Annex C. Previous NFI inventory design 12 km 2 3 1000 m Location system of the first stage sample clusters. 12 x 12 km grid applied. 30 m 1 Design of the first stage cluster. Field measurements carried out if the lower left second stage cluster occurs in forest. 4 50 m Design of the second stage cluster with three square plots, 30 x 30 m each. 53
Annex D: Inventory Design for the <strong>FRA</strong> in <strong>Nepal</strong> 1. General There will be a 2-phase sampling with stratification and in some areas even 3-phase sampling with stratification. The country is stratified to few sampling strata by e.g. physiographical zones. Different sampling intensity and design can be applied for the different strata. HR and VHR satellite images are used for the 1 st phase sampling to classify clusters and sample plots as regards forest coverage. Based on that, sample plots for the field inventory work will be defined. The 2 nd phase sampling refers to the actual clusters and sample plots to be measured in the field. Sample plots are used for soil, site, forest, biomass and biodiversity measurements, partially also for NTFP measurements. The TOFs are defined using the same 2-phase sampling design as for forest, possibly completed by additional measurements. For biodiversity measurements (mapping) line inventories inside clusters will be considered. The assessment of NTFPs requires additional field work, which should be adjusted with seasonality and existence of NTFPs. Partly the NTFP assessment is based on indicator habitats and species. The LiDAR Working Area will be laser scanned (wall-to-wall or strips) and classified using measured 2 nd phase field data, possibly completed with additional plots. For the LiDAR Working Area, either very accurate wall-to-wall forest resource map or improved forest resource statistics in tabular form will be achieved, depending on the information need and the selected scanning strategy. 2. Two - Phase sampling with stratification 1 st Phase sampling On general level medium resolution satellite images are available for the entire country to have “wall-to-wall” overview of areas. The ALOS/Avnir satellite images (resolution 20 m) could be available at the Ministry of Land Reform and Management for another project, but options should be found to have them also free of charge available for the <strong>FRA</strong>. The Project will have for the most forest covered areas high resolution (HR, resolution around 5 m) or even very high resolution (VHR, resolution around 1 m) remote sensing data. For the images, there are various options available like ALOS/Prism, IKONOS/Geo, RapidEye, GeoEye, IRS, Quick Bird or Kompsat. Additional option is to have LiDAR data scanned from the most important areas. A large number of sample points could be assessed on the images as the first phase sample. It could be assumed that broad land use classes (e.g. forest, other wooded land, other land with tree cover, treeless land, built up land) can be classified with visual interpretation very accurately, but forest types cannot be classified with a reasonable accuracy. In addition, the accessibility (Note: definition needed) using the RS data & digital maps will be assessed. As an example, Google Earth with a KML application for the interpretation could be used. Having with the digital RS data, digital maps and the KML application, the visual classification of the first phase sample is very quick. This means that a large number of 1 st phase sample points could be spotted to guarantee accurate estimates for the total forest area, other wooded land, other land with tree cover (TOFs), other land, and built up land. While starting the work a proper intensity of visual classification should be decided. A grid of 6 km by 6 km for the 1 st phase sample point clusters gives us 4 083 clusters (for the entire country). If there are 6 points per cluster, the total number of 1 st phase sample points would be 24 500. Most of the points (clear forest and clear non-forest) are very easy to classify. Some points need more time. Assuming 2 minutes per cluster the required workload is around one man-month for the classification task (equals to 140 hours). Respectively, a grid of 4 km by 4 km for the 1 st phase sample point clusters gives us 9 188 clusters (for the entire country). If there are 6 points per cluster, the total number of 1 st phase sample points would be 55 100. Assuming once more 2 minutes per cluster the required workload is around two and half man-months for the classification task (equals to 310 hours). 2 nd Phase sampling The second phase sample is a sub-sample of the 1 st phase sample. Clusters selected for the 2 nd phase are measured in the field. The idea of clustering is to reduce travelling costs between sample plots. In the mountain areas clusters should work as camp units: one cluster of sample plots should correspond e.g. 4 - 6 days of work (+ 2–3 days travel/walking time to reach the cluster). If the clusters have 6 plots with 200 or 300 m distance, in easiest mountain terrains one might measure the cluster in 3 - 4 days. If the terrain is difficult, 6 days are needed. Such variability cannot be avoided. In the Terai and lowest hills, 54