A geospatial approach for estimating hydrological connectivity of impervious surfaces

Abstract

Recent studies have reported that connected impervious areas – those impervious surfaces that contribute directly to runoff in a storm network or stream – are a better indicator of hydrologic response, stream alteration, and water quality than total impervious area. However, most methods for quantifying connected impervious areas require major assumptions regarding the definition of ‘connection’, potentially over-simplifying the role of variable climates, slope gradients, soils conditions, and heterogeneous flow paths on impervious surface connectivity. This study presents a new conceptual model and method for estimating hydrologically connected impervious areas (HCIA) that explicitly considers the effect of landscape and storm variability. The model separates impervious surfaces into two categories: directly or physically connected (Aphys) and variably connected (Avar) (impervious that drains to pervious). Of these categories, we investigated the sensitivity of Avar connectivity to varying soil conditions, slope gradients, rainfall properties, and hillslope geometry using PySWMM (a python interface for SWMM5). Simulations spanned a large parameter space with varying soil, slope, rainfall properties and geometries (i.e., relationships between the impervious and downslope pervious areas). PySWMM simulations were used to train and test a regression tree that predicts infiltration and connectivity of runoff from Avar surfaces, which provides excellent fidelity with PySWMM outcomes. To enable use of these methods in practice, we developed an ArcGIS tool that (1) delineates subcatchments; (2) extracts the impervious surface categories Aphys and Avar; (3) applies the regression tree algorithm to predict the fraction of incident rainfall that produces runoff across Avar; and (4) summarizes the resulting HCIA by subcatchment. Analysis of the regression feature importance shows that, in general, Avar connectivity is highly sensitive to the soil type, rainfall depth, area fraction, and antecedent soil moisture conditions of the downslope pervious area. We find that temporally varying parameters (e.g., rainfall and antecedent soil moisture) control Avar connectivity in areas with low permeability soils, while spatial flow path variability (e.g., relative quantity of disconnecting pervious area) controls Avar connectivity in areas with highly permeable soils. The methods developed in this study can be used to identify impervious surface connectivity more accurately in urban watersheds, representing an important step forward for incorporating spatial heterogeneity in stormwater modeling and planning.