Journal Paper Digests 2026 #14
- Integrating soil health and manure practices to advance sustainability for US dairy farms
- A function-semantic oriented heterogeneous graph aggregation framework with weighted spatial relationships for urban functional zone mapping
- Linking Soil Threats and Mitigating Management Practices
- Mid-Infrared Spectroscopy to Estimate Ratios of Soil Organic Carbon to Clay
- Challenges in reliable bias correction of hydro-climatic and energy data: Towards a reconstruction of datasets
Challenges in reliable bias correction of hydro-climatic and energy data: Towards a reconstruction of datasets
Regular gridded datasets commonly exhibit significant biases and therefore require correction before being used in impact studies. This study highlights the challenges involved in improving the adjustment of climate variables (precipitation and temperature), hydrological variables (soil water content, streamflow, evapotranspiration), and radiative variables (longwave and shortwave radiation). Eighteen bias-correction (BC) methods were selected, covering a wide range of approaches, including classical statistical techniques (e.g., quantile mapping), distribution-based methods (e.g., kernel smoothing), and machine learning approaches (e.g., random forests). BC outputs from CMIP6 models and ERA5 reanalyses were evaluated against observations from contrasting climatic regions of Benin. The evaluation relied on Kling–Gupta Efficiency (KGE), Pearson correlation coefficient, relative-mean-absolute-error (RMAE), and fraction within tolerance (FTI), computed at hourly, daily, and monthly temporal scales. It also considered the representation of diurnal and monthly cycles, as well as consistency with mean values and trends of various climatic extreme indices. Three calibration strategies were explored: (i) using the first half of the time series for calibration to assess the preservation of future trends; (ii) the second half to evaluate the ability to reconstruct missing data; and (iii) 80% of the series to investigate the effect of sample size. Finally, bias corrections were applied by stratifying the data by month and intensity ranges to assess the limitations of global BC methods. Results show that machine learning methods do not systematically outperform classical bias-correction approaches. Performance varies considerably by variable, station, and climatic context. It is therefore important to select methods suited to the specific conditions and intended use, to improve the reliability of future climate and hydrological projections.
Mid-Infrared Spectroscopy to Estimate Ratios of Soil Organic Carbon to Clay
The index based on the ratio of soil organic carbon to clay concentration (SOC/clay) has been confirmed as an effective tool for assessing the organic matter status of mineral soils. The index contains four classes: very good (SOC/clay ≥ 1/8), good (1/10–1/8), moderate (1/13–1/10) and degraded (≤ 1/13). Conventional analyses of SOC and clay concentrations are resource intensive, however. Here, we investigate the use of mid-infrared spectroscopy (MIRS) as an alternative for quantifying SOC/clay ratios and the index. We use data from the National Soil Inventory of England and Wales (n = 1301). We obtained RMSE values of 3.6 and 57 g kg−1 for estimations of SOC and clay concentrations by MIRS, respectively. We found that a simplified three-class index was satisfactory for predictions using MIRS. The accuracy of clay concentration estimations was a limiting factor, particularly for the combined middle class (approx. 60%–70% success). Over 80% of the samples in the other two classes were correctly predicted. The probability that a sample was in the estimated index class can be used to filter out estimates likely to be misclassified. The filtering removed at least twice the number of misclassified samples as the number of correctly classified that had below-threshold probability. We found that uncertainty estimates are more informative if calculated from the component estimates of SOC and clay rather than the SOC/clay ratio. An important source of uncertainty is the mismatch between MIRS estimates of clay concentration based on clay mineralogy and conventional estimates based on particle size fraction. The additional information on clay mineralogy given by MIRS means it provides a better characterisation of SOC–clay interactions affecting soil functioning.
