the best (R² = 0.78) when previous DF data were incorporated. We found that a nonlinear machine-learning model incorporating temperature and humidity variables had the best prediction accuracy for DF, regardless of the availability of previous DF data. This finding can inform DF early warning systems and preparedness in Saudi Arabia. Full article

Fire danger poses a pressing threat to ecosystems and societies worldwide. Adequate preparation and forewarning can help reduce these threats, but these rely on accurate prediction of extreme fire danger. With the knowledge that climatic conditions contribute heavily to overall fire danger, this study evaluates the skill with which episodes of extreme fire danger in Australia can be predicted from the activity of large-scale climate driver patterns. An extremal dependence index for extreme events is used to depict the historical predictive skill of the Australian Bureau of Meteorology’s subseasonal climate prediction system in replicating known relationships between the probability of top-decile fire danger and climate driver states at a lead time of 2–3 weeks. Results demonstrate that the El Niño Southern Oscillation, Southern Annular Mode, persistent modes of atmospheric blocking, Indian Ocean Dipole and Madden-Julian Oscillation are all key for contributing to predictability of fire danger forecasts in different regions during critical fire danger periods. Northwest Australia is found to be particularly predictable, with the highest mean index differences (>0.50) when certain climate drivers are active, compared with the climatological index mean. This integrated approach offers a valuable resource for decision-making in fire-prone regions, providing greater confidence to users relying on fire danger outlooks for key management decisions, such as those involved in the sectors of national park and forest estate management, agriculture, emergency services, health and energy. Furthermore, the results highlight strengths and weaknesses in both the Australian Fire Danger Rating System and the operational climate model, contributing additional information for improving and refining future iterations of these systems. Full article

This study aims to identify the optimal regression techniques for downscaling among ten commonly used methods in climatology, including SVR, LinearSVR, LASSO, LASSOCV, Elastic Net, Bayesian Ridge, RandomForestRegressor, AdaBoost Regressor, KNeighbors Regressor, and XGBRegressor. For the Köppen climate classification system, including A (tropical), B (dry), C (temperate), and D (continental), synoptic station data were collected. Furthermore, for the purpose of downscaling, a general circulation model (GCM) had been utilized. Additionally, to enhance the performance of downscaling accuracy, mutual information (MI) was employed for feature selection. The downscaling performance was evaluated using the coefficient of determination (DC) and root mean square error (RMSE). Results indicate that SVR had superior performance in tropical and dry climates and LassoCV with RandomForestRegressor had better results in temperate and continental climates. Full article

Regression algorithms are regularly used for improving the accuracy of satellite precipitation products. In this context, satellite precipitation and topography data are the predictor variables, and gauged-measured precipitation data are the dependent variables. Alongside this, it is increasingly recognised in many fields that combinations of algorithms through ensemble learning can lead to substantial predictive performance improvements. Still, a sufficient number of ensemble learners for improving the accuracy of satellite precipitation products and their large-scale comparison are currently missing from the literature. In this study, we work towards filling in this specific gap by proposing 11 new ensemble learners in the field and by extensively comparing them. We apply the ensemble learners to monthly data from the PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks) and IMERG (Integrated Multi-satellitE Retrievals for GPM) gridded datasets that span over a 15-year period and over the entire contiguous United States (CONUS). We also use gauge-measured precipitation data from the Global Historical Climatology Network monthly database, version 2 (GHCNm). The ensemble learners combine the predictions of six machine learning regression algorithms (base learners), namely the multivariate adaptive regression splines (MARS), multivariate adaptive polynomial splines (poly-MARS), random forests (RF), gradient boosting machines (GBM), extreme gradient boosting (XGBoost) and Bayesian regularized neural networks (BRNN), and each of them is based on a different combiner. The combiners include the equal-weight combiner, the median combiner, two best learners and seven variants of a sophisticated stacking method. The latter stacks a regression algorithm on top of the base learners to combine their independent predictions. Its seven variants are defined by seven different regression algorithms, specifically the linear regression (LR) algorithm and the six algorithms also used as base learners. The results suggest that sophisticated stacking performs significantly better than the base learners, especially when applied using the LR algorithm. It also beats the simpler combination methods. Full article

Salt pans are unique wetland ecosystems. In the Austrian Seewinkel region, salt pans are in an increasingly vulnerable state due to groundwater drainage and heightened climatic pressures. It is crucial to model how seasonal and long-term hydrological and climatological variations affect the salt pan dynamics in Seewinkel, yet a comprehensive understanding of the driving processes is lacking. The goal of this study is to develop random forest machine learning models driven by hydrological and meteorological data that allow us to predict in early spring (March) of each year the inundation state in the subsequent summer and fall. We utilize Earth observation data from Landsat 5 (L5), 8 (L8), and 9 (L9) to derive the time series of the inundation state for 34 salt pans for the period 1984–2022. Furthermore, we demonstrate that the groundwater level observed in March is the strongest predictor of the salt pan inundation state in summer and fall. Utilizing local groundwater data yields a Matthews correlation coefficient of 0.59. Models using globally available meteorological data, either instead of or in addition to groundwater data, provide comparable results. This allows the global transfer of the approach to comparable ecosystems where no in situ data are available. Full article