Evaluation of digital surface model data to improve forest biomass estimation from SPOT HRG
Remote sensing techniques play a crucial role to upscale aboveground biomass estimates from local, regional to global scale. The objective of the present research was to use previously not evaluated canopy height model (CHM) data to enhance aboveground biomass estimation from SPOT HRG imagery (HRG)....
| Autor principal: | |
|---|---|
| Formato: | H1 |
| Lenguaje: | Inglés |
| Publicado: |
SLU/Dept. of Forest Resource Management
2010
|
| Materias: |
| _version_ | 1855570367313084416 |
|---|---|
| author | Hlaing, Myint |
| author_browse | Hlaing, Myint |
| author_facet | Hlaing, Myint |
| author_sort | Hlaing, Myint |
| collection | Epsilon Archive for Student Projects |
| description | Remote sensing techniques play a crucial role to upscale aboveground biomass estimates from local, regional to global scale. The objective of the present research was to use previously not evaluated canopy height model (CHM) data to enhance aboveground biomass estimation from SPOT HRG imagery (HRG). The different CHMs data evaluated were digital surface models mapped using photogrammetric processing of data acquired by the airborne Digital Mapping Camera from Zeiss/Intergraph (DMC), SPOT High Resolution Stereo (HRS) and Airborne Laser Scanning (ALS) data. The pixel sizes range from one half meter to twenty meter. The study site is the watershed of the Krycklan stream located in the North Eastern part of Sweden (Lat. 64°14’ N, Long. 19°50’ E). The study area covers approximately 7800 hectares and is characterized by boreal forest dominated by Norway spruce (Picea abies) and Scots pine (Pinus sylvestris).
The remotely sensed data derived spectral bands and canopy heights (CHMs) were used to fit regression models and to perform cross validation at plot level to estimate aboveground biomass. The resulting models were used to produce raster maps. Furthermore, accuracy assessment in terms of root mean square error (RMSE) of stand level estimations was computed based on an independent field measured dataset.
The adjusted R2 for stand level estimates of above ground tree biomass was 60% and the RMSE was 31.8% when using SPOT HRG alone. The corresponding values of CHM data were 23.0% R2 (adj) and 35.4% RMSE for SPOT HRS; 77% R2 (adj) and 18.8% RMSE for Z/I DMC; and 80.7% R2 (adj) and 20.2% RMSE for ALS respectively. The results of cross validation of all models comply with the standard limit falling between 1.04 and 1.15. The former corresponds to a model with one explanatory variable and the latter was for 5 or 6 explanatory variables.
Fusing the data sources of HRG and CHM improved aboveground biomass prediction in terms of both R2 and RMSE for all sensors data. For HRS, R2 improved from 23.0% to 50.2% and RMSE improved from 35.4% to 26.9%. R2 of Z/I DMC increased from 77.0% to 80.0% and RMSE improved from 18.8% to 16.9%. ALS derived canopy height measurements without vegetation ratio increased R2 from 80.0% to 84.5% and RMSE improved from 20.2% to 15.6%. Using ALS data including vegetation ratio decreased R2 from 90.5 % to 90.2% but RMSE improved from 15.7% to 14.1%. HRS and DMC increased the coefficient of determination and improved mapping accuracy when combined with the multi-spectral bands from HRG. ALS derived measurements had much higher R2 and accuracy when the canopy height was combined with vegetation ratio in estimating aboveground biomass. The use of digital CHM do appear promising to estimate dry biomass content and monitor carbon uptake for many important future applications.
