hyperspectral classification of plateau zokor molehills at different successional stages in alpine meadows

Abstract The distribution of molehills of the plateau zokor (Eospalax fontanierii) in alpine meadows and the age of their formation influence the structure and function of alpine meadow ecosystems. To realize a remote sensing survey of large grassland molehills, in this experiment, the Plateau zokor molehills in alpine meadows with heavy rodent infestation in Cooperative City, Gannan Tibetan Autonomous Prefecture, Gansu Province, classified according to successional stage based on soil characteristics and vegetation conditions, and categorized them into one-year-old molehills (ZM0), two-year-old molehills (ZM1), three-year-old molehills (ZM2), and perennial molehills (ZMM). The normal vegetation of the meadows adjacent to the molehills was used as a control (CK) to study the distribution status of the mouse and the control. The control (CK) was used to study the total cover, height and biomass of the vegetation of the molehills and the control group. The spectral data of the Zokor molehills and the control group at different stages of succession were collected using the ASD spectrometer to analyze the total cover, height and biomass of the vegetation of the molehills as well as the vegetation of Patrick, Shannon-Wiener, Pielou and other species. The correlation between total cover, height, biomass, Patrick, Shannon-Wiener and Pielou vegetation traits and spectral vegetation indices was analyzed, and multiple stepwise regression and random forest (RF) inversion models of vegetation traits were constructed. The results of the experiment showed that: 1) with the increase of the successional period of molehills, their Pielou, Patrick, height, biomass and the total cover showed an increasing trend, and except for Pielou, the differences with the control group were significant (P<0.05); 2) the original spectral characteristics of the surface of molehills were different in different successional periods, and the predictive ability of the multivariate stepwise regression model for the total cover was the best; the predictive ability of the random forest inversion model was the best; and the random forest modeling was based on the relationship between vegetation characteristics and spectral data of zocores in different successional periods. Total cover; Random Forest could Patrick (R2=0.537, RMSE=0.823), height (R2=0.459, RMSE=0.143), total cover (R2=0.635, RMSE=24.287), biomass (R2 = 0.948, RMSE = 15.173); TCARI (Enhanced Chlorophyll Absorption Vegetation Index) plays an important role in the inversion of Patrick, height and total cover, and SR (the ratio of NIR/R is known as Simple Ratio) plays an important role in the inversion of biomass.3) The use of vegetation spectral indices to classify molehills can distinguish molehills in different successional stages, and their out-of-bag data OOB is 23.91%, SR and TCARI are important spectral indices for classification modeling of molehills. The above results provide a theoretical basis for remote sensing methods for the classification of plateau zokor moles, and can be used to analyze the succession of zokor moles through the random forest model using the vegetation indices SR and TCARI, which can help us analyze the succession of plateau zokor more quickly and thus maintain the balance of grassland ecosystems.