K-NN FOREST: a software for the non-parametric prediction and mapping of environmental variables by the k-Nearest Neighbors algorithm

Abstract

In the last decades researchers investigated the possibility of extending the information collected in sampling units during a field survey to wider geographical areas through the use of remotely sensed images. One of the most widely adopted approaches is based on the non-parametric k-Nearest Neighbors (k-NN) algorithm. This contribution describes the software K-NN FOREST we developed to provide a complete tool for the implementation of the k-NN technique to generate spatially explicit estimations (maps) of a response variable acquired in the field by sampling units through the use of remotely sensed data or other ancillary variables. K-NN FOREST is designed to guide the user through a graphic user interface in the different phases of the process. K-NN FOREST is freely available for download and it is designed to run under Windows environment in conjunction with the GIS software IDRISI.

Keywords:

• Environmental Inventory and Mapping
• Prediction
• Remote Sensing
• k-Nearest Neighbors