Can CATPCA be utilized for spatial modeling? a case of the generation susceptibility of gully head in a watershed.

[Display omitted] • CATPCA was first used in the spatial modeling with gully head as a case. • CATPCA can evaluate the importance of each classification and factors. • CATPCA can identify the correlation between categorical and numerical factors. • Elevation, human activities, lithology and soil type were the key factors. • CATPCA has great potential in the spatial modeling. Many spatial modeling methods have emerged; however, they require dependent variables, cannot reflect the relationship between categorical variables and numerical variables, and are limited by the interference of data collinearity. Categorical principal components analysis (CATPCA) has the potential to overcome these issues. Therefore, in order to investigate the suitability of CATPCA for spatial modeling, we conducted a case study based on the generation susceptibility of gully heads was determined, including 2310 gully head and 23 variables. CATPCA was first used in the spatial modeling of gully head generation. The first six principal components retained 76.4% of data trends. The area of the training and validation sensitivity curve s were 75.4% and 75.7%, respectively, which reflected good levels. CATPCA can simulate gully head spatial differences, and thus, has great potential for spatial modeling applications. Among the 23 factors, elevation, distance to residential areas, human footprint, lithology, and soil type were identified as the main controlling factors affecting the generation susceptibility of gully heads, with high correlations between them. Low altitudes, close proximity to residential areas, high human footprint, and poor vegetation were associated with high susceptibility. The findings of this study provide a better understanding of the applicability of CATPCA for spatial modeling. CATPCA is a novel solution for spatial modeling strategies that can improve understanding and has great potential for various spatial modeling applications in the future. [ABSTRACT FROM AUTHOR]