Use of CERES-Maize to study effect of spatial precipitation variability on yield
Introduction
Corn [Zea mays (L.)] yield varies significantly from year to year on a site-specific basis because of physical and management factors. Yield maps based on GPS (global positioning system)-guided yield monitors show spatial variability, but using these maps to manage agronomic inputs is challenging because of uncontrollable factors affecting yield. Precipitation is one such factor.
The total effect of precipitation variability on yield may be divided into two components: spatial variation from point to point and temporal variation between crop years. The effect of temporal (year-to-year) variability of precipitation on yield variability is well known, but the effect of spatial precipitation variability is not well understood. Quantification of spatial precipitation variability is needed to determine whether precipitation data are needed on the farm or can be obtained from off-farm climate stations, for uses such as yield map interpretation and crop simulation.
Many farmers have begun to measure site-specific precipitation on their farms (Reetz, 1999). Precision farmers have been encouraged in the popular farm press to purchase their own weather monitoring equipment (Bechman, 1998), but National Weather Service (NWS) data from nearby stations are free or inexpensive and can be used to estimate on-farm precipitation. However, if spatial variability at the scale of NWS stations is significant, then relying on NWS data may not accurately estimate precipitation at the farm.
In Indiana, the spacing of NWS stations averages 17.7 km. Precipitation data at a point between NWS stations can be estimated with the distance-weighted mean (Serrano, 1997):where Pm=precipitation for the central unknown station, Pi=known precipitation for each of the surrounding stations, and Di=the distance from the central station to each of the surrounding stations. Studies of precipitation in non-mountainous, agricultural regions have suggested that precipitation variability at scales of 2–5 km is not significant, but beyond that scale, precipitation may vary substantially between rain gauges (Huff, 1979, McConkey et al., 1990, Hubbard, 1994). Spatial precipitation variability at this small scale has not usually been tied to crop yield variability.
Growth of corn plants depends on precipitation throughout the season, with periods of varying importance. Corn water needs peak in the 4 weeks prior to silking, ending about 20 days after silk (Dale and Daniels, 1992). Because of changing crop needs for precipitation, some researchers have aggregated precipitation according to corn phenological phases for studies of yield response to precipitation (e.g., Dai, 1991, Dale and Daniels, 1995).
Sophisticated relationships between precipitation and yield have been incorporated into the CERES(Crop-Environment REsource Synthesis)-Maize simulation model (Jones and Kiniry, 1986). CERES-Maize is a module of DSSAT (Decision Support System for Agrotechnology Transfer), version 3.5 (Hoogenboom et al., 1999). CERES-Maize was developed by the US Department of Agriculture (USDA), with extensive testing in the Corn Belt (Hodges et al., 1987). Successful adaptations of the model to areas such as Brazil (Liu et al., 1989), China (Wu et al., 1989), and Nigeria (Jagtap et al., 1999) have proven the model to be accurate under very different conditions, indicative of having modeled true physical relationships.
CERES-Maize models the carbon, water, and nitrogen balances of the corn plant over the growing season, split into nine phenological phases (crop stages). Infiltration comes from the difference between precipitation and runoff, which is calculated with the USDA Soil Conservation Service (SCS) curve number method. Infiltrated water is apportioned with a cascading principle through up to 10 soil layers. The Priestley–Taylor method is used to calculate potential evapotranspiration, and a two-stage Ritchie model is used to estimate actual evapotranspiration (Xevi et al., 1996). After evapotranspiration occurs, capillary water flows upward to depleted layers according to a function of moisture content and lower limit of moisture (Gabrielle et al., 1995).
Sadler et al. (2000) found that manipulating infiltration was the most direct link to crop water stress in CERES-Maize, and infiltration was overestimated for an extreme storm. However, for seven years tested, yield generally increased with precipitation, except for two cases when yield was reasonably limited by drought or saturation. They found the sensitivity of the model to water supply via precipitation to be “plausible” and “promising”, which suggests it is appropriate for modeling yield-precipitation relationships.
To determine significance of farm-scale precipitation variability, O'Neal et al. (2001) developed an analysis of precipitation variability by phenological phase, comparing spatial to temporal variability within the farm. This study extends the work to the scale of NWS stations. The effect of precipitation variability on crop yield is also analyzed.
The objective was to determine the usefulness of on-farm precipitation measurement compared to off-farm estimates from surrounding NWS stations, through (1) determining spatial variability of precipitation at the scale of nearest NWS stations, and comparing with temporal variability; and (2) determining the effect of that variability on corn yield. Yield was simulated with CERES-Maize using DSSAT v. 3.5.
