ObjectiveFiltering is an important part of data preprocessing when using discrete-return LiDAR to derive leaf area index (LAI). Laser penetration index (LPI), which responses to the canopy’s gap fraction, is a pivotal argument, and can be defined by echoes intensity or count, and is directly influenced by filter precision. So, filter algorithms can affect deriving LAI indirectly.
MethodIn this paper, we used the open source filter algorithms without manual operation to filter the error points. Using the LPI defined on count, we built model in larch forest and elm forest, Maor Mountain National Park, based on Beer-Lambert law. We compared the filter algorithm of adaptive triangulated irregular network, morphology, local slope, using hybrid filtering as standard. In order to avoid the subjective influence during modelling, we built 100 models by choosing samples randomly.
ResultIn larch forest, the models’ R-squared under larch was 0.900 3, 0.876 3, 0.892 5,0.877 0, root mean squared error (RMSE) was 0.105 6, 0.134 5, 0.109 7,0.133 2; in elm forest, the models’ R-squared was 0.914 4, 0.903 0, 0.887 2, 0.900 0, root mean squared error (RMSE) was 0.269 0, 0.201 7, 0.189 4, 0.207 0, respectively.
ConclusionConsidering the sample’s topography, when using discrete-return LiDAR data derive LAI based on LPI, the hybrid algorithm has a better performance on deriving LAI. II error has more influence on deriving LAI than I error.
DownLoad: