Forest Change Type Detection Based On Bi-Temporal Airborne Laser Scanning Data

Forests are an important component of global sustainable development and play an important role in protecting biodiversity,cleaning up the environment and regulating climate change.Timely understanding of forest cover and its structural changes is important for the effective use of forest resources and sustainable forest management.Optical remote sensing imagery has been shown to be highly accurate in detecting forest cover type changes,but it has low resolution,tends to saturate when analysing vegetation and does not allow for detecting of forest changes from a vertical perspective.LiDAR(Light Detection and Ranging)has been successfully demonstrated as a means of deriving structural properties of forests at the plot level.It is well suited to provide information on changes in forest structure because it can emit pulsed signals that penetrate the dense canopy and reproduce stand structure.As the availability of multitemporal ALS(Airborne Laser Scanning)data increases,more and more studies are demonstrating that bi-temporal ALS data can be used to estimate forest height growth and above-ground biomass change.However,current ALS-based forest dynamics detecting still has some problems,such as little concern for coordination between different temporal ALS data,neglecting the accumulation and transmission of uncertainty during processing,and the quantification of changes in existing structural attributes that do not meet the needs of forestry applications.This paper therefore takes the low hills of northeastern Song County,Henan Province,as the study area,and uses two phases of ALS and field survey data from 2020 and 2021 as data sources to investigate methods for forest change detection and to explore the ability and effectiveness of ALS in identifying forest changes caused by different management activities.This paper firstly establishes a multi-temporal optimal digital elevation model(DEM)generation technique,and compares the accuracy of DEMs with different spatial resolutions generated by different spatial interpolation algorithms and single/dual phase ALS data sources;secondly constructs a Canopy Height Model(CHM)difference Secondly,an uncertainty model was constructed to quantify the error propagation of canopy height model(CHM)variation;finally,the effects of forest change detection based on CHM and ALS structural indicators were compared based on field survey data and forest management.The main findings of this paper are as follows:(1)The optimal DEM is a DEM with a spatial resolution of 0.5 m generated by the irregular triangular mesh interpolation method and ground points from dual-temporal LiDAR data;the error of the DEM decreases with increasing resolution;the DEM generated using ground points from dual-temporal ALS data outperforms the DEM generated using ground points from either single-temporal ALS data;the effect of the interpolation method on the error of the DEM is the same as the combined effect of spatial resolution and ground point source.The effect of the interpolation method on the DEM error is a combination of spatial resolution and ground point source,with the irregular triangular mesh interpolation method outperforming kriging and inverse distance weighting when the spatial resolution is 0.5 m and the ground points are from bi-temporal ALS data.(2)A method based on the error propagation law of statistical medium error can effectively quantify the uncertainty in the amount of CHM variation over time and space.The medium error is positively correlated with the degree of canopy height variation,and is influenced by exogenous factors such as topography and data quality,as well as the stand’s own state such as tree height and densities.(3)For the four types of forest changes caused by management activities:replanting,harvesting,fast growth and slow growth,the multi-temporal ALS data are well identified and the detection based on ALS structural indicators is significantly better than the CHM-based results.The overall accuracy of the CHM-based model for detecting forest change was 47.0%,with a kappa coefficient value of 0.26,and could only detect the harvesting type.The overall accuracy of the model for detecting forest change based on ALS structural indicators was 89.7%,with a Kappa coefficient value of 0.86,and it was able to identify all types of forest change well.Among the ALS structural indicators,point cloud height standard deviation,skewness and kurtosis,as well as 90% and 95% quantile height played an important role in the detection of forest change types.