|
| format | H1 |
| id | RepoSLU1928 |
| institution | Swedish University of Agricultural Sciences |
| language | Inglés |
| publishDate | 2010 |
| publishDateSort | 2010 |
| publisher | SLU/Dept. of Forest Resource Management |
| publisherStr | SLU/Dept. of Forest Resource Management |
| record_format | eprints |
| spelling | RepoSLU19282012-04-20T14:16:23Z Evaluation of digital surface model data to improve forest biomass estimation from SPOT HRG Hlaing, Myint aboveground biomass digital surface models remote sensing prediction mapping accuracy Remote sensing techniques play a crucial role to upscale aboveground biomass estimates from local, regional to global scale. The objective of the present research was to use previously not evaluated canopy height model (CHM) data to enhance aboveground biomass estimation from SPOT HRG imagery (HRG). The different CHMs data evaluated were digital surface models mapped using photogrammetric processing of data acquired by the airborne Digital Mapping Camera from Zeiss/Intergraph (DMC), SPOT High Resolution Stereo (HRS) and Airborne Laser Scanning (ALS) data. The pixel sizes range from one half meter to twenty meter. The study site is the watershed of the Krycklan stream located in the North Eastern part of Sweden (Lat. 64°14’ N, Long. 19°50’ E). The study area covers approximately 7800 hectares and is characterized by boreal forest dominated by Norway spruce (Picea abies) and Scots pine (Pinus sylvestris). The remotely sensed data derived spectral bands and canopy heights (CHMs) were used to fit regression models and to perform cross validation at plot level to estimate aboveground biomass. The resulting models were used to produce raster maps. Furthermore, accuracy assessment in terms of root mean square error (RMSE) of stand level estimations was computed based on an independent field measured dataset. The adjusted R2 for stand level estimates of above ground tree biomass was 60% and the RMSE was 31.8% when using SPOT HRG alone. The corresponding values of CHM data were 23.0% R2 (adj) and 35.4% RMSE for SPOT HRS; 77% R2 (adj) and 18.8% RMSE for Z/I DMC; and 80.7% R2 (adj) and 20.2% RMSE for ALS respectively. The results of cross validation of all models comply with the standard limit falling between 1.04 and 1.15. The former corresponds to a model with one explanatory variable and the latter was for 5 or 6 explanatory variables. Fusing the data sources of HRG and CHM improved aboveground biomass prediction in terms of both R2 and RMSE for all sensors data. For HRS, R2 improved from 23.0% to 50.2% and RMSE improved from 35.4% to 26.9%. R2 of Z/I DMC increased from 77.0% to 80.0% and RMSE improved from 18.8% to 16.9%. ALS derived canopy height measurements without vegetation ratio increased R2 from 80.0% to 84.5% and RMSE improved from 20.2% to 15.6%. Using ALS data including vegetation ratio decreased R2 from 90.5 % to 90.2% but RMSE improved from 15.7% to 14.1%. HRS and DMC increased the coefficient of determination and improved mapping accuracy when combined with the multi-spectral bands from HRG. ALS derived measurements had much higher R2 and accuracy when the canopy height was combined with vegetation ratio in estimating aboveground biomass. The use of digital CHM do appear promising to estimate dry biomass content and monitor carbon uptake for many important future applications. SLU/Dept. of Forest Resource Management 2010 H1 eng https://stud.epsilon.slu.se/1928/ |
| spellingShingle | aboveground biomass digital surface models remote sensing prediction mapping accuracy Hlaing, Myint Evaluation of digital surface model data to improve forest biomass estimation from SPOT HRG |
| title | Evaluation of digital surface model data to improve forest biomass estimation from SPOT HRG
|
| title_full | Evaluation of digital surface model data to improve forest biomass estimation from SPOT HRG
|
| title_fullStr | Evaluation of digital surface model data to improve forest biomass estimation from SPOT HRG
|
| title_full_unstemmed | Evaluation of digital surface model data to improve forest biomass estimation from SPOT HRG
|
| title_short | Evaluation of digital surface model data to improve forest biomass estimation from SPOT HRG
|
| title_sort | evaluation of digital surface model data to improve forest biomass estimation from spot hrg |
| topic | aboveground biomass digital surface models remote sensing prediction mapping accuracy |