Accurate spatio-temporal information on the soil water balance is critical for an efficient and sustainable irrigation. Large effort requirements limit the applicability of complex simulations for precision irrigation. The spatially distributed application of one-dimensional models can reconcile the need for precise soil water balance simulations with the complexity of root-zone water flow processes. This study uses HYDRUS-1D to simulate the daily depth-specific (0 cm to 60 cm, in 10 cm increments) soil water balance from 1st April to 30th September 2021 (2022). Simulations at 70 m spatial resolution covered a 1600 ha farm in Mecklenburg-Western Pomerania, Germany. Results were validated against in-situ soil water content (SWC) and two remotely-sensed SWC data sets ("Soil Moisture Active Passive", SMAP; Sentinel-1, S1-SWC). Further analysis explored crop-specific irrigation efficiencies and potential farm-scale water savings. Spatially distributed HYDRUS-1D simulations showed good accuracy compared to in-situ SWC (RMSEmean = 0.020 m3 m-3; MAEmean = 0.017 m3 m-3; R2mean = 0.676; bias =-0.008 m3 m-3). The agreement with remotely-sensed SWC was moderate to weak (RMSEmean = 0.059 (0.150) m3 m-3, MAEmean = 0.049 (0.123) m3 m-3, R2mean = 0.208 (0.141), mean bias = 0.021 (0.108) m3 m-3 for SMAP (S1-SWC)). Irrigation efficiencies were 65.0 % (potato), 47.3 % (wheat), 40.5 % (rye), and 58.2 % (sugar beet). Potential water savings amounted to 87,006.9 m3 (11.2 % of total irrigation water; 2021) and 71,396.6 m3 (10.4 %; 2022). The proposed approach reduces the trade-offs between accurately representing the soil water balance in the root-zone and keeping the practical effort reasonable.

Urban green spaces (UGS) are widely promoted as nature-based solutions to reduce heat risk in Urban Heat Islands. To quantify the cooling effects of UGS the use of remote sensing-based Land Surface Temperature (LST) indicators, such as the Park Cool Island Intensity (PCII) or Cooling Effect Intensity (CEI), has proven beneficial. However, these indicators heavily depend on the urban reference the UGS LST is compared with. This limits the comparability of the cooling performances of UGS across cities and urban forms and furthermore obscures the fact that UGS may themselves experience warming effects from adjacent urban areas in return, a mechanism which is still underrepresented in quantitative research.In this study, we therefore develop and test a transferable approach for determining UGS cooling deficits (ΔLST) by consistently deriving UGS LST relative to a standardized rural baseline derived from Local Climate Zones (LCZ). Using a on a long-term (1984–2025) Landsat LST time series, we analyze several municipalities in Hesse, Germany, and compare summer patterns of ΔLST within and between municipalities.UGS are characterized by size, tree cover, and vegetation state according to the Normalized Difference Vegetation Index. Surrounding urban structure is quantified using buffer-ring metrics and indicators of built form and land cover (e.g., imperviousness and building density) to capture how different urban contexts modulate ΔLST. This allows the evaluation of the warming effects of different urban areas on different UGS. To disentangle drivers of cooling deficits, we fit multivariate models that account for nested spatial structure (mixed-effects regression) and complement them with a nonlinear benchmark (e.g., random forest). Finally, we analyze to what degree antecedent weather conditions (air temperature, precipitation, and relative humidity) in different time periods (e.g., 7, 14, 21 days) prior to a Landsat acquisitions modulate ΔLST.This approach provides a transferable, planning-relevant metric that allows UGS to be classified not only as "cool" or "warm," but also as more or less effective relative to a clearly defined rural reference state. This improves comparability across time, space, and different urban structures, and creates a robust basis for prioritized adaptation measures.

