Landscape fragmentation reduces but shape complexity enhances soil loss: Evidence from the plain grain-producing area along the Yangtze River in China

• The systematic influencing mechanism of landscape pattern on soil loss is investigated at the sub-watershed level in Jianghan Plain. • Positive spatial autocorrelation of soil loss is identified and the spatial regression model is utilized to explore the driving factors. • Landscape fragmentation in Jianghan Plain is revealed to attenuate soil loss with a decreasing effect. • It is found that high magnitude of shape complexity and landscape diversity lead to high level of soil loss in Jianghan Plain. Soil loss poses a significant threat to global food security, necessitating the investigation of its driving mechanisms to develop effective mitigation strategies. The spatial-temporal patterns of soil loss in the Jianghan Plain, one of the major grain-producing area in China, were calculated using the Revised Universal Soil Loss Equation (RUSLE) model for the period from 2005 to 2019. Then the influencing mechanism of landscape pattern on soil loss at the sub-watershed level was revealed by ordinary least square (OLS) and spatial error model (SEM). The results showed that: (1) A decrease in total soil loss from 2005 to 2015, associated with rapid farmland decline due to intensified anthropogenic activities, followed by a sharp increase from 2015 to 2019, largely driven by extensive farmland reclamation projects. (2) Contrary to conventional understanding, patch density (PD) was found to be negatively correlated with soil loss, attributed to human interventions such as agricultural facility construction and ecological ridges that disrupt runoff pathways and increase surface roughness. Meanwhile, perimeter area fractal dimension (PAFRAC) was positively correlated with soil loss and played a dominant role, due to the fact that complex patch shapes increased erosion-prone edge surfaces in plains with the dominance of farmland. Shannon homogeneity index (SHEI) was positively correlated with soil loss. In addition, the regression results showed that for each 1 % increase in PD, soil loss decreased by approximately 2 %, whereas 1 % increase in PFARAC and SHEI would lead to the increase of soil loss by 20 % and increased by 2 %, respectively. (3) Soil loss in Jianghan Plain exhibited spatial clustering, especially in the western areas where the terrain was predominantly hilly. By prioritizing the conversion of approximately 5,400 ha of cropland with slopes greater than 15° in these areas into forest, soil loss was expected to be reduced by about 2,606 t. Accordingly, the findings will contribute to providing a basis for formulating soil conservation policies integrated with integrated landscape management in the plain grain-producing areas.