The LULC maps and simplified land zoning maps were transmitted to the fishnet sampling procedure to prepare the dataset for the probability-fitting study. The workflow is shown in Figure 3 and the governing equations for computing samples' impervious fractions are presented in Equations (3) and (5). The amount of fishnet samples is proportional to the area of different land zoning categories in the study catchment, which aligns with the theory of stratified random sampling for avoiding the ‘bias’ caused by the larger area land zoning category (e.g. low- to medium-density residential) (Vries 1986). Table 3 illustrates the number of raw and downscaled fishnet samples (after applying the area reduction ratio).