Forest stand volume is the sum of growing stock of all the living trees in forest land, which is an essential element in the National Forest Inventory. It is an important indicator of the total quantity of regional and national forest resources. Explicitly spatial information on forest stand volume is critical for estimation of above ground biomass for quantifying carbon sequestration and carbon dioxide exchange. However, it is costly and spatially limited to estimate it by the conventional field-based approach.

A study led by Ph.D. candidate CHEN Lin and Prof. REN Chunying from the Northeast Institute of Geography and Agroecology of the Chinese Academy of Sciences build a hybrid model for mapping forest volume by remote sensing modeling. The study was based on multi-source satellite and inventory data for spatially continuous and temporally uniform predictions.

The hybrid model makes up for the shortcoming of machine learning algorithm, which is a well adopted method to generate explicit spatially estimates of forest stand volume but ignores the spatial autocorrelation of neighboring observed data.

Researchers develops the hybrid model as Support Vector Machine for Regression Kriging (SVRK) to map stand volume of the Changbai Mountains mixed forests covering 171,450 ha area based on a small training dataset.

This Support Vector Machine for Regression Kriging (SVRK) model consists of Support Vector Machine for Regression (SVR) and its residuals interpolated by Ordinary Kriging, which considered both the environmental factors and the spatial autocorrelation of neighboring observed data. The SVRK model improved the accuracy of 9% than SVR based on Root Mean Square Error (RMSE) values. Topographic indices from L band Interferometric Synthetic Aperture Radar (InSAR), backscatters of L band Synthetic Aperture Radar (SAR), and texture features of VV channel from C band Synthetic Aperture Radar (SAR), as well as vegetation indices of the optical sensor were contributive to explain spatial variations of stand volume.

This study shows that SVRK was a promising approach for mapping stand volume in the heterogeneous temperate forests with limited samples. “This method has the potential to retrieve stand volume with archived Advanced Land Observing Satellite (ALOS) and Sentinel imagery and determine the temporal variation of stand volume at a large scale. It can be used to map stand volume with limited samples and support proper management of forest stand” said Prof. REN Chunying.

The study entitled "Multi-Sensor Prediction of Stand Volume by a Hybrid Model of Support Vector Machine for Regression Kriging" has been published online in Forests.