Abstract
Semi-operative or even operative methods are already available for the estimation of the stem volume of forests. However, a drawback of most of the methods is that they need extensive ground data. This paper presents a new method for the estimation of the forest stem volume. The method is based on fitting a stem volume model function to Landsat ETM+ reflectance data and ground data. Separate models were estimated for coniferous and broad-leaved trees. The models were validated using Landsat ETM+ data and ground data from another study site. The root mean square errors in the two study sites for coniferous trees were 50.5 m3/ha and 62.2 m3/ha. The errors for the broad-leaved models were 12.7 m3/ha and 78.4 m3/ha, respectively. The best models were applied to a Modis image mosaic of Finland and Sweden. The resulting estimates were compared to the Finnish and Swedish forestry statistics. The results were better in Finland than in Sweden. For the whole Finland the total bias of the average stem volume was -2.3 m3/ha. In total stem volume the bias was -34.4 mill. m3. In Sweden the total bias of the average stem volume was -15.6 m3/ha, which caused a bias in the total stem volume of -598.0 mill. m3. A major reason for the large underestimation in Sweden was the remaining clouds in the mosaic.
| Original language | English |
|---|---|
| Title of host publication | ForestSAT 2002 Conference |
| Subtitle of host publication | CD-ROM |
| Number of pages | 11 |
| Publication status | Published - 2002 |
| MoE publication type | Not Eligible |
| Event | ForestSAT 2002 Conference: Operational tools in forestry using remote sensing techniques - Edinburgh, United Kingdom Duration: 5 Aug 2002 → 9 Aug 2002 |
Conference
| Conference | ForestSAT 2002 Conference |
|---|---|
| Country/Territory | United Kingdom |
| City | Edinburgh |
| Period | 5/08/02 → 9/08/02 |
Keywords
- stem volume
- landsat ETM+
- modis