Automatic multi-sensor forest vegetation height mapping methods for Tanzania

This note describes new methods for automatic mapping of forest cover in Tanzania, in the form of yearly estimates of average vegetation height, from time-series of Landsat, Palsar and Sentinel-2 satellite images.
Multispectral images of 30 m and 10 m resolution may be downloaded from the Landsat and Sentinel-2 archives, respectively. A number of processing steps are applied in order to estimate the mean vegetation height within each pixel: (1) cloud masking, (2) scaling of digital numbers to top-of-the-atmosphere reflectance, (3) computation of the specific leaf area vegetation index (SLAVI), (4) estimation of vegetation height, and (5) Kalman filtering to reduce the variance of the estimates.
By using airborne laser scanning (ALS) data and Landsat-8 data from 2014, a regression between average vegetation height and the SLAVI vegetation index is established. If only one Landsat image is used, then the variance is high. However, by using all available Landsat acquisitions of the same area within one year, and producing a yearly estimate of vegetation height, the variance is reduced. The variance may be further reduced by applying Kalman filtering on the sequence of yearly estimates from 1985 to date. If L-band SAR data is available, as is the case with Palsar data from 2007-2010, then a multisensor version of the method may be used to reduce the variance for those years.
The method has been implemented in an automatic processing chain and evaluated with airborne laser scanning data from 2012 and 2014. Estimation of mean vegetation height is indeed possible, but lower variances in the estimates are needed. For 2015 and onward, this may be obtained by including Sentinel-2 data in the processing chain. We plan to quantify the improved performance of this in a future study.