Document Actions
Experiences with the new RapidEye satellite for forest damage assessment
In the context of a comprehensive scientific investigation it was tested whether an (as automated as possible) assessment of snow breakage in spruce stock using satellite image data (RapidEye) is possible. The study area in the western part of the Erzgebirge (Saxony, Germany) was, for comparative purposes, also well documented with other data (stereo and ortho aerial images, before and directly after the occurrence of damage, forest stock data, data of site survey etc.).
The project constituted a big challenge in several aspects.
- There were only few experiences with the new RapidEye imagery for complex inventory tasks. The satellite system was then only a few months in operation.
- The lighting conditions were not optimal at the date of capture (2009/4/12). A very low altitude of the sun resulted in an inferior radiometric quality and furthermore clouds and cloud shadows covered parts of the investigation area.
- The damage symptoms in the stock were not specific: gaps in crowns, which can also occur by other causes such as thinning or old damages.
- The symptoms of snow breakage were not necessarily usable for remote sensing. Apart from damage as total loss of stock also individual trees or groves are involved; broken crowns do not leave necessarily gaps in the crowns because sometimes only crown tips are broken while the rest of the tree remains with a reduced crown.
- The spatial resolution of the RapidEye satellites (about 6m) usually does not permit to identify individual crowns.
Since the georeferencing of the satellite data was not sufficient and moreover difficult to correct a pixel based comparison was not possible. Thus only larger areas could be compared. Even the identification of training and validation areas for the classes of "soil", "shadow" and "crown" was difficult due to their distribution at a small scale and the low influence of small scaled damages on the mixed signatures. Finally the parameter "incompleteness" (or "level of crown aggregation") was analysed on stock level as (regrettably not strictly specific) indicator for snow breakage. The supervised classification methods "Maximum Likelihood" and "SAM" generated feasible results. Unsupervised classification methods achieved almost the same quality of results. The method "Linear Spectral Unmixing" was even better for the specific objective, since the percentage of soil specific signatures can be calculated pixel by pixel. The extrapolation to the stock results in the level of crown aggregation with an average variance of 7% compared to a visual evaluation in the area.
RapidEye is continuously improving the radiometric and geometric quality of their satellite data so that advanced applications are expected for the future. For the specific matter of the "assessement of snow breakage" satellite and aerial images with a better resolution as e.g. Quickbird images would be more appropriate, but a short-termed data provision can be expected with a higher probability from the system RapidEye. The confusion of snow breakage and clearance in the stock by reason of other causes (old damages, thinning) can be overcome only by identifying the transformation of the level of crown aggregation (before and after the occurrence of damage).