This study presents a novel framework for quantifying uncertainties and variabilities related to the monitoring of crop phenology via Synthetic Aperture Radar (SAR) time series at the field scale. Therefore, the study investigated multi-orbit, multi-feature time series derived from Sentinel-1 (S1) VV/VH polarizations. This multi-feature approach encompasses backscatter intensity, interferometric coherence and alpha/entropy decomposition features. Crop phenology tracking is crucial for assessing agricultural resilience under climate change, yet existing approaches face challenges due to uncertainties and variability in SAR signal interpretation as well as in situ data. Building on previous landscape-level analyses, this work introduces the concept of trackability, defined as the temporal range during which SAR-derived time-series metrics (TSM), such as breakpoints in backscatter intensity or interferometric coherence, align with key phenological stages (e.g., stem elongation in winter wheat). A growing degree day (GDD)-based normalization contextualizes field-specific deviations relative to landscape averages, enabling quantification of uncertainties inherent in both SAR signals and ground observations. The framework captures the spatio-temporally variable nature of crop development by estimating the first and last phenologically relevant TSM occurrence within a defined uncertainty window, thus providing relational and relative indicators of phenological tracking. This approach reduces dependencies of extensive in situ data and enhances comparability across studies with differing SAR processing methods and their acquisition geometries. Results reproduce known feature-stage relationships (e.g., tracking for stem elongation by interferometric coherence) and reveal inter-seasonal variability influenced by weather conditions and acquisition parameters. On average relevant TSM occurrences were found at approximately 90% of GDD progression of in situ reported phenological stages, while systematic differences of around 5% by relative orbit were discovered. The study highlights the potential of integrating multiple S1 features and orbits without optimization-induced information loss, producing quality masks that identify optimal tracking performance at the field level. This framework advances SAR-based phenology monitoring by offering scalable, transferable insights for precision agriculture, while practical implementation still requires detailed field boundaries and early-season crop management information.

Central Asia is highly vulnerable to increasing drought frequency and intensity due to climate change, strong dependence on irrigated agriculture, and complex transboundary water systems. Effective drought risk management in the Aral Sea Basin (ASB), therefore, requires timely, spatially explicit, and policy-relevant information that can be accessed and interpreted by water managers, environmental experts, and hydrometeorological services. In this study, we present Droughtmap-ASB, an operational, Earth observation–based drought monitoring and decision support tool designed to support drought assessment, near real time warning, and policy-relevant planning across multiple spatial and temporal scales.Droughtmap-ASB integrates satellite-derived vegetation and evaporative stress indicators with climate reanalysis data to provide a comprehensive characterization of agricultural and meteorological drought. The core framework combines Sentinel-3–based NDVI and land surface temperature-dependent Evaporative Stress Index (ESI) with a dynamic ten-year baseline to compute a Drought Severity Index (DSI), capturing drought onset, duration, and intensity. In addition, the system implements SPI, SPEI, and the Hydrothermal Coefficient (HTC) derived from ERA5 reanalysis data, enabling consistent assessment of meteorological drought conditions. Drought conditions are classified into eight standardized drought classes ranging from initial mild to long-term severe drought.A key strength of Droughtmap-ASB is its multi-scale spatial design, which allows analyses at resolutions of 5 × 5 km grids up to rayon, oblast, national, and basin-wide levels, ensuring compatibility with both operational water management and policy frameworks. The web-based dashboard provides interactive visualization, while an automated bulletin module generates bi-weekly, monthly, and seasonal drought reports, supporting routine information dissemination for end-users.By translating complex Earth observation data into actionable indicators, standardized drought classes, and policy-ready bulletins, Droughtmap-ASB bridges the gap between scientific monitoring and decision-making. The tool supports evidence-based water allocation, agricultural risk management, and climate adaptation planning, contributing to improved drought preparedness and resilience in Central Asia.Keywords: Agricultural drought; Meteorological drought; Web-based drought monitoring; Earth observation data; Drought bulletins; Central Asia

WaterBalanceR is an R package developed within the BMEL-funded project AgriSens-DEMMIN 4.0 and is distributed under the AGPL-3.0-only license. It offers the calculation of spatially distributed, daily water balance maps for starch potatoes at a precision farming scale. This is achieved by combining Normalized Distance Vegetation Index (NDVI) data derived from multispectral UAV DJI Phantom 4 M, PlanetScope and/or Sentinel-2 imagery with freely available meteorological data together with reference evapotranspiration (ETref) from the German Weather Service (DWD) using the FAO56 Penman-Monteith method along with specified or downloaded irrigation data from earlier events. The correlation of the underlying measured and modelled crop evapotranspiration was proven with R² ≥ 0.95*** and 0.99 ≥ slope ≥ 0.96 with a low noise of RSE ≤ 0.31 for five different POIs in 2023.

