A Realistic Three-dimensional Scene Model (RTDS) for simulating the Surface Temperature over Sparsely Vegetated Surface from a Remote Sensor and First Results

Abstract:

Validations of the remotely sensed land surface temperature (LST) are necessary due to the strong sensitivity of temperature as an input parameter for many models. LST products with large errors may lead to significant uncertainties in subsequent applications. Most of the current validation methods rely on direct LST comparisons between the ground measurements at a single site and its corresponding pixel. However, due to the scale-mismatch, it is inappropriate to use the single-site data to evaluate the whole pixel value, especially over sparsely vegetated pixels or other heterogeneous pixels. Based on the geometric optical model, a ground-based realistic three-dimensional scene model (RTDS) is developed by taking into account the sun-ground target-sensor geometry, to calculate the surface temperature for a mixed pixel. In this model, the ground-based realistic three-dimensional scene is reconstructed, and the surface of the scene is divided into little surface elements. The sun light is treated as parallel rays in order to calculate the Boolean value of each surface element observed by the remote sensor to judge shadow or illumination, and thus we get the fraction of each component of the scene using face-factor. The RTDS modeled LST has the same extent as the field-of-view of the remote sensor. Therefore, it is comparable with the estimates of remotely sensed LST. The measurements over a sparsely vegetated surface in the downstream area of Heihe River basin are used to test the RTDS model. The results show that the calculated LST with the RTDS model agree well with the ground measurements. When the size of element is less than 0.1 m², the modeled surface temperatures have negligible differences. The influences of different distributions of ground scene on the results are analyzed and the results indicate that the distribution of ground scene has a great impact on the calculated surface temperature. The proposed RTDS model provides a bridge to solve the problem of scale-mismatch between the ground measurement and the pixel value for heterogeneous surfaces. It is helpful to the validation of remotely sensed land surface temperature products.