Predictors and variance parameters for the selected unit level models.

When the error variance was not constant, mcp was the first predictor, associated to β1. Elev_mean and Elev_sd are the mean and standard deviation of the lidar heights. Elev_mean_sq is the square of the LiDAR height. Elev_P50, Elev_P70 and Elev_P80 are the 50th, 70th and 80th percentile of the LiDAR heights. all_co_ab_2m and 1st_co_ab_mean are the proportion of returns above 2m and the proportion of first returns above the mean LiDAR height respectively. See S1 Appendix for additional details on the variables and model selection procedure.