This study provides quality control black sky albedo (BSA) and white sky albedo (WSA) corresponding to 13 AWS. ASP, ASN, ACO, DSL, EBO, HZS, HCG, JYL, DMS, BJT, SSW, ZYS, and HZZ in the Heihe River Basin from 2013 to 2014.
MCD43A3 (V006) is the most widely used surface albedo product. This product adopts a semi empirical linear kernel driven bidirectional reflectance distribution function model, which can obtain bidirectional reflectance under any solar irradiation and satellite observation conditions. Then, by integrating the bidirectional reflectance of the observed hemisphere, the black sky albedo (BSA) can be obtained; By integrating the bidirectional reflectance of the incident hemisphere and the outgoing hemisphere twice, the daytime albedo (WSA) can be obtained. The MCD43A3 used in this article has a spatial resolution of 500 meters and a temporal resolution of daily. It provides BSA and WSA in the local solar noon 1-7 band, as well as three wideband albedo in the visible light (300-700 nanometers), near-infrared (700-3000 nanometers), and shortwave (3000-5000 nanometers) bands. This article uses shortwave albedo because its spectral range is comparable to in-situ based albedo [1]. We extracted the quality control BSA and WSA corresponding to 13 AWS from 2013 to 2014. Due to the fact that in-situ measurements provide blue sky albedo, the BSA and WSA of MCD43A3 are linearly combined using the sky diffuse light ratio.
This study describes the use of ground residual measurements based on automatic weather stations (AWS) to validate MCD43A3 and albedo tolerance products on heterogeneous landscapes in the Heihe River Basin, China. Due to the fact that the footprint of ground observation albedo is much smaller than the spatial resolution of albedo products, high-resolution albedo images are used as amplification bridges to reduce scale differences. Based on this scheme, we propose the results of MODIS and GLASS accuracy evaluation. The verification results indicate that the root mean square of MODIS and GLASS measurements over a large area and a whole year is less than 0.05.
The performance of three error correction models, RF, CDF, and KF, was evaluated due to their wide applicability. To achieve this, 70% of the data (i.e. matching MCD43A3, pixel scale ground truth, and auxiliary data) is used to train an error correction model, which is defined as the training dataset. The remaining 30% of the data is used to validate the error correction model, defined as the validation dataset.
The data quality is good.
CSTR:11738.11.NCDC.ZENODO.DB6461.2024
10.5281/zenodo.6702596
