Agronomy Journal  Volume 105, Issue 2  2013 419 Biometry, Modeling & Statistics Validating the FAO AquaCrop Model for Rainfed Maize in Pennsylvania Valerie J. Mebane,* Rick L. Day, James M. Hamlett, Jack E. Watson, and Greg W. Roth Published in Agron. J. 105:419–427 (2013) doi:10.2134/agronj2012.0337 Copyright © 2013 by the American Society of Agronomy, 5585 Guilford Road, Madison, WI 53711. All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. D rought is the primary contributor to crop failure in the United States (Kaltenbacher, 1994), resulting in estimated annual losses between US$6 and 8 billion (Federal Emergency Management Agency, 1995). Although drought is typically associated with the western United States, it is prevalent in the east as well. Much of the eastern United States, including Pennsylvania, experienced an extremely severe drought in 1998 and 1999 (Wilhite et al., 2005). Te USDA Farm Service Agency estimated that the 1999 drought alone resulted in US$500 million in associated crop losses in Pennsylvania (Susquehanna River Basin Commission, 2001), and in some counties crop losses reached unprecedented levels of 70 to 100% (Susquehanna River Basin Commission, 2001). Agricultural drought occurs when there is inadequate soil moisture to meet a plant’s needs, ultimately resulting in a reduc- tion in crop yield. A plant’s water demand is dependent on multiple factors, including the biological characteristics of the specifc plant, its stage of growth, prevailing weather conditions, and the soil characteristics (Sastri, 1993; Wilhite, 1993). Many widely used drought indices and assessment tools fail to provide an accurate assessment of soil variability across the landscape and neglect the interaction of crops and soil type on soil moisture reserves. Process-oriented crop models have the ability to capture this critical relationship and can thus be utilized as valuable tools in the study and modeling of agricultural drought. Te applicability and degree of complexity of the numerous crop models in existence today are highly variable. Te more physically based mechanistic crop models are extremely data intensive and require accurate input parameters that are ofen unattainable for analyses without measured feld data. Tere- fore, simpler soil–plant–atmosphere models that can sufciently simulate the efects of water stress on various plant growth and development processes can provide a good alternative to data- intensive and complex simulation models (Ines et al., 2001). An agricultural drought vulnerability study for Pennsylvania soils was developed in which the AquaCrop (Raes et al., 2009b; Steduto et al., 2009) crop growth simulation model was utilized to simulate the efects of soil moisture stress on corn produc- tion under a wide range of soil and climate conditions across a span of 30 yr (Mebane, 2011). Te efects of soil moisture stress on corn production were quantifed to develop an agricultural drought vulnerability index with the ability to capture the vul- nerability of Pennsylvania’s diverse soils to agricultural drought under a variety of climatic conditions. Te vulnerability index was developed to be utilized as a long-term risk assessment tool applicable at the feld scale to help Pennsylvania farmers make educated decisions based on their farm’s potential vulnerability to agricultural drought. Te index was based on local soil condi- tions, local long-term historical climatic conditions, and the ABSTRACT It is widely known that a close relationship exists between crop production and water stress. In this study, feld-measured data were used to test the performance of AquaCrop and its ability to capture this relationship for rainfed maize (Zea mays L.) in Pennsyl- vania. Te objectives were to evaluate AquaCrop’s ability to simulate the progression of cumulative biomass and grain yield with time, fnal biomass and harvestable yield, and volumetric water content at six depths. Two years of data from a study conducted in Rock Springs, PA, were used to validate AquaCrop’s ability to accurately simulate the progression of cumulative biomass and grain yield with time, as well as fnal biomass and harvestable yield. Data collected from January 2004 to March 2006 from a study conducted near Landisville, PA, were used to assess AquaCrop’s ability to efectively simulate soil moisture content at six depths. In addition, the 2004 and 2005 seasonal fnal biomass measurements obtained from the Landisville location were compared with the model’s simulated values. Te results indicated that AquaCrop was able to accurately simulate the progression of cumulative biomass and grain yield with time, with index of agreement values ranging from 0.96 to 0.99. Comparisons between simulated and measured fnal biomass and fnal harvestable yield produced biomass deviations ranging from 2.4 to 20.7% and yield deviations of 2.9 and 15.3%. Te water balance evaluation indicated that, averaged across all depths, the results were consistent with other validation studies of soil water balance models, with RMSE ranging from 1.5 to 9.8% (v/v). V.J. Mebane, R.L. Day, and J.E. Watson, Dep. of Ecosystem Science and Management, Pennsylvania State Univ., University Park, PA 16802; J.M. Hamlett, Dep. of Agricultural and Biological Engineering, Pennsylvania State Univ., University Park, PA 16802; and G.W. Roth, Dep. of Plant Science, Pennsylvania State Univ., University Park, PA 16802. Received 30 Aug. 2012. *Corresponding author (vjp112@psu.edu). Abbreviations: ET 0 , reference evapotranspiration; GDD, growing degree days; MAE, mean absolute error; MBE, mean bias error, RMSE, root mean squared error; RMSE s , root mean squared error systematic; RMSE u , root mean squared error unsystematic. Published January 29, 2013