Testing the impact of using auxiliary variables in the definition of k-Nearest Neighbor imputation

In this study, we use the k-Nearest Neighbor (kNN) method to test the potential of computing Georgia forest volumes based from the FIA plot data and Landsat ARD with auxiliary variables such as maximum and minimum temperature, land cover, and elevation. The total number of plots used is 941, we divided them into the training dataset (708) and testing dataset (233). The training dataset was used to calibrate the kNN model, and the testing dataset was used to test the model. The results showed that the best kNN model was a combination of ARD and all auxiliary variables, the RMSE (%) and bias are 77.61% and -5.40%, respectively. Using the derived results, we have created a volume map, used subsequently for the county-level volume estimates.