Spatial scale applicability of 4-Scale geometrical optics model canopy reflectance simulation.

Abstract

Accurately clarifying the applicable spatial scale of 4-Scale model is conducive to improving the accuracy of its application in canopy reflectance simulation of different vegetation types, and to further improving the inversion accuracy of leaf area index, canopy density, and other parameters. Two forest plots (one for broad-leaved forest and one for mixed forest) with each area of 100 m×100 m in Maoershan Experimental Forest Farm, Shangzhi, Heilongjiang, were divided into the spatial scales of 10, 20, 30, 40 and 50 m, respectively. The 4-Scale model was used to simulate forest canopy reflectance. Local mean method, the nearest neighbor method, bilinear interpolation method, and cubic convolution method were used to convert Sentinel-2 images with spatial resolution of 10 m to other scales, with the results being evaluated. The simulated canopy reflectance and remote sensing pixel reflectance were compared and analyzed. The spatial scale of mixed forest and broad-leaved forest suitable for high-precision inversion parameters of 4-Scale model was determined. The results showed that the 4-Scale model underestimated the pixel forest canopy reflectance as a whole. The canopy reflectance of mixed forest and broad-leaved forest had the worst simulation effect at the 20 m scale. Both the root mean square error (RMSE) and the mean absolute error from (MAE) of red and near-infrared band were large. When the scale was >20 m, the simulation effect became better. The applicability of the model was the best when the mixed forest was 40 m and the broad-leaved forest was 30 m. The mean and standard deviation of the reflectance difference between the simulated value and the remote sensing pixel were the minimum in the red and near near-infrared bands, with the minimum RMSE and MAE. The simulation results of mixed forest and broad-leaved forest at 10 m scale were not stable, the rule of mean and standard deviation was inconsistent, and the difference between RMSE and MAE was large under the same band.