Inventories of tropical forest aboveground biomass (AGB) are often imprecise and sparse. Increasingly, airborne Light Detection And Ranging (LiDAR) and satellite optical wavelength sensor data are used to map tree height and to estimate AGB. In the tropics, cloud cover is particularly prevalent and so several years of satellite observations must be considered. This may reduce mapping accuracy because of seasonal and inter-annual changes in the forest reflectance. In this paper, the sensitivity of airborne LiDAR and Landsat-8 Operational Land Imager (OLI) based dominant canopy height and AGB 30 m mapping is assessed with respect to the season of Landsat acquisition for a ~10,000 Km2 tropical forest area in the Democratic Republic of the Congo. A random forest regression estimator is used to predict and assess the 30 m dominant canopy height using LiDAR derived test and training data. The AGB is mapped using an allometric model parameterized with the dominant canopy height and is assessed by comparison with field based 30 m AGB estimates. Experiments are undertaken independently using (i) only a wet season Landsat-8 image, (ii) only a dry season Landsat-8 image, and (iii) both Landsat-8 images. At the study area level there is little reported sensitivity to the season of Landsat image used. The mean dominant canopy height and AGB values are similar between seasons, within 0.19 m and 5 Mg ha−1, respectively. The mapping results are improved when both Landsat-8 images are used with Root Mean Square Error (RMSE) values that correspond to 18.8% of the mean study area mapped tree height (20.4 m) and to 41% of the mean study area mapped AGB (204 Mg ha−1). The mean study area mapped AGB is similar to that reported in other Congo Basin forest studies. The results of this detailed study are illustrated and the implications for tropical forest tree height and AGB mapping are discussed.
Democratic Republic of the Congo Tropical Forest Canopy Height and Aboveground Biomass Estimation with Landsat-8 Operational Land Imager (OLI) and Airborne LiDAR Data: The Effect of Seasonal Landsat Image Selection