Thermal remote sensing is a valuable tool for assessing Surface Urban Heat Islands (SUHI). To quantify the SUHI intensity, the Urban Thermal Field Variance Index (UTFVI) is increasingly used as a proxy for urban heat distribution, e.g., by public authorities in Germany. The UTFVI is an ordinal-scaled metric that shows the spatial variability of LST in relation to the average LST of an area of interest. Numerous scientific studies utilize the UTFVI for detecting spatiotemporal increases in SUHI intensities attributed to Land Use/Land Cover (LULC) changes, such as rapid urbanization. However, UTFVI analyses often rely on only a few time steps over extended periods, ignoring the effects of weather patterns on actual UTFVI distributions. To address this research gap, this study investigates the influence of weather conditions of varying durations (up to 21 days) on seasonal UTFVI distributions in four Hessian municipalities (Germany) with less than 300,000 inhabitants and only minor LULC changes in the urban area over time. The analysis is based on more than 100 Landsat 4–9 Level‑2 datasets spanning a 40-year period. To reduce rural influences, only urban areas are considered. Results reveal high seasonal and intra-seasonal variability in UTFVI. Statistical tests (Friedman and Wilcoxon) show significant differences of UTFVI distributions even within one summer (2023). Spearman’s rank correlation coefficients indicate that spatial patterns of the UTFVI are influenced by the temperature intensity of preceding weather phases: short-term warming leads to an increase of the UTFVI categories indicating high LST levels, while they are less frequent during prolonged warmth. The presented study provides for the first time a comprehensive analysis of long-term UTFVI developments that focusses on factors altering the UTFVI which have not been investigated so far. These new insights support a better understanding and a more distinct interpretation of the UTFVI for potential users.

Accurate spatio-temporal information on the soil water balance is critical for an efficient and sustainable irrigation. Recent irrigation scheduling approaches are often limited to a representation of (i) the local or point scale soil water balance by in-situ measurements, (ii) solely surface soil water contents at a coarse spatial resolution by microwave remote sensing technologies, or (iii) only selected components of the soil water balance by simple crop evapotranspiration models. To reconcile the need for accurate estimates of different components of the soil water balance with feasible effort, this study proposes the application of physically-based one-dimensional soil water balance models in a spatially-distributed manner.The HYDRUS-1D software environment is applied at 70 m spatial resolution across a 1,600 ha study farm in Mecklenburg-Western Pomerania, Germany, with heterogeneous soil textures and different crops. Depth-specific (0 cm to 60 cm, in 10 cm increments) soil water balance simulations were conducted from 1st April to 30th September 2021 and 2022 to estimate the soil water content, plant available water content, infiltration, crop evapotranspiration, root water uptake, and deep percolation, at daily intervals. Simulated soil water contents were validated against in-situ measurements and two microwave remote sensing surface soil water content datasets (“Soil Moisture Active Passive”, SMAP; Sentinel-1, S1-SWC). Spatially distributed irrigation demands and irrigation timings at daily intervals, crop-specific irrigation efficiencies and potential farm-scale water savings are estimated using the simulated soil water balance to explore the contribution of this simulation framework for precision irrigation.The average simulation performance metrices were Root Mean Square Error (RMSE) = 0.020 m3 m-3, Mean Absolute Error (MAE) = 0.017 m3 m-3, coefficient of determination (R²) = 0.676, and bias = -0.008 m3 m-3, showing a good accuracy of spatially-distributed HYDRUS-1D simulations. The agreement with remotely-sensed data was moderate to weak (RMSEmean = 0.059 (0.150) m3 m-3, MAEmean = 0.049 (0.123) m3 m-3, R2mean = 0.208 (0.141), mean bias = 0.021 (0.108) m3 m-3 for SMAP (S1-SWC)). Average crop specific irrigation efficiencies were 65.0% (potato), 47.3% (wheat), 40.5% (rye), and 58.2% (sugar beet). Potential water savings amounted to 87,006.9 m³ (11.2 % of the applied irrigation water; 2021) and 71,396.6 m³ (10.4 %; 2022).The proposed simulation framework offers an easy-to-adopt and physically-based foundation for the estimation of crop-specific irrigation demands and irrigation timings at high spatial resolution. Further accuracy improvements by using depth-specific remote-sensing derived soil water contents (“Soil Water Index”) for model calibration are under ongoing investigation.

