BRDF Applications in Semiarid Grassland Monitoring with the AVHRRs

---

---

BRDF Modelling applied to entire AVHRR subscenes over Xilingol Aimag, Inner Mongolia P.R.C.

The effects of BRDF adjustment and BRDF modelling on community type differentiation applications using 1km AVHRR were assessed by establishing ten training areas for extraction of visible and near-infrared channel signatures from :

1. uncorrected AVHRR visible and near-infrared (21 orbits)
2. corrected or "adjusted" AVHRR visible and near-infrared (21 orbits)
3. modelled scenes (2 channels)
4. the model's anisotropic parameters (4 channels)
5. a maximum NDVI value composite using just the PM scenes and
6. a maximum NDVI value composite using both AM and PM scenes

The geographical area covered by the subscenes is bounded by 112.4 - 116.1 degrees east by 42.8 - 44.7 degrees north. The training areas comprise 64 samples, cloud-screened to provide 900 uncontaminated samples from the 21 orbits of uncorrected and corrected AVHRR; the single modelled scene does not require screening and so comprises 64 samples per training area (per channel). The majority of training sites correspond to grids where field radiometry was carried out in August 1996; the remainder were chosen as representative of distinct grassland and surface types in this region.

Follow these links for

• locations of the training sites
• information about grassland types and the training sites
• a VISNIR scatterplot from uncorrected AVHRR (using 21 orbits)
• a VISNIR scatterplot from corrected AVHRR (using 21 orbits)
• a VISNIR scatterplot from a modelled scene
• a VISNIR scatterplot from an MVC_NDVI composite