Fuel value indices of fuelwood species and accuracy of K-nearest neighbour technique in woody biomass estimation in Masindi and Nebbi Districts, Uganda

Abstract

Biomass is still the major energy source in developing countries like Uganda. This implies that prioritizing high fuel potential fuelwood species and routine monitoring of stocks is paramount. This study identified ten commonly used fuelwood species, examined their Fuel Value Indices and assessed the accuracy of K-NN technique for estimating woody biomass in the savannahs of Uganda. The Fuel Value Indices were determined from percentage moisture content, density and gross calorific value. Systematic point sampling in TNTmips Software was used to locate 26 field plots. The coordinates of the points were extracted and recorded in a hand held Garmin Global Positioning System. Plots (50 x 50 m) were demarcated within each point and tree height, diameter at breast height and crown cover measured. The generic biomass equations for Uganda were used to determine the individual tree and plot biomass. A GeoEye satellite image (0.5 m) was used in the K-Nearest Neighbour analysis. The percentage moisture content and density varied significantly amongst species (F(df=9)=92.927, P=0.0001) while gross calorific value does not. The highest correlation coefficient between K-NN and field plot estimates was 14.8 % (Root Mean square Error=43 and K=1) using Euclidean Distance. The Fuel Value Indices ranged from 1.10 in Ficus natalensis to 13.09 in Albizia grandibracteata making the latter the most suitable fuelwood species. The use of GeoEye in the K-NN technique poorly estimates the woody biomass stocks. There is need to test it with other satellite datasets such as LiDAR, LANDSAT 5TM, LANDSAT 7 ETM+.