Fang, Hongliang; Zhang, Yinghui; Li, Sijia; Wang, Yao: Field measurements and high resolution reference data of leaf area index, fraction of vegetation cover and clumping index in Honghe, China, 2019 [dataset bundled publication]. PANGAEA, https://doi.pangaea.de/10.1594/PANGAEA.967411 (dataset in review)

The file includes both field measured and satellite derived high resolution leaf area index(LAI), fraction of vegetation cover(FVC) and clumpling index(CI) data obtained over the Honghe farm in northeastern China. The Honghe farm (centered at 47°39′N, 133°31′E) is located in the east of the Heilongjiang province, northeast China. Five plots in 400 m × 600 m were selected in the Honghe farm in 2019. Within each plot, about 50 - 60 elementary sampling units (ESUs) about 20 m ×20 m in size were selected in different weeks with a moving sampling strategy to avoid the sampling disturbance. Field LAI/FVC/CI measurements were performed weekly June 22 to August 26, 2019. All ESU measurements made with Digital hemispherical photography (DHP) within a plot were averaged to represent the plot LAI/FVC/CI. High-resolution reference data was generated with cloud-free harmonized Landsat, and Sentinel-2 (HLS) imagery based on random forest models for 2019, respectively. The HLS V1.4 dataset with a spatial resolution of 30 m (L30 and S30) was used for upscaling (https: //hls.gsfc.nasa.gov). In the random forest model, the reflectances of the blue, green, red, and near-infrared bands were used as explanatory variables, whereas LAIe, LAI, and FVC were used as the explained variables. The model was trained with yearly ground measurements and assessed by 10-fold cross-validation, for which nine-tenths of the samples were used for training and the remainder of the samples were used to evaluate the model. This step was run 10 times until all samples were looped. The optimal hyperparameters of the model (e.g. the number of trees and the maximum depth of the tree) were determined by selecting parameters corresponding to the best performance after iterative running.