Estimation of forest volumes by integrating Landsat TM imagery and forest inventory data

Accurate information about forest volumes is essential for forest management planning. The survey interval of the Forest Resource Inventory of China (FRIC) is too long to meet the demand for timely decision-making required for forest protection, management, and utilization. Analysis of satellite imagery provides good potential for more frequent reporting of forest parameters. In this study, we describe an application of the k-nearest neighbors (kNN) method to Landsat TM imagery for improving estimation of forest volumes. Several spectral features were tested and compared in forest volume estimations, including normalized difference vegetation index, environmental vegetation index, and the combination of the spectral features. The combined index resulted in the most accurate volume estimations. The kNN estimator and the combined index were then used in forest volume estimation. The estimation error (RMSE) of the total volume was 44.2%, much lower than those for Larix forest (the RMSE was 51.7%) and those for the Korean pine and broadleaved forests (the estimation errors were over 71.7% and 88.19%, respectively). This preliminary study demonstrates the potential of forest volume estimations with remote sensing data to provide useful information for forest management if only limited ground information is available.

Estimation of forest volumes by integrating Landsat TM imagery and forest inventory data

Accurate information about forest volumes is essential for forest management planning. The survey interval of the Forest Resource Inventory of China (FRIC) is too long to meet the demand for timely decision-making required for forest protection, management, and utilization. Analysis of satellite imagery provides good potential for more frequent reporting of forest parameters. In this study, we describe an application of the k-nearest neighbors (kNN) method to Landsat TM imagery for improving estimation of forest volumes. Several spectral features were tested and compared in forest volume estimations, including normalized difference vegetation index, environmental vegetation index, and the combination of the spectral features. The combined index resulted in the most accurate volume estimations. The kNN estimator and the combined index were then used in forest volume estimation. The estimation error (RMSE) of the total volume was 44.2%, much lower than those for Larix forest (the RMSE was 51.7%) and those for the Korean pine and broadleaved forests (the estimation errors were over 71.7% and 88.19%, respectively). This preliminary study demonstrates the potential of forest volume estimations with remote sensing data to provide useful information for forest management if only limited ground information is available.