Soil moisture (SM) is a key parameter for irrigation monitoring, scheduling, and supporting precision agriculture. In this study, we used Sentinel-1 synthetic aperture radar (SAR) and Sentinel-2 optical data to estimate SM at high spatial resolution (20 m) in the Lower Chenab Canal Command (LCC) area of Punjab, Pakistan. To achieve this, we applied a semiempirical water cloud model (WCM) using both the VV and VH polarizations from SAR data. Additionally, two widely used machine learning (ML) models, random forest (RF) and support vector machines (SVM), were employed to estimate SM at the field scale. To assess model reliability and transferability, reference data from two field sites were split using two approaches: (1) stratified random sampling, with 50

Soil organic carbon (SOC) estimation often lacks multi-sensor integration with improved bare soil compositing, and robust uncertainty assessment. This study integrates 6 years multi-temporal Synthetic Aperture Radar (SAR) from Sentinel-1 and optical data from Sentinel-2, with confidence interval (CI)-based bare soil compositing for SOC prediction in an agricultural landscape in northeast Germany to enhance SOC estimation at the example of an agricultural area in northeast Germany (Demmin). Four Random Forest models, each incorporating terrain attributes (TA), were developed to assess the independent contributions and combined potential of optical and SAR data. The local soil samples collected from 2013 to 2022 were split into training (70%) and testing (30%) datasets, with independent validation performed using samples from 2024 across the study area. An uncertainty map based on 100 model repetitions accompanies the final SOC map. Results demonstrate that combining SAR and optical data slightly improves model calibration, while CI-based compositing further enhances accuracy. The best-performing model, based on important features identified through recursive feature elimination, achieved an R-2 of 0.79 and RPD of 2.23 in independent validation. SAR data exhibited higher uncertainty due to its sensitivity to surface conditions, still produced satisfactory SOC mapping (R-2 = 0.57, RPD = 1.54, RMSE = 6.01 g/kg). Although SAR data penetrates only the first few centimeters of soil (<50 & micro;m for optical data), the results provide a reliable basis for estimating topsoil SOC.

Groundwater-dependent vegetation (GDV) plays a vital role in maintaining biodiversity and ecosystem services in the Mediterranean biome but is increasingly threatened by climate and land use change. Large-scale GDV mapping in arid regions is critical for effective conservation and water resource management but remains challenging due to limited ground-truth data and the lack of high-resolution remote sensing-based spatial models. In this study, we mapped GDV across the world's five Mediterranean climate regions using a multi-level approach combining species-occurrence, vegetation-plot, and remote-sensing data. At the species level, we compiled a global list of phreatophyte and groundwater-associated species, which we used to extract occurrence records from GBIF. At the community level, we classified vegetation-plot data from the sPlot database based on phreatophyte presence and coverage, resulting in a unique ground-truth species-community dataset. At the biome level, we trained Random Forest models with eleven predictor variables in order to map GDV distribution at 30 m resolution for the period 2018-2023. We identified 482,000 km2 of GDV, covering 28% of the study area and predicted GDV hotspots for the Western Iberian Peninsula, Southern France, California, Chile, and the West Coast of Australia. The largest absolute area of GDV was found in the Mediterranean Basin (306,000 km2), while the highest relative coverage was found in California (42%) and Chile (40%). Notably, only about a quarter of Mediterranean GDV lies within protected areas. Key environmental predictors include soil sand content, dry season vegetation vitality, and elevation. By integrating species, community, and remote-sensing data, our high-resolution GDV map provides a crucial basis for monitoring ecosystem response to global change, conservation planning, sustainable groundwater management, and risk assessment in the drought-stressed Mediterranean biome.

