Forest monitoring with airborne laser scanning in Tanzania

The research presented in this report was conducted in three separate study areas in Tanzania. First, an area in Amani Nature Reserve located in the Eastern Arc Mountains was selected to analyze to what extent airborne laser scanning (ALS) could be used to estimate biomass in dense sub-montane rainforests. A number of other issues arose during the course of the work, such as influence of pulse density of ALS systems on the quality of the digital terrain model (DTM) derived from the ALS data. The DTM is of fundamental importance because ALS data for the tree canopies are usually calculated as heights relative to the derived DTM surface. Another important issue was the influence of field plot size for precision of biomass estimates based on ALS. Second, a large study area in Liwale (Lindi District) was selected to quantify how well ALS data could characterize biomass in tropical savanna. The Liwale area is mainly miombo woodlands. The study area had a size of around 16,000 km2. It was also used to analyze the potential of use of ALS to estimate change in biomass over short time periods (two years) and to assess the cost-efficiency of use of ALS to enhance biomass and biomass change estimates. Third, a smaller area of 365 km2 in miombo woodlands was used to quantify the contribution of different remote sensing technologies to improve estimates of biomass. The technologies were (i) ALS, (ii) interferometric synthetic aperture radar (InSAR) derived from the TanDEM-X satellites, (iii) RapidEye optical imagery, and global forest map products derived from (iv) Landsat and (v) ALOS PALSAR L-band radar imagery. The overall findings in the project were that biomass observed on ground plots could be modelled with ALS-derived metrics such as canopy height and canopy cover as explanatory variables with model performance similar to what has been reported in other tropical studies. That even holds true for the high-biomass situations in the rainforest with recorded biomass densities up to around 1000 Mg/ha. Small plot sizes will tend to result in poorer models and therefore larger uncertainty of the final biomass estimates. For high-biomass situations plots larger than 2000 m2 seem to be suitable for improved biomass estimates. We did not evaluate the cost-efficiency of larger plots. Even for dense forests, it seems sufficient to use ALS data with a pulse density of 1 pulse per meter square. Lower densities should be avoided to ensure accuracy of the DTM suitable for extraction of canopy properties from the ALS data. For large area surveys for which acquisition of ALS data with complete coverage is cost prohibitive, sampling with ALS carried out by acquisition of ALS data along strips that may be spaced many kilometers apart, can improve the precision of biomass estimates by an order of magnitude compared to pure field sampling, like for example the national forest inventory of Tanzania (NAFORMA). In this situation ALS can also be highly cost-efficient. For change in biomass the cost-efficiency of use of ALS to enhance estimates was low, and pure field inventory is probably more cost-efficient – at least if the same field plots are visited on both occasions for which an estimate of change is sought. The current design of NAFORMA strongly restricts the benefits of remotely sensed data to enhance estimates. Further, transaction costs caused by problems with flight permissions to be granted by public authorities hamper efficient use of ALS technology in an operation setting in Tanzania. Different remote sensing techniques will have very different contributions to improve the precision of the biomass estimates, but the cost-efficiency of using different types of remotely sensed data remains an unexplored issue. Stratification of the entire Tanzanian land area in a manner that is consistent with the current NAFORMA stratification may open up for improved cost-efficiency of remote sensing.