Numerical assessment of effective evapotranspiration from maize plots to estimate groundwater recharge in lowlands

Abstract

To maximize the irrigation efficiency and to protect groundwater from agrochemical pollution, two variables must be known with good accuracy: effective evapotranspiration and infiltration, especially in lowland areas were the run-off is minimal. Three different experimental plots cultivated with maize were equipped with tensiometers and soil moisture probes to monitor every day the water movement in the unsaturated zone. Other relevant parameters of the various soil layers, as hydraulic conductivity and water retention curve, were obtained in laboratory experiments, while boundary conditions, as precipitations, temperature and root growth, were obtained on site. Inverse modeling was performed using HYDRUS-1D to assess the degree of uncertainty on model parameters. Results showed a good model fit of water content and head pressure at various depths, in each site, using Penman–Monteith formula for daily potential evapotranspiration calculation, but poor fit applying the Hargreves and Turk formulas. Best performance of model fit was observed for S-shaped equation employed to simulate the root water-uptake reduction with respect to Feddes equation. The soil parameters uncertainty was limited and remained within analytical errors, thus a robust estimation of cumulative infiltration and evapotranspiration has been derived. This study points out that evapotranspiration is the most important variable in defining groundwater recharge for maize crops in lowlands.

Introduction

Aquifer recharge assessment is essential to quantify water resources and to estimate contaminants flux towards aquifers (Scanlon et al., 2002). Presently, the most effective tool to quantify recharge flux is to model the unsaturated soil water dynamics; although this process faces many challenges in field conditions (Youngs, 1995). The most important are the soil heterogeneity, the development of aggregations/cracks and pore discontinuity, which causes deviations from Darcy's Law. Despite of these limitations, numerical models are more and more frequently employed to quantify soil water dynamics, often applying several simplifying assumptions to cope with these limitations. Deterministic numerical models require hydraulic conductivity and characteristic curve functions for each soil layer (Jarvis, 1994, Jacques et al., 2002, Suleiman and Ritchie, 2003, Šimunek et al., 2008): in field studies, these parameters are routinely calculated from soil physical properties, because their quantification is often difficult, time consuming and expensive (Suleiman and Ritchie, 2001). Nevertheless, a multiplicity of direct measurement methods have been developed to estimate hydraulic input parameters (Šimunek et al., 1998a, Dane and Topp, 2002, Peters and Durner, 2008). In this study an evaporative method was used to assess hydraulic parameters. Once these parameters have been characterized, other variables had to be addressed such as evapotranspiration, root growth and stress functions.

Several techniques have been developed to measure evapotranspiration: by weighting lysimeters (Gavilan et al., 2007), by field water balance equation (Lenka et al., 2009) or micrometeorological methods (Drexler et al., 2005), but these techniques are threatened to be time consuming and expensive. Evapotranspiration can also be estimated from climatic data, linking evapotranspiration with one or more climatic variables (Sharma, 1985). The performance of various equations has been assessed under a variety of climates (Allen et al., 1998, Berengena and Gavilàn, 2005). The Penman–Monteith equation is now considered the best approximation to estimate evapotranspiration over most of climates (Jensen et al., 1990). This equation is now become the United Nations Food and Agriculture Organization (FAO) standard equation to estimate evapotranspiration (Allen et al., 1998). In addition to evapotranspiration, studies on water stress on plants and roots have been performed at various scales, from microscopic to field sites and ecosystems (Passioura, 2002, Duchemin et al., 2006). These studies reveal that maize exerts a very high water demand mainly in the upper soil horizons (Lenka et al., 2009, Doorenbos and Pruitt, 1977) and that an important variable is the plants stress function due to water deficit and salinity stress (Maas, 1986, Katerji et al., 2000). There is a close connection between groundwater depth and crop evapotranspiration, in fact plants can use shallow groundwater, if present during the growing season, which can improve or threaten crop performance as recently demonstrated (Nosetto et al., 2009).

To simulate the abovementioned processes, numerical models have often been implemented and employed in the recent past (Azzaroli Bleken et al., 2009, Lopez-Cedron et al., 2005). Overall, the model accuracy depends on the accuracy of the input data and on the assumptions made to simplify the complex modeled environment (Panigrahi and Panda, 2003). Here residuals, i.e. differences between simulated and measured values, were used to evaluate the accuracy of each model run. The most common statistical tools, used to compute model accuracy are the average mean residual error and the root-mean square residual error (Whitmore, 1991). These statistics were used in this work to assess the effect of different potential evapotranspiration (PET) formulas and root water-uptake reduction functions on actual groundwater recharge in the Ferrara province, an area located in the Po Plain lowlands, characterized by a flat topography and by a temperate humid climate. The main goal of this study were to assess whether simple approaches to calculate the PET, like Hargreves and Turk ones, can substitute complex ones like Penman–Monteith and to assess the variability of the groundwater recharge estimated with different PET formulas. The same assessment was performed for root water-uptake reduction functions. In addition, simulations were run with minimum and maximum observed saturated hydraulic conductivities, to quantify its influence on groundwater recharge flux. To achieve these aims, three field sites consisting of different soils representative of the Po Plain lowlands were monitored for 400 days and modeled with HYDRUS-1D.