Abstract Groundwater-dependent ecosystems are biodiversity hotspots that provide habitat for specialised species. The EU Water Framework Directive (WFD) stresses the importance of identifying and protecting these ecosystems. However, they remain poorly mapped in temperate regions, as most studies have focused on (semi-) arid regions, where groundwater use by vegetation is both more prevalent and easier to detect from remote sensing. In this study, we transfer mapping approaches for groundwater-dependent vegetation (GDV) from dry climates into a novel framework for humid climates. To do so, we integrated, ECOSTRESS evapotranspiration data, together with high-resolution remote sensing data, regional geospatial data and field data to identify GDV. To test our framework, we trained and validated Random Forest models with eight predictor variables using 166 ground-truth vegetation plots to map GDV in Saxony-Anhalt (Germany). The final model achieved an overall accuracy of 0.97, identifying 2,067 km 2 (41%) of GDV. Currently, only 19% are protected under the EU WFD. The proposed mapping framework offers a new solution for identifying GDV in temperate regions. The new GDV maps can contribute to managing groundwater resources and preserving biodiversity hotspots in regions facing increasing droughts, ultimately supporting implementation of the EU WFD.

Trees on agricultural land are key structural components of agroecosystems, contributing to essential ecosystem services like microclimate regulation, erosion control, biodiversity conservation, and the mitigation of climate-induced abiotic stresses, thereby enhancing the resilience of agricultural landscapes. However, existing inventories are often outdated, incomplete, and lack the spatial resolution necessary for in-depth analysis and effective decision-making. Therefore, we apply a semantic segmentation approach based on the U-Net architecture, to quantify the current spatial distribution of trees on agricultural lands across southern Saxony-Anhalt (approximately 4,000 km²). The model is based on official digital orthophotos (DOP) with 20 cm spatial resolution and a spectral resolution of four channels (RGBI). Given the large study area and the coarse repetition rate of aerial imagery, we further evaluate model performance across different acquisition dates, ranging from the beginning of the 2023 vegetation period (30.04. - spring) to the peak of the 2024 vegetation period (29.08. – late summer).Training data generation uses a semi-automatic workflow: a normalized surface model is clipped into 512×512-pixel tiles, filtered to retain objects >4m height, and masked to exclude impervious surfaces. This produces 7,894 tiles containing 14,360 annotated features, which are manually verified against true-color imagery. An independent test set is created through manual digitization of agricultural trees, stratified by image acquisition date. Model performance is evaluated using Precision, Recall, F1-score, and Intersection over Union (IoU).The dataset is split 70/30 for training/validation. Input data includes four channels (RGBI) and the Normalized Difference Vegetation Index (NDVI) as a fifth channel. Data augmentation applies random horizontal/vertical flips and rotations (±15°). The U-Net model is trained using focal Tversky loss (weighted to penalize both false positives and negatives) and the Adam optimizer with default learning rate.Lowest model errors were reached after 48 epochs. The best-performing model is selected and subsequently applied to each DOP tile intersecting the study area, resulting in predictions for 1182 DOP tiles. First validation results on approximately 8000 reference polygons show an average F1-Score of 0.5 which is comparable to recent studies. A total area of 195 km² of trees on agricultural land are mapped. Despite the heterogeneity of acquisition dates, the model produces accurate segmentations and successfully identifies trees on agricultural land in different compositions. The results indicate that semiautomatic training data generation can compensate for seasonal variability in aerial images, which often hinders the application of deep learning models to larger spatial scales.

