R lidR package grid_terrain() giving weird plot

Question

I am working on the pre and post lidar earthquake data from Kumamoto, Japan. I would like to generate a DTM with the grid_terrain(pre_online, res = 1, algorithm = knnidw(k = 10L, p = 0.5)) function in R Studio. When I plot the data with plot_dtm3d() I get a weird looking one. Is someone familiar with this problem?

The data is available at https://portal.opentopography.org/lidarDataset?jobId=pc1620409032095.

The plot looks like this and it should be a 3D terrain model.

pre_online
#> class : LAS (v1.2 format 1) 
#> memory : 633 Mb 
#> extent : -12000, -9000.001, -21500, -19500 (xmin, xmax, ymin, ymax) 
#> coord. ref. : JGD2000 / Japan Plane Rectangular CS II (with axis order normalized for visualization) 
#> area : 6 km² 
#> points : 20.74 million 
#> points density : 3.46 points/m²

Answer 1

Your point cloud contains a lot of region with 0 ground points.

library(lidR)
las = readLAS("~/Téléchargements/points.laz", filter = "-keep_class 2")
plot(las)

Moreover if I look at ground points they do not look really "on the ground". I don't know how the classification have been made but it looks poor including a lot of points that are unlikely to be ground points

sub = clip_circle(las, -10000, -20000, 200)
plot(sub, axis = TRUE)

This may explain why grid_terrain failed and found infinite values. You can replace infinites by NA

dtm = grid_terrain(las, 2, tin())
dtm
#> class      : RasterLayer 
#> dimensions : 1000, 1500, 1500000  (nrow, ncol, ncell)
#> resolution : 2, 2  (x, y)
#> extent     : -12000, -9000, -21500, -19500  (xmin, xmax, ymin, ymax)
#> crs        : +proj=tmerc +lat_0=33 +lon_0=131 +k=0.9999 +x_0=0 +y_0=0 +ellps=GRS80 +units=m +no_defs 
#> source     : memory
#> names      : Z 
#> values     : -2147484, 392.234  (min, max)
dtm[dtm < 0] = NA

Yet, your DTM is terrible and looks more like a CHM...

plot_dtm3d(dtm)