University of Twente Student Theses
LiDAR remote sensing of aboveground biomass using a plot-based approach in the tropical forest of Nepal: A comparison of regression and geo-statistical approach
Adefurin, Olusola (2012) LiDAR remote sensing of aboveground biomass using a plot-based approach in the tropical forest of Nepal: A comparison of regression and geo-statistical approach.
PDF
2MB |
Abstract: | Tropical forests contribute significantly to the global carbon balance, attaining high rate of carbon sequestration through their huge densities. In order to develop a better understanding and quantification of carbon stocks and flux dynamics, estimation of aboveground biomass becomes very crucial. Remote sensing technologies have proven superiority over other methods because of their extensive coverage capability making estimation of aboveground biomass possible over broad spatial scales. To achieve broad scale mapping with remote sensing, statistical relationship between sensor-extracted tree parameters and field measurements are used. With the advancement in remote sensing technology, laser scanning evolved with the possibilities of acquiring three dimensional information of the forest structure. The aim of this study is to demonstrate the use of low density data (less than 1 point/m2 ) for the estimation of aboveground biomass in the tropical forest of Nepal (Kayarkhola watershed in Chitwan Province) while comparing between regression and geo-statistical approach (ordinary kriging and regression kriging).. The plot-based approach was adopted and 81 metrics were extracted using LiDAR’s elevation and intensity values including canopy density computations. Intensity metrics were excluded from further regression analysis because of their poor relationship with aboveground biomass. Data reduction technique (PCA) was used to select independent and uncorrelated LiDAR metrics. Seven predictors were selected including height maximum, height average and absolute deviation, height L-moment (L2), Height L-moment skewness, 40th, 80th, and 95th height percentile. For this study, it was possible to employ only the 95th height percentile for predicting aboveground biomass without additional variables. The model of 95th height percentile and aboveground biomass showed a moderately strong relationship with aboveground biomass, with an R2 of 0.54. The performance of the two approaches was assessed using their RMSE values and ME estimates. Using regression kriging, this study showed an improvement in the accuracy prediction of aboveground biomass with a lower RMSE of 0.20 LN(Mg/ha) and ME of 0.00023 LN(Mg/ha). Regression analysis resulted in an RMSE 0.68 LN(Mg/ha) and ME of 0.42 LN(Mg/ha). Regression kriging showed an improvement in the estimation because of its ability to account for some variations in aboveground biomass. |
Item Type: | Essay (Master) |
Faculty: | ITC: Faculty of Geo-information Science and Earth Observation |
Programme: | Geoinformation Science and Earth Observation MSc (75014) |
Link to this item: | https://purl.utwente.nl/essays/93661 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page