In the face of unabated urban expansion, understanding the intrinsic characteristics of landscape structure is pertinent to preserving ecological diversity and managing the supply of ecosystem services. This study integrates machine-learning-based geospatial and landscape ecological techniques to assess the dynamics of landscape structure in cities of the rainforest (Akure and Owerri) and Guinea savanna (Makurdi and Minna) ecological regions of Nigeria between 1986 and 2022. Supervised classification using the random forest (RF) machine-learning classifier was performed on Landsat images on the Google Earth Engine (GEE) platform, and landscape metrics were calculated with FRAGSTATS to assess landscape composition, configuration, and connectivity. The results reveal a consistent pattern of urban expansion in all four cities at varying intensities. The proportion of the built-up class exhibited positive correlations with the largest patch index (r = 0.86, p < 0.05) and aggregation (r = 0.39, p < 0.05), indicating a concurrent rise in landscape densification as urban expansion persists. For the agricultural and vegetation landscapes, landscape proportion correlates negatively with fragmentation (r = −0.88, p < 0.05) and connectivity (r = −0.77, p < 0.05), but positively with aggregation (r = 0.89, p < 0.05). The increased patch density indicates a rising magnitude of landscape fragmentation and heterogeneity over time with varying implications for ecosystem functioning. These findings demonstrate the complex interplay between urbanisation and ecological processes within and across different ecoregions, highlighting the need for targeted ecological management, sustainable urban planning, and regionally informed landscape conservation strategies.

Additional water supply by irrigation is increasingly applied to high-value crops as potatoes in humid and temperate biomes to maximize yield and quality. In the light of increasing drought risks, innovative adjustments are required in irrigation scheduling to increase the efficiency and sustainability of supplemental irrigation. Hence, this study aims to (i) propose an easy-to-adopt, soil-property based approach to estimate the application efficiency under open-field conditions, and (ii) compare the accuracy of simulated soil moisture dynamics provided by two different physically based water balance models (AMBAV, HYDRUS-1D). Variable rate gross irrigation depths were applied during the growing seasons 2021 and 2022 on two loamy-sand potato fields (27 ha and 35 ha) in Mecklenburg-Western Pomerania, Germany. The average application efficiency of the utilized gun sprinkler irrigation system was 77.2%, ranging between 71.6% and 81.4% (71.3 %-80.9 %) in 2021 (2022). It was higher than expected in agricultural practice, which allowed for potential seasonal water savings of 6.3 L m(-2) on loamy sands, in comparison with the customary potato irrigation. HYDRUS-1D provided a significantly increased accuracy of simulated soil moisture dynamics, when compared to AMBAV. The overall average model performance metrics in comparison with in-situ measurements were: Root Mean Square Error (RMSE): 0.044 (AMBAV), 0.021 (HYDRUS-1D); Mean Absolute Error (MAE): 0.041 (AMBAV), 0.018 (HYDRUS-1D); coefficient of determination (R-2): 0.567 (AMBAV), 0.649 (HYDRUS-1D). Simulated soil moisture dynamics of both models showed greater variations in topsoil (0 cm-30 cm) than in subsoil (40 cm-60 cm) and diverged in their accuracy in different months of the growing season of the potato crop. These novel insights on unproductive water losses during irrigation and on the simulation accuracy of different physically based soil water balance models significantly improve the irrigation scheduling under open-field conditions. The results of this study may hence serve as important cornerstones for establishing water-saving irrigation strategies in the humid and temperate biomes of Central Europe.

