Quantification and driving mechanism of cultivated land fragmentation under scale differences

Abstract

Cultivated land fragmentation (CLF) has received widespread attention, which often leads to increased planting costs, the excessive use of agricultural chemicals, and hinders agricultural mechanization and modernization. This study investigated the severity of CLF in the Yangtze River Delta region (YRD), one of China's economic growth poles, where land use conflict is prominent. First, the cultivated land fragmentation index (CLFI) was calculated and visualized on a grid scale of 15 km and an administrative scale of counties. Second, the variance inflation factor was analyzed to verify collinearity of six potential drivers, which were gross domestic product density (GD), population density (PD), road density (RD), Slope, Elevation and Precipitation. Finally, multiple geographically weighted regressions (MGWR) were applied to explore driving mechanisms of CLF on two different scales, and their differences have been fully demonstrated. The results showed that the average value of CLFI on the grid scale was 0.32, which was lower than 0.36 on the county scale. There were also significant differences in the spatial distribution of CLFI on the two scales. Among the driving factors under two scales, PD had the largest variance inflation factor (VIF), with a value of 8.142 (<10). All driving factors had the required VIF values of MGWR and passed the significance test (pvalue<0.05). The bandwidths of PD, Slope and Precipitation on the grid scale are 47, 47, 163, respectively, and bandwidths of Elevation and Precipitation on the county scale are 44 and 43, respectively. GD and RD on both scales had a weak positive effect on CLF. On the grid scale, the maximum value of the regression coefficient of PD was 1.67, and the minimum value was −2.33, which had the most drastic effect. PD had an obvious negative effect on CLF on the county scale, and its regression coefficient distribution was around −0.3. Elevation and Precipitation had the widest range of regression coefficients (2.86 and 2.37, respectively), indicating that they had the most dramatic impact on the county scale. The regression coefficients of RD, Slope and Elevation had little difference in spatial distribution on different scales, while the differences of GD, PD and Precipitation were obvious. These findings can enrich research on CLF, highlighting differences in scale and providing valuable insights into the formulation of cultivated land governance policies at different scales.