Mapping stand volumes of Pinus halepensis Mill in a semi-arid region using satellite imagery of the Sénalba Chergui forest in north-central Algeria

We developed an approach using remote sensing and modeling, applicable to Algerian forest inventory, for estimating the volume of timber in Aleppo pine stands. We used ordinary linear regression (OLR) and reduced major axis (RMA) regression to assess an operational model to map stand volume from satellite images. Our analysis was supported by measurements from 151 sample plots and spectral values from remote sensing imagery. Fifteen candidate models were tested through the Akaike Information Criterion to assess their predictive power. For the 2009 Landsat TM image, we found that the best models for both regression methods used the NDVI as the independent variable. The RMSEs were 20.3% (16.10 m3 ha−1) and 22.5% (17.83 m3 ha−1), respectively, for OLR and RMA. We chose the RMA regression models because they had realistic standard deviation values for the estimated volumes, and they gave lower RMSEs in volume classes over 40 m3 ha−1. Our method gave similar results for two other images, which demonstrated that our approach was robust when applied to data from a different year (2006 Landsat TM), but from the same sensor, and also to data from a different sensor (2005 Alsat-1).

► Method for forest inventory of countries in semi-arid environments.
► Use of satellite images to map forest volume.
► Alternative to conventional forest inventory.
► Linear regression to model stand volume with spectral bands and indices.