Climate change and increasing weather and seasonal dynamics challenge agricultural landscapes. To cope with this challenge information on crop performance is key. This study presents a novel framework for bridging landscape-scale vegetation dynamics with field-level crop phenology using Sentinel-1 radar time series. Unlike previous approaches that focus on local algorithm optimisation or SAR feature selection, this work integrates two scales: (1) landscape patterns derived from annual distributions of time series metrics (TSMs) and (2) field-level phenology, both linked to growing degree days (GDD). TSMs were generated through breakpoint analyses over different smoothing intensities for Sentinel-1 polarisation (PolSAR) and interferometric coherence (InSAR) features, capturing crop, orbit and sensor-specific responses. The framework quantifies uncertainties inherent in both remote sensing and ground observations, and evaluates trackable progress (phenological stage detectability) and tracking range (GDD variance around stages) to assess accuracy under variable acquisition geometries, weather and smoothing parameters. Applied to the DEMMIN site (Germany), the analysis revealed consistent TSM-GDD relationships for wheat, rape, and sugar beet, with descriptors such as soil fertility and water availability explaining spatial patterns (R2 ≈ 0.8). Key novelties include the identification of low tracking ranges in drought years, the demonstration of the impact of orbit-specific incidence angles on monitoring fidelity, and the highlighting of Sentinel-1’s ability to resolve phenological variance across fragmented landscapes. By harmonising multi-scale SAR time series with agro-meteorological data, this approach advances transferable methods for operational crop monitoring, supporting precision agriculture and regional yield assessment beyond localised models.

The measurement of available water for agricultural plants is a crucial parameter for farmers, particularly to plan irrigation. However, an area-wide measurement is often not trivial as there are several inputs and outputs of water into the system. Here, we present a high-resolution, remote sensing-based water balance model for starch potato cultivation, combining multispectral ground station data with UAV and satellite imagery. Over a three-year period (2021–2023), data from Arable Mark 2 ground stations, DJI Phantom 4 MS drones, PlanetScope satellites, and Sentinel-2 satellites were collected in Mecklenburg–Western Pomerania, Germany. The model utilizes NDVI-based crop coefficients (R2 = 0.999) to estimate evapotranspiration and integrates on-farm irrigation and precipitation data for precise water balance calculations. A correlation with reference NDVI observations by Arable Mark 2 systems can be shown for UAV (R2 = 0.94), PlanetScope satellite data (R2 = 0.94), and Sentinel-2 satellite data (R2 = 0.93). We demonstrate the model’s ability to capture intra-site heterogeneity on a precision farming scale. Our spatially comprehensive model enables farmers to optimize irrigation strategies, reducing water and energy use. Although the results are based on sprinkler irrigation, the model remains adaptable for advanced irrigation methods such as drip and subsurface systems.

Maintaining an equilibrium between the rapid pace of urbanisation and the demand for urban ecological well-being amid climate change remains a global challenge. This study integrates machine learning and geospatial techniques with biophysical models to investigate the changes in ecosystem regulating services (ERS), such as carbon stock and climate regulation, in cities of the Rainforest (Akure and Owerri) and Guinea savanna (Makurdi and Minna) ecological regions of Nigeria in 2002 and 2022. Landsat images were processed using the random forest (RF) machine learning classifier, with the Normalised Difference Vegetation Index (NDVI) and Land Surface Temperature (LST) serving as indicators of landscape changes. The Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) platform was deployed to assess carbon storage and sequestration, and cooling and heat mitigation (HMI) services. Urban and agricultural expansion was associated with a drastic depletion of ERS within a 5 km-10 km radius of the urban core, resulting in an 8.60 %-33.83 % decline in carbon stock and a 5 %-13 % decline in HMI across cities. Correlation and geographically weighted regression models revealed that in the Rainforest (Akure and Owerri), carbon sequestration and heat mitigation are more influenced by LST, with strong correlations in Akure (r = 0.499) and Owerri (r = 0.408). In the Guinea savanna, carbon sequestration pattern in Makurdi is influenced by LST (r = 0.419), while Minna shows a stronger influence of NDVI on both carbon stock and heat mitigation. This highlights the influence of urbanisation and ecological variations in providing urban ERS and underscores the importance of enhancing vegetation biomass through existing urban and rural afforestation frameworks and sustainable agricultural practices. These measures are crucial for improving carbon stock,

