A remote sensing-guided forest inventory concept using multispectral 3D and height information from ZiYuan-3 satellite data

Kurzfassung

Increased frequencies of storms and droughts due to climate change are changing central European forestsmore rapidly than in previous decades. To monitor these changes, multispectral 3D remote sensing (RS) data canprovide relevant information for forest management and inventory. In this case study, data of the multispectral3D-capable satellite system ZiYuan-3 (ZY-3) were used in a RS-guided forest inventory concept to reduce the fieldsample size compared to the standard grid inventory. We first pre-stratified the forest area via the ZY-3 datasetinto coniferous, broadleaved and mixed forest types using object-based image analysis. Each forest type wasthen split into three height strata using the ZY-3 stereo module-derived digital canopy height model (CHM).Due to limited sample sizes, we reduced the nine to six strata. Then, for each of the six strata, we randomlyselected representative segments for inventory plot placement. We then conducted field inventories in theseplots. The collected field data were used to calculate forest attributes, such as tree species composition, timbervolume and canopy height at plot level (terrestrially measured tree height and height information from ZY-3CHM).Subsequently,wecomparedtheresultingforestattributesfromtheRS-guidedinventorywiththereferencedata from a grid inventory based only on field plots. The difference in mean timber volumes to the reference was+30.21 m3ha−1(8.99 per cent) for the RS-guided inventory with terrestrial height and−11.32 m3ha−1(−3.37per cent) with height information from ZY-3 data. The relative efficiency (RE) indicator was used to comparethe different sampling schemes. The RE as compared to a random reduction of the sample size was 1.22 forthe RS-guided inventory with terrestrial height measurements and 1.85 with height information from ZY-3 data.The results show that the presented workflow based on 3D ZY-3 data is suitable to support forest inventories byreducing the sample size and hence potentially increase the inventory frequency.