The objective of this Forum article is to argue that one of the key justification for developing theories and models of soil processes is that these theories are needed to determine what it is relevant to measure in soils. Without these theories and models to guide us, in particular if we rely on machine-learning and artificial intelligence methods to make progress, we are navigating in the dark, and are likely to base decisions on more correlations that do not reflect true mechanisms.

Soil erosion by water threatens food security, but soil conservation practices can help protect arable land. We tested a soil erosion model that simulates sediment yields in micro-scale watersheds with soil conservation in place. The model captured the very low sediment yields but showed limited accuracy on an annual time scale. However, it performed well when applied to larger areas over longer timeframes, demonstrating its suitability for strategic long-term soil conservation planning.

When glaciers retreat, new land surface is revealed. Using detailed glacial retreat maps, it is possible to determine for how long a location has been ice-free. This age is used in this study to analyse how fast carbon is incorporated into the soil. Our results show that the wetness of the soil strongly determines the CO₂ uptake and carbon incorporation rates. Wetlands cover a small percentage of the land surface but are nonetheless important for the carbon storage in the deglaciated area.

We studied how soil-building practices like cover cropping, reduced tillage, and crop diversity affect soil carbon on farms in California. We find that maintaining living plants throughout the year and reducing tillage can boost multiple forms of soil carbon with distinct functions and longevity. While soil type matters, how the land is managed often had a greater effect. This suggests that improving farming practices can meaningfully increase carbon in the soil and draw down atmospheric carbon.

We investigated how the soil degradation status of smallholder grazing, classified using remote sensing, in two districts of Western Kenya, compared with measured soil parameters at 90 sites. Grouping sites using soil data had some agreement with the remote sensing (RS) classification. Overall, our results suggest that supplementing RS methods with microbial biomass C, soil P, percent C and N, and soil pH, could enhance our ability to identify degraded soils and target restoration efforts.

Olive groves face severe soil erosion, threatening their sustainability. This study reviewed existing research to understand how and why erosion happens. By analysing data from various measurement methods, we found that erosion rates are unsustainable and that while natural factors contribute, farming practices, especially keeping soil bare, drive significant soil loss. Adding ground cover can cut soil loss by over half. This urges the adoption of soil conservation practices.

We compared soil properties and water storage capacity across these three forest types, finding that conversion from primary forests significantly reduced water retention capacity. While secondary forests exhibited better water retention than rubber plantations in shallow soil layers, they showed weaker capacity in deeper soils, with distinct seasonal variations in water storage. Soil porosity was identified as the key determining factor, with primary forests maintaining higher porosity.

The study evaluates the effect of trachyte and basalt powders on soil and maize yield in the Sudano-Sahelian zone of Cameroon. Texture, base saturation, and organic carbon indicate moderate limitations. The climate is favourable for maize cultivation. Rock powder application significantly improves yield, with basalt being more effective than trachyte. The advantage of using rock powders lies in their environmental benefits, long-term effectiveness, and more affordable cost.

This study compares the performance of different neural network architectures and a random forest model for modelling soil erosion by water at a small spatial scale in agricultural cropland. The results indicate that a complex convolutional neural network outperforms all other tested models across standard evaluation metrics. A permutation importance analysis identified the digital elevation model as the most influential predictor across all models.

The soil stoichiometric characteristics in the study area are predominantly governed by niche-microtopographic factors. Although plant life forms influence certain elements through processes such as nutrient return, their overall effect is relatively weak, and their impact is often contingent upon the microtopographic context. This combined action of biotic and abiotic factors reflects the complexity inherent in the karst soil nutrient cycling mechanism.

This study examines organic carbon (OC)–mineral interactions in permafrost soils undergoing thermokarst degradation in Cape Bounty (Melville Island, Canada). Chemically stabilized OC accounts for 13 ± 5 % as organo-metallic complexes and 6 ± 2 % as associations with iron oxides. Including physical protection, up to 64 ± 10 % of OC is mineral-protected. Deeper layers show a sharp decline in mineral-bound OC, suggesting increased vulnerability to degradation when exposed by deep thaw features.