An indicator of relative distribution probability of field-scale permafrost in Northeast China: Using a particle swarm optimization (PSO)-based indicator composition algorithm

Under the influence of climate changing, permafrost in Northeast China (NEC) has been consistently degrading in recent years. Numerous scholars have investigated the spatial and temporal distribution patterns of permafrost in the NEC region. However, due to constraints in data availability and methodological approaches, only a limited number of studies have extended their analyses to the field scale. In this study, we established a particle swarm optimization (PSO)-based indicator composition algorithm (PSO-ICA) to obtain an indicator factor, η, that indicates the relative distribution probability of permafrost at the field scale. PSO-ICA screened and combined 12 high-resolution environmental variables to compose η. The spatial distribution data of permafrost with a length of 765.378 km provided by the engineering geological investigation report (EGIR) of six highways were used to train and validate the effectiveness of η in indicating permafrost. At the field scale, η was found to be similar to the surface freezing number (SFN) in its ability to indicate permafrost, with AUC values of 0.7046 and 0.7063 for the two by the ROC test. In addition, η has a good performance in predicting highway distresses in the permafrost region in the absence of survey data. This study also confirmed that the resolution and accuracy of permafrost mapping results can be improved by utilizing η. After downscaling the 1 km resolution SFN to 30 m resolution using η, the R² of the linear relationship between SFN and permafrost temperatures from 43 monitoring boreholes was improved from 0.7010 to 0.8043. If η can help understand the distribution of permafrost at field scale, many engineering and environmental practices could potentially benefit.