Satellite images have been successfully applied in the Forest Resource Assessment 1990 and 2000 in tropics. So far, the global wall-to-wall forest cover maps have been based on low or medium resolution images. It is expected that global wall-to-wall mapping applications with high-resolution images will appear in few years (Zhu and Walter 2001). However, this might not reduce the need for an independent remote sensing aided forest resource survey. The strength of this type of scientific survey would be the forest inventory with exact FAO definitions and estimates with current state and changes, together with standard errors, which are computed under strict quality control. A representative field sample would support this goal. Country capacity building could be an additional benefit of this type survey. Existing wall-to-wall forest cover maps could be utilised in planning the design of the survey. For a possible global survey, its role should be assessed against the existing and coming other global surveys.
The satellites launched after the FRA 1990 and FRA 2000 surveys, and new satellites under planning, increase noticeably the availability of optical area satellite images for a possible remote sensing aided survey in FRA 2010. The most interesting satellite programs are Landsat and MODIS, and possibly some programs with very high-resolution images, like Ikonos and QuickBird. Landsat 7 with the ETM+ sensor was launched in 1999. The next Landsat launch has been planned for year 2006 (Appendix 2). Imagery from Spot and IRS-1 satellites can be used for completing the image cover if Landsat images are not available, e.g., due to cloud cover. The medium resolution instruments, e.g., MODIS, MERIS and IRS-1C/D WiFS can be used for global wall-to-wall mapping. The spatial resolution for estimating changes may be problematic. The instruments have, however, good potential for planning global remote sensing and field sampling designs.
Sampling might be the only relevant approach with high-resolution images (Landsat ETM+, Spot, IRS-1 LISS-III). The number of the needed Landsat ETM+ scenes with 10 % sampling for the entire globe is about 400 - 450 and the estimated costs about 255 000 US$. The price can be compared to the costs of pure field measurement based survey. The minimum field measurement costs for this type of survey could be around 10 million US$ and likely costs around 100 million US$ (Table 5, Chapter 5).
The operational use of satellite images in forest inventories began about ten years ago, in spite of the fact that satellite image research has been active almost 30 years. The progress in estimation, error estimation method and image analysis methods can be utilised in the survey.
The main emphases in our study are 1) outlining how the evaluation of a sampling design in a remote sensing based survey could be accomplished and 2) presenting rough error estimates for the variables 'Area of Forest land' and 'Area of Other wooded land' in our test area (Europe and CIS). A hypothetical simulation model based on FRA 2000 land cover map was applied. Our study shows promising results in the applied test area, e.g., high resolution and very high resolution image based forest survey can meet the accuracy requirements of a possible independent remote sensing aided forest survey with a moderate costs. The final conclusions will need further analysis and more resources than what was available in this study. The procedure described here could be repeated in the other regions of the globe as well, with some other remote sensing based material, and with some other variables than which were used here.
One fact which should be taken into account in planning the remote sensing based survey is the availability of satellites images from an earlier time point, particularly if very high resolution satellite images will be used. Very high-resolution images are not necessarily archived and the images are taken only upon request. There are two different options to handle this problem, 1) either to order the image from time 0 early enough, or 2) to develop a multi-resolution technique for change detection part (which can, of course, be also utilised in analysing the current status). The early order presumes fast decisions and might be unrealistic taking into account the current status of very high-resolution satellites. Thus the option 2 in which high resolution images are acquired from time 0 and very high resolution together with high-resolution images are acquired from time 10 may be a more realistic option.
The existing and upcoming medium resolution satellite images, in addition to existing and coming wall-to-wall land cover maps, provide information sources for planning both a possible field measurement based and remote sensing based sampling designs. Based on this fact and our experiment, we are convinced that an independent remote sensing aided global forest survey could be carried out with reasonable costs. We are also convinced that this type of inventory would have its role both in independently testing the accuracy of estimates obtained from other sources and in country capacity building.
