Abstract
A workflow for combining airborne lidar, optical satellite data and National Forest Inventory (NFI) plots for cost efficient operational mapping of a nationwide sample of 5 × 5 km squares in the National Inventory of Landscapes in Sweden (NILS) landscape inventory in Sweden is presented. Since the areas where both satellite data and lidar data have a common data quality are limited, and impose a constraint on the number of available NFI plots, it is not feasible to perform classifications in a single step. Instead a stratified approach where canopy cover and canopy height are first predicted from lidar data trained with NFI plots is proposed. From the lidar predictions a forest stratum is defined as grid cells with more than 3 m mean tree height and more than 10% vertical canopy cover, the remaining grid cells are defined as open land. Both forest and open land are then classified into broad vegetation classes using optical satellite data. The classification of open land is trained with aerial photo interpretation and the classification of the forest stratum is trained with a new set of NFI plots. The result is a rational procedure for nationwide sample based vegetation characterization.
Highlights
The National Inventory of Landscapes in Sweden (NILS) is a sample based nationwide landscape inventory [1,2]
The aim of the present study is to develop and evaluate a cost efficient method for semi-automated mapping of the nationwide sample (n = 631) of NILS 5 × 5 km squares by using the combination of airborne lidar data and optical satellite data (Landsat 8 and SPOT 5)
The lidar predictions of tree height and canopy cover were used to stratify the NILS 5 × 5 km squares in the test area into open land and forest areas, where forest was defined as having more than 10% canopy cover and trees over 3 m in height
Summary
The National Inventory of Landscapes in Sweden (NILS) is a sample based nationwide landscape inventory [1,2]. The aim of NILS is to be a data source for monitoring the fulfilment of some of the Swedish national environmental objectives [3], as well as for international habitat reporting and for research. NILS data are collected from a stratified sample of 631 clusters. Each cluster contains 12 field surveyed sample plots within a 1 × 1 km square that is photo interpreted. The original ambition was, to do a photo interpretation of the 5 × 5 km square, in order to obtain data that could be used for providing the landscape context to the data collected in the field and in the inner square. One potential data source is optical imagery from earth observation satellites such as SPOT or Landsat, which have been much used for land cover classifications in the past [4,5,6,7,8]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
