The Bavarian Forest National Park is a protected area in the southeast of Germany, where forest stands are unmanaged and are subject to long-term undisturbed natural processes. Scientists have chosen this site to study ecological and environmental processes on various levels. Remotely sensed data collected by airborne and spaceborne sensors allows them to monitor forest parameters such as tree species mixture, carbon fluxes and forest structure. LiDAR is particular useful for retrieval of vertical forest structure thanks to the laser ‘s ability to penetrate the canopy and capture both over- and understorey of forest stands. Various metrics computed from discrete return LiDAR data under leaf-off and leaf-on conditions are used to assess the extent and development of trees, shrubs and regeneration layers in natural forests.
This project will extract relevant metrics from two sets of leaf-off and leaf-on LiDAR to make an accurate and unbiased estimation of canopy density in tree, shrub and herbal layers within the Bavarian Forest National Park. LAStools will be used to initially process the raw point cloud data and create DTMs, DSMs and CHMs and to derive LiDAR metrics from normalized LiDAR points over the entire area of the National Park. Performing these LiDAR processing steps over the extend of the entire National Park is computationally intense. The full version of LAStools is needed to assure timely processing of the vast amount of raw data. The results of this study will be used as a benchmark to compare with those previously achieved by Latifi et al. (2015) using leaf-on data across the same study area. The hypothesis is that using leaf-off LiDAR data together with complementary modeling approaches (e.g. beta regression and machine learning) will lead to improved results.
+ Two LiDAR data sets covering the entire area of Bavarian Forest National Park (24369 hectare = 243.69 square kilometers).
+ Leaf-off LiDAR from 2009 / 2010 flight campaign split in first- and last return data from the “Bayrisches Landesvermessungsamt”, the state surveying office of Bavaria. Average point density is 4 – 5 points/m² and points are classified in 5 categories: 1 = certain ground point, 2 = uncertain ground point, 3 = no ground point (object point), 4 = point on building, 9 = invalid point.
+ Leaf-on LiDAR from 2012 flight campaign recorded in full waveform and processed into high-density point cloud. The statistical metrics are already available for this dataset.
LAStools processing (leaf-off data only):
1) set classifications to 0 (= unclassified) and merge first and last return files [las2las]
2) tile data into 1000 x 1000 m² tiles with 25 m buffer to avoid edge artifacts [lastile]
3) extract ground points on many cores in parallel [lasground]
4).generate DTM from ground points on many cores in parallel [las2dem]
5) height-normalize tiles on many cores in parallel [lasheight]
6) derive metrics (percentiles, proportions and possibly density metrics) from height-normalized tiles on many cores in parallel. also measure pre-defined height strata to characterize the forest vertical layers as measured in the field campaign including 0-2 m (herbal layer), 2-5 m (shrub- and regeneration layer), 5-12 m (lower tree layers) and > 12 m (top tree layer) [lascanopy]
7) create a Canopy Height Model (CHM) using the pit-free method of Khosravipour et al. (2014) with the workflow described here [lasthin, las2dem, lasgrid]
Khosravipour, A., Skidmore, A.K., Isenburg, M., Wang, T.J., Hussin, Y.A., 2014. Generating pit-free Canopy Height Models from Airborne LiDAR. PE&RS = Photogrammetric Engineering and Remote Sensing 80, 863-872.
Latifi, H., Heurich, M., Hartig, F., Müller, J., Krzystek, P., Jehl, H., Dech, S., 2015, Estimating over- and understorey canopy density of temperate mixed stands by airborne LiDAR data. Forestry (Article in Press). DOI. 10.1093/forestry/cpv032