Predictive Evapotranspiration Equations in Rain Gardens

Abstract

Current stormwater control measure (SCM) design often does not include the dynamic process of evapotranspiration (ET) for vegetated systems. This study compared two reference ET equations with a three-year data set from rain garden weighing lysimeters. The outcome was a tool to incorporate ET into SCM design. The weighing lysimeters at Villanova University, located in southeastern Pennsylvania, were used to measure water budget parameters for three scenarios: sandy loam with UO, sand with an unconstricted outflow (UO), and sand with internal water storage (IWS). The two ET models explored were the ASCE Penman-Monteith equation (a robust model) and the Hargreaves equation (a simple model). Estimated ET values from these two equations, both with and without modifications for water availability and crop presence, were compared and calibrated (if modified) with observed data. Comparisons and calibrations were performed on a daily and storm basis to explore the applicability of the two ET models for continuous and storm approaches. The observed ET was 28%–52% of inflow over the continuous three-year period and 16–30 mm on a storm scale, making ET a significant part of the lysimeters’ water budget. Due to the experimental nature of the lysimeters, 12 of the 36 study months had additional simulated runoff, such that a smaller ET as a percentage of inflow was expected in the rain garden SCM’s water balance. The Hargreaves and ASCE Penman-Monteith equations without modification provided an adequate estimate for rain garden ET for all systems at the storm scale. Modifications to ET estimations produced by both equations through crop coefficients and a soil moisture extraction function provided a good model for storm-scale ET by reducing errors and increasing efficiencies for all weighing lysimeter types. Evapotranspiration estimates from both unmodified equations provided, at best, a marginally better estimate than the average observed rate for continuous daily rain garden ET. The application of crop coefficients and a soil moisture extraction function to both equations reduced errors in ET estimates and increased the equations’ predictive power (Nash-Sutcliffe efficiency) for all weighing lysimeter types. Both equations with modifications on a daily scale produced good ET estimates for the IWS system. For both equations, crop coefficients were found in an expected range for UO systems (0.3–1.5) but were high in the IWS system (1.6–2.0). Soil moisture extraction functions were not needed to calibrate the IWS equations on the storm scale. Both the Hargreaves equation and the ASCE Penman-Monteith equation provided an adequate model (especially with modifications) to incorporate ET into a design-storm approach to SCM design. Use of both predictive models on a daily scale has potential use in continuous simulation, as in most cases the ET estimations predicted by the equations provided a better estimate than the average of the observed daily ET rates.