Impact of Spatial Resolution of Environmental Covariates on the Accuracy of Digital Soil Mapping: A Review Based on the SCORPAN Conceptual Framework

Background and Objective

Digital Soil Mapping (DSM), as a modern data-driven approach, predicts the spatial distribution of soil physical and chemical properties based on the SCORPAN model. However, one of the keys and often overlooked factors influencing mapping accuracy is the spatial resolution of environmental covariates, which may either enhance or distort the true soil-forming patterns. This review article aims to examine and analyze the impact of the spatial resolution of environmental variables on the accuracy of DSM, particularly in arid and semi-arid regions of Iran.

The specific objectives include: (i) identifying scalability challenges such as scale mismatch, noise amplification at very fine resolutions, and computational costs; (ii) providing optimized resolution recommendations based on landscape type; and (iii) proposing multi-scale approaches and advancements in machine learning to improve local accuracy and global generalizability.

This review emphasizes the importance of adaptive spatial resolution selection for practical applications such as sustainable agriculture and evidence-based environmental policymaking under climate change. The focus on arid and semi-arid regions stems from the high sensitivity of these ecosystems to micro-scale variations, where inappropriate resolution may increase prediction errors by 30-50%. Ultimately, this study seeks to bridge theory and practice to enhance DSM as a more operational and effective tool.

Materials and Methods

This targeted review was conducted in accordance with the PRISMA 2020 statement. A comprehensive search was performed across major international and Persian databases, covering the period from 2000 to 2025. Search terms consisted of combinations of key DSM-related terminology.

A total of 438 articles were initially identified. After removing duplicates, 302 articles remained for preliminary screening. Inclusion criteria comprised studies that directly or indirectly examined the effect of spatial resolution on DSM accuracy and evaluated at least one SCORPAN factor. Following full-text assessment, 150 articles were reviewed in detail, and ultimately 56 studies were included in the final analysis.

Data extraction involved categorizing variables according to the SCORPAN framework, evaluating methodological approaches, validation metrics, strengths and limitations, and identifying emerging trends.

Results

The findings indicate that, in complex arid and semi-arid terrains, the spatial resolution of topographic variables should be as fine as 30 m to adequately capture local features such as rills and erosion patterns and to prevent excessive smoothing. Otherwise, the prediction accuracy of properties such as clay content or soil water storage may decline by 30-40%.

For climatic variables, a spatial resolution finer than 250 m is essential in these regions to better model microclimates and their interactions with topography, thereby reducing unexplained variance.

Biological and remote sensing covariates require a spatial resolution of 10-30 m to capture seasonal and patchy vegetation dynamics in dry ecosystems, potentially improving prediction accuracy by up to 25%.

Parent material and geological variables are generally adequate at 90-100 m resolution; however, in highly heterogeneous settings, integration with topographic data is necessary to improve the prediction of soil chemical properties.

Soil age and spatial position variables play complementary roles, and their integration at moderate resolutions may reduce uncertainty by 10-20%.

Recent advancements in machine learning algorithms and multi-scale modeling approaches have improved prediction accuracy across multiple spatial scales while addressing challenges such as scale mismatch. The recommendation framework suggests that in humid and temperate lowland regions, moderate spatial resolution is generally sufficient, whereas in arid and rugged landscapes, high spatial resolution for topography and vegetation is essential.

Discussion

The analysis underscores that spatial resolution selection should be adaptive and dependent on landscape complexity, modeling objectives, and practical constraints to balance local accuracy, computational efficiency, and generalizability.

In arid and semi-arid regions, high spatial resolution more effectively captures micro-scale patterns of erosion, salinization, and moisture distribution. However, it also introduces challenges such as increased noise, overfitting, and large data processing costs, which require careful methodological management.

Scale mismatch among covariates increases unexplained variance and highlights the need for spatial harmonization. Advances in deep learning and three-dimensional modeling are transforming DSM from a static to a dynamic framework, improving predictive performance in environmentally sensitive ecosystems. Nevertheless, critical gaps remain, particularly the scarcity of historical soil age data in specific biomes.

Ultimately, this study demonstrates that spatial resolution is not merely a technical parameter, but a key determinant of uncertainty reduction, enabling digital soil mapping to evolve into a more effective tool for environmental policymaking and sustainable agriculture.