Section snippets
Procedure
The procedure consisted of two parts: determining spatial and temporal precipitation variability, and determining the effect of that variability on corn yield. Spatial precipitation variability was determined from observations of precipitation among three data sources (discussed below). Temporal precipitation variability was determined from 31 years of precipitation at an on-farm NWS station. Yield was simulated with CERES-Maize based on precipitation from the three data sources. Yield
Spatial and temporal precipitation variability
Precipitation aggregated by phenological phase is summarized in Table 2. The median percent absolute difference of precipitation among data sources, representing spatial variability of precipitation, varied from 21 to 104% from phase to phase. The two periods around silking (V3–R1), most critical for corn moisture, had differences of 21 and 49%, suggesting precipitation values could vary by almost one-half depending on the location. This spatial variability may arise from the spatially
Conclusions
This study determined spatial and temporal precipitation variability, and the possible effect of that variability on corn yield, among three different precipitation data sources for a farm in east central Indiana: on-farm data, the nearest non-urban National Weather Service station with electronic reporting (27 km from the farm), and a weighted mean of the three nearest such stations (27–35 km away).
Spatial variability of phenological phase accumulations of precipitation from the three sources
Acknowledgements
The authors thank the following: the staff at Davis-Purdue Agricultural Center, for collecting, organizing, and sharing weather and crop data; R. Ogoshi, G. Hoogenboom, G. Tsuji, and M. Habeck, for their indispensable assistance with providing and using DSSAT; D.K. Morris, for processing and providing yield and soil data; K. Scheeringa, for NWS information; I. Aly, for GIS work; and J. Boyer, S. Brouder, B. Engel, R. Grant, J. Lowenberg-DeBoer, M. Morgan, S. Parsons, and E.J. Sadler, for
References (43)
- et al.
Large-scale simulation of wheat yields in a semi-arid environment using a crop-growth model
Agricultural Systems
(1999) Regression of a model on real-system output: an invalid test of model validity
Agricultural Systems
(1990)- et al.
Using the CERES-Maize model to estimate production for the U.S. Corn Belt
Agricultural and Forest Meteorology
(1987) Spatial variability of daily weather variables in the high plains of the USA
Agricultural and Forest Meteorology
(1994)- et al.
Long-term assessment of nitrogen and variety technologies on attainable maize yields in Nigeria using CERES-Maize
Agricultural Systems
(1999) - et al.
Testing simulation models for the assessment of crop production and nitrate leaching in Hungary
Agricultural Systems
(1995) - et al.
Application of CERES-Maize model to yield prediction of a Brazilian maize hybrid
Agricultural and Forest Meteorology
(1989) - et al.
Small area variability of warm-season precipitation in a semiarid climate
Agricultural and Forest Meteorology
(1990) Misuse of regression for empirical validation of models
Agricultural Systems
(1997)- et al.
Comparing genetic coefficient estimation methods using the CERES-Maize model
Agricultural Systems
(2000)
Site-specific modeling of corn yield in the SE coastal plain
Agricultural Systems
On the application of the CERES-Maize model to the North China Plain
Agricultural and Forest Meteorology
Comparative study of two crop yield simulation models
Agricultural Water Management
Capturing precision farming's missing link: weather is the missing variable in the yeilds-soils-environment equation
Indiana Praire Farmer
Application of simulation models and weather generators to optimize farm management strategies
Concepts for calibrating crop growth models
Simulation of within variablility of corn yield with CERES-Maize model
The agroclimatology of a moisture stress index and average corn (Zea mays L.) yields in Tippecanoe County, Indiana 1961–1991
Proceedings of the Indiana Academy of Sciences
A weather-soil variable for estimating soil moisture stress and corn yield probabilities
Agronomy Journal
Cited by (25)
Can crop simulation models be used to predict local to regional maize yields and total production in the U.S. Corn Belt?
2016, Field Crops ResearchCitation Excerpt :Third, it is challenging to define the inference domain for a crop model simulation based on weather, soil, and crop management data for a specific location and how to upscale this ‘point’ estimate to larger spatial levels (Van Wart et al., 2013a; van Bussel et al., 2015). Finally, simulation models should be able to reproduce yields across a wide range of environmental and management conditions without need of site-year specific calibration of internal parameters related to leaf area expansion, carbon assimilