Satellite Albedo products Validation by Upscaling Multi-nodes in situ Data into a Satellite Pixel Scale over Heterogeneous Land Surface

Abstract:

Land surface albedo is a key parameter for energy budgets. There are many available products from remote sensing sensors, such as Moderate Resolution Imaging Spectroradiometer (MODIS), Visible Infrared Imaging Radiometer Suite (VIIRS), and Advanced Very High Resolution Radiometer (AVHRR) and so on.

Their accuracy should be carefully quantified before being used. Most validations directly use a single-point in situ measurement in the relatively homogeneous land surface. However, it is not valid over heterogeneous cases. A multi-scale validation strategy using a high-resolution albedo imagery as a bridge is alternative, with several uncertainties from high-spatial-resolution albedo imagery, geometric registration, and the upscaling process. It results a relative precision. Hence, for more effective validation, the albedo absolute value based on ground measurements is still required, which can be conceptualized as the “truth” value of pixel scale albedo.

In this study, a sampling strategy based on using wireless sensor network (WSN) technology to measure albedo at multiple nodes is proposed to capture the land surface heterogeneity in Huailai remote sensing test station, Hebei province, China, which is one station of a Chinese validation network (fig. 1).

The nodes are distributed in an optimal layout determined by a sequential selection method using theirs representativeness. The first six nodes with the highest degree of representativeness are finally selected (fig. 2). Upscaling functions with different weights for each node, calculated by the ordinary least squares (OLS) linear regression, are used to upscale them to a coarse pixel scale.

Application is exemplified by the validation of the MODIS albedo product (fig. 3), and VIIRS albedo product (fig.4), from Jul. 18, 2013 to Jul. 31, 2014. The RMSEs are 0.025 and 0.020 for MCD43B3 full inversion and magnitude inversion, respectively. The overall accuracy of VIIRS albedo is 0.021 and 0.014 under clear sky and strict clear sky, respectively.

Compared with the multiple scale validation strategy, the WSN albedo “truth” at t pixel scale reveals a good quality both in stability and continuity. And this validation strategy has been one key element in the LAnd surface remote sensing Product VAlidation System (LAPVAS) in China.