Retrieval of the Leaf Area Index from Visible Infrared Imaging Radiometer Suite (VIIRS) Surface Reflectance Based on Unsupervised Domain Adaptation

Abstract

Several global leaf area index (LAI) products were generated using neural networks, but the training dataset for the neural networks was sensor specific, and the construction of the training dataset was time consuming. In this paper, an unsupervised domain adaptation-based method was proposed to estimate LAI from the Visible Infrared Imaging Radiometer Suite (VIIRS) surface reflectance dataset based on a training dataset constructed from the Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance dataset. A transfer component analysis (TCA) algorithm was first utilized to map the MODIS and VIIRS surface reflectance into the same subspace to reduce the distribution discrepancies between the MODIS and VIIRS surface reflectance. Then, the embedded data obtained from MODIS surface reflectance dataset, along with the LAI values produced by fusing the MODIS and the Carbon cYcle and Change in Land Observational Products from an Ensemble of Satellites (CYCLOPES) products, were employed to train general regression neural networks (GRNNs). Finally, for retrieving the LAI values, the embedded data acquired from the VIIRS surface reflectance dataset was input into the trained GRNNs. For multiple field sites with different biome types, we used this developed method to retrieve LAI values based on the VIIRS surface reflectance dataset. The results indicate that, based on the training dataset built from MODIS surface reflectance dataset, the domain adaptation-based retrieval method can effectively estimate LAI values from VIIRS surface reflectance dataset. By comparison with the VIIRS and MODIS LAI products, the retrieved LAI values with TCA are more consistent with the reference LAI values acquired from high-resolution remote sensing images. The coefficient of determination (R2) and root mean square error (RMSE) of the retrieved LAI values with TCA at all selected sites are 0.88 and 0.68, respectively. Furthermore, the accuracy of the retrieved LAI values with TCA is higher than the retrieved LAI values without TCA with the R2 0.81 and the RMSE 0.79.

A High Spatiotemporal Enhancement Method of Forest Vegetation Leaf Area Index Based on Landsat8 OLI and GF-1 WFV Data

“…Neural networks have great advantages when used to deal with linear and complex nonlinear problems [9,10]. Many scholars have introduced artificial neural networks (ANN) into the model research of LAI inversion [11,12], among which the backpropagation neural network (BPNN) has a multi-layer nonlinear network structure and strong robustness of the model, and is widely used [13,14]. Yang Min et al used Landsat8 OLI and measured data to simulate the BPNN model, and its inversion accuracy is higher than that of the traditional regression model [15].…”