Spatial heterogeneity and partitioning of soil health indicators in the Northern Great Plains using self-organizing map and change point methods

Boluwade, Alaba

DOI: https://doi.org/10.1007/s12145-023-01007-6

2023-04-28

Earth Science Informatics

Abstract:The Nelson-Churchill River Basin (NCRB) is one of the largest river basins in North America, covering four provinces and four states in Canada and the United States, respectively. This study's primary objective is to evaluate the spatial variability and derive the optimal number of spatially contiguous regions of the NCRB using important soil health indicators (SHI) such as organic carbon, bulk density, cation exchange capacity, elevation, pH, and percentage of clay, gravel, silt, and sand extracted from the Unified North American Soil Map database. Several soil parameters are input into various watershed activities such as precision farming and water balance determination through hydrologic modeling. Therefore, information about their heterogeneities (homogeneities) is important from a best management practices (BMPs) perspective. This study applied a self-organizing map (SOM), an unsupervised artificial neural network technology that can visualize high-dimension datasets in a 2-D format and preserve the topology of multivariate datasets. The derived result (SOM nodes) from the SOM procedure was then clustered using the hierarchical clustering method (HCM). Using SOM and HCM, all the variables mentioned above were partitioned into a possible number of clusters that did not follow the geographical boundaries of seven sub-basins inside the NCRB. In addition, change point methods based on Markov Chain Monte Carlo and nonparametric E-Divisive change point methods were used to partition the within-cluster sum of squares into an optimal number of clusters. The study reveals two partitions (two and five regions) as optimal, meaning there would be no further striking changes to the number of clusters after these partitions. Therefore, both the two-region and five-region partitions would help in the decision process, especially when there are project initiatives or interventions at a scale larger than a watershed or sub-basin. A proper understanding of the spatial heterogeneity and the optimal number of clusters in a large regional river basin such as the NCRB could aid BMPs, climate change adaptation efforts, precision agriculture, and water resources management through hydrologic modeling.

geosciences, multidisciplinary,computer science, interdisciplinary applications