Linking Soil Threats and Mitigating Management Practices
Climate change, population growth, and intensifying land use are exerting increasing pressure on soil resources worldwide. Despite growing knowledge of soil degradation processes, the relationship between soil threats and the effectiveness of management practices remains poorly synthesized. Here we present a comparative assessment of soil expert perspectives on soil threats and mitigation practices based on a survey of 162 soil experts from 38 countries. Across tropical, arid, and temperate climates, soil experts evaluated the importance of soil threats and the perceived effectiveness and use of key soil management practices, whereas also highlighting examples of locally adapted innovations. Across all responses, the most important soil threats were organic matter decline (overall rating of 4.2 on a five-point scale), soil erosion (4.1), and biodiversity loss (3.9), indicating broad agreement on the primary drivers of soil degradation. Mitigating management strategies (crop diversification, reduced tillage, organic inputs, and agroforestry) were consistently perceived as effective across climates (scores from 3.7 to 4.0); however, their implementation was not as widespread (2.6 to 3.2). Differences among climate groups were detectable for some soil threats, such as the higher perceived importance of organic matter decline in tropical climates and the greater relevance of salinization in arid climates. Soil experts also highlighted locally developed climate-smart farming systems, including Milpa intercropped with fruit trees in alternating strips (MIAF), Zero-budget natural farming, and agrivoltaic systems, with potential for upscaling and wider application across contexts. Overall, this global soil expert survey identifies common priorities in soil threats and management practices across three major climate groups, while illustrating how local environmental and socio-economic conditions shape the selection of sustainable soil management practices. These findings provide a comparative perspective to inform future research, monitoring, and soil management strategies.
A function-semantic oriented heterogeneous graph aggregation framework with weighted spatial relationships for urban functional zone mapping
Urban Functional Zones (UFZs) serve as spatial carriers for urban economic and social activities. Accurate and fine-grained UFZ mapping is critical for urban planning, governance, and sustainable development. The complementarity between Remote sensing images and Points of Interest (POIs) provides important support for UFZ mapping. However, existing UFZ identification methods exhibit notable limitations in integrating multimodal features from remote sensing images and POIs due to inherent data differences. These methods rely on pre-defined UFZ units and lack the capability to model spatial relationships between geographic objects, thereby cause functional mixing. Moreover, extracting multi-source heterogeneous features from images and POIs is essential for constructing multidimensional semantic representations of geographic objects. To address these challenges, this study proposes a Function-Semantic Oriented Geographic Object Heterogeneous Graph Aggregation (FOGA) framework for accurate UFZ mapping. FOGA models urban space as a heterogeneous graph composed of an image layer and a POI layer. MGFMamba is designed and Bidirectional Encoder Representations from Transformers (BERT) is employed to extract functional semantic features from images and POIs, respectively. By embedding both types of semantic features into the graph and constructing intra-layer and cross-layer connections, the framework achieves unified modeling of multimodal feature fusion spatial relationships. A Heterogeneous Edge-Attention Relational Graph Netual Network (HEA-RGNN) is proposed to perform UFZ mapping. Within this heterogeneous graph structure, adaptive multimodal feature propagate enables dynamic learning of feature representation ranges and cross-modal semantic weights, orienting the aggregation of geographic objects into UFZs without pre-defined UFZ units. Experimental results from four cities demonstrate the effectiveness and stability of FOGA. FOGA is further applied to construct the first Chinese UFZ dataset that requires no predefined spatial units, at 2.4 m resolution across 31 major cities in 2024, with results outperforming existing products in both mapping accuracy and fine-grained spatial representation. The UFZ data and source code produced in this study have been made public at the following link: https://github.com/haha123haha460/foga.
Integrating soil health and manure practices to advance sustainability for US dairy farms
The US dairy industry has committed to advancing environmental sustainability by reducing greenhouse gas (GHG) emissions, enhancing water use efficiency, and improving water quality. Feed production accounts for approximately 12% of GHG emissions and 99% of consumptive water use from dairy operations, making it a focus area for potential resource use and overall efficiency improvements. However, few studies report changes to GHG emissions or water quantity and quality outcomes from adopting soil health management systems and use of novel manure products for dairy feed production. The Dairy Soil and Water Regeneration (DSWR) project is exploring whether soil health management systems and novel manure products can help advance environmental sustainability outcomes across major dairy-producing regions in the United States. Through a suite of coordinated studies, including regional soil benchmarking and large-plot- to field-scale experiments, DSWR is evaluating the performance and scalability of reduced tillage, cover crops, and novel manure products applied to row crop feed production systems. By integrating high-resolution data on soil, water, GHG emissions, and crop production, this project is generating actionable insights to support decision-making for farmers, farm managers, dairy cooperatives, retailers, and consumer packaged goods companies. We introduce the project by summarizing its purpose, the conceptual framework guiding its design and implementation, and its approaches to hypothesis testing about soil health, hydrology, and yield responses.