West Africa’s vulnerability to climate change is influenced by a complex interplay of socio-economic and environmental factors, exacerbated by the region’s reliance on rain-fed agriculture. Climate variability, combined with rapid population growth, intensifies existing socio-economic challenges. Migration has become a key adaptive response to these challenges, enabling communities to diversify livelihoods and enhance resilience. However, spatial patterns of migration in response to climate risks are not fully understood. Thus, the study evaluates the applicability of the IPCC risk assessment framework to map and predict migration patterns in Ghana and Nigeria, with a focus on identifying areas of potential out-migration. By integrating geospatial environmental, socio-economic, and population data, the study highlights areas that have a higher likelihood of migration for the current baseline and near future (2050). Future climate is modeled using CMIP6 projections under the RCP4.5 scenario, while population projections providing insight into future exposure. The results from the baseline assessment are compared with actual migrant motivations, providing a ground-level perspective on migration drivers. In northern Ghana and Nigeria, elevated hazard, vulnerability, and exposure scores suggest a higher likelihood of migration due to the overall risk faced by the population. This pattern is projected to persist in the future. However, migrant responses indicate that environmental factors often play a secondary role, with vulnerability factors cited more frequently as migration drivers. The findings highlight the importance of developing localized adaptation strategies that address the specific needs of vulnerable areas. Additionally, management strategies that enhance community resilience and support sustainable migration pathways will be critical in addressing future climate-induced migration challenges.

Random Forest (RF) is a widely used machine learning algorithm for crop type mapping. RF’s variable importance aids in dimension reduction and identifying relevant multisource hyperspectral data. In this study, we examined spatial effects in a sequential backward feature elimination setting using RF variable importance in the example of a large-scale irrigation system in Punjab, Pakistan. We generated a reference classification with RF applied to 122 SAR and optical features from time series data of Sentinel‑1 and Sentinel‑2, respectively. We ranked features based on variable importance and iteratively repeated the classification by excluding the least important feature, assessing its agreement with the reference classification. McNemar’s test identified the critical point where feature reduction significantly affected the RF model’s predictions. Additionally, spatial assessment metrics were monitored at the pixel level, including spatial confidence (number of classifications agreeing with the reference map) and spatial instability (number of classes occurring during feature reduction). This process was repeated 10 times with ten distinct stratified random sampling splits, which showed similar variable rankings and critical points. In particular, VH SAR data was selected when cloud-free optical observations were unavailable. Omitting 80

This paper presents a Spatial Decision Support System (SDSS) designed to assist stakeholders in West Africa in analysing critical climate and land use indicators for risk management in agriculture and further sectors being affected by extreme precipitation and temperature events. Developed as part of the WASCAL WRAP 2.0 project LANDSURF, the SDSS makes scientific data accessible and comprehensible to non-scientific audiences, facilitating informed decision-making among communities affected by climate change. From the beginning of the development process, the web portal was co-designed with relevant West African stakeholders. Due to the challenging conditions during the COVID-19 pandemic, alternative online communication tools, e.g. ZOOM, online surveys and email, successfully were utilized to interact with stakeholders instead of on-site activities. The co-design process carried out with stakeholders includes several steps such as stakeholder analysis, identification of their information needs using specific climate, crop and remote sensing indicators, and the evaluation of the SDSS in a dedicated workshop. In total, the co-design process involved nine different steps, recorded and described in a stakeholder interaction protocol.The SDSS integrates observational data, including CHIRPS and ERA5-Land datasets, and state-of-the-art high-resolution climate model outputs under two greenhouse gas concentration scenarios (RCP2.6 and RCP8.5) and remote sensing data. It enables the comparison of model outputs with observations and facilitates the assessment of regional climate variability and trends. Two concept studies illustrate the SDSS’s functionality: one focusing on a farmer in Burkina Faso assessing irrigation needs for millet cultivation, and another involving a regional planner analysing drought and heat wave impacts in coastal West Africa. These examples highlight the SDSS’s usability in supporting adaptive strategies and enhancing resilience to climate-related challenges, underscoring the importance of integrating local knowledge with scientific data for effective climate adaptation and mitigation.