Abstract: 【Aim】 Over the years, the typical larch pest of the Mongolian Plateau, Erannis jacobsoni, has occurred frequently, and caused severe damage to forest ecosystem. Population density can directly describe the severity of forest pests, and it is extremely important to obtain information on pest population density in a timely and rapid manner. The purpose of this study is to construct a population density estimation method based on the hyperspectral data of larch trees and the population density data of E. jacobsoni in the outbreak areas of E. jacobsoni. 【Methods】 Outbreak areas of E. jacobsoni in four localities in Hangay Province and Khentiy Province in Mongolia were selected as the test areas. Firstly, from the four test areas 110 sample trees of Siberian larch (Larix sibirica) with different degrees of damage were selected, and the population density and the spectral reflectance of the canopy was measured. Secondly, the differential spectral reflectance (DSR) and the modified spectral index (MSI) were calculated through spectral reflectance data. Thirdly, the polynomial curve fitting method was used to analyze the sensitivity of DSR and MSI to population density. Then, the successive projection algorithm (SPA) was used to extract sensitive DSR and MSI. Finally, the sensitive DSR and MSI, polynomial regression (PR) and support vector machine regression (SVMR) algorithms were used to establish the population density estimation model of E. jacobsoni, and the accuracy was evaluated. 【Results】 The sensitive bands of DSR are mainly in the yellow edge and red edge, and the sensitivity at 572 nm was the most significant (R²=0.5821, P<0.001). The most sensitive index of MSI was TVI (R²=0.5386, P<0.001). The TVI (R²_CV=0.6323, RMSE_CV=0.1513) was more accurate than DSR₅₇₂ (R²_CV=0.5581, RMSE_CV=0.1649), while multiple DSRs (R²_CV=0.7309, RMSE_CV=0.1347) had higher estimation potential than multiple MSIs (R²_CV=0.6537, RMSE_CV=0.1453), and the performance of its SVMR model was always better than the PR model, suggesting that SVMR is suitable for population density estimation. 【Conclusion】 MSI and DSR can be used as sensitive indicators for population density estimation, and the polynomial curve fitting method can tap the sensitivity of MSI and DSR to population density. SPA is an effective method for POPD (pest population density) sensitive spectral feature extraction. The extracted DSR sensitivity index and MSI sensitivity index have fully captured the chlorophyll absorption characteristics and water absorption characteristics of larch needles and the reflection characteristics caused by the damage of needle cells. This study not only provides an experimental theoretical basis for the aerial and aerospace remote sensing monitoring of forest pest population density, but also develops a new way for remote sensing and monitoring of forest pests.

Key words: Erannis jacobsoni, larch, population density, modified spectral index, differential spectral reflectance, estimation model