kNN

Estimating Forest Inventory Using Satellite Imagery

EARLY RESULTS

Cover Type Classification

Maps with variables such as cover type or volume can be produced based on satellite imagery, if an adequate number of field plots are available. The cover type map on the left has been derived using three dates of Landsat imagery, a forest/non-forest mask, an upland/lowland mask, three nearest neighbors, and a band optimization routine. The field data is from USDA and the cover types classified here are the same as those used in the USDA Forest Inventory and Analysis. Click the image to view a larger version.

Volume Estimation

Three nearest neighbors have also been used for this volume map. Both land use and lowland masks were utilized and the data was run through band optimization. All thematic maps produced with the kNN method have a the pixel size of 30 meters, based on the Landsat imagery that, together with the field data, are the basis for these classification. The white areas inside maps are either non-forested lands or clouds. Click the image to enlarge it.

Volume Estimation: Before and After

Click the image to see the satellite image before and after volume estimation.