Optimized sampling strategy of Wireless sensor network for validation of remote sensing products over heterogeneous coarse-resolution pixel

Abstract:

Coarse-resolution satellite albedo products are increasingly applied in geographical researches because of their capability to characterize the spatio-temporal patterns of land surface parameters. In the long-term validation of coarse-resolution satellite products with ground measurements, the scale effect, i.e., the mismatch between point measurement and pixel observation becomes the main challenge, particularly over heterogeneous land surfaces. Recent advances in Wireless Sensor Networks (WSN) technologies offer an opportunity for validation using multi-point observations instead of single-point observation. The difficulty is to ensure the representativeness of the WSN in heterogeneous areas with limited nodes.

In this study, the objective is to develop a ground-based spatial sampling strategy through consideration of the historical prior knowledge and avoidance of the information redundancy between different sensor nodes. Taking albedo as an example. First, we derive monthly local maps of albedo from 30-m HJ CCD images a 3-year period. Second, we pick out candidate points from the areas with higher temporal stability which helps to avoid the transition or boundary areas. Then, the representativeness (r) of each candidate point is evaluated through the correlational analysis between the point-specific and area-average time sequence albedo vector. The point with the highest r was noted as the new sensor point. Before electing a new point, the vector component of the selected points should be taken out from the vectors in the following correlational analysis. The selection procedure would be ceased once if the integral representativeness (R) meets the accuracy requirement. Here, the sampling method is adapted to both single-parameter and multi-parameter situations.

Finally, it is shown that this sampling method has been effectively worked in the optimized layout of Huailai remote sensing station in China. The coarse resolution pixel covering this station could be well-represented by six nodes with an integral representativeness high to 99.1%. The measurements have been applied to the validation of MODIS albedo products for crop. Comparison results demonstrate that the RMSE of MODIS albedo products in growing season is 0.011, in harvest time was nearly 0.05, and for bare soil in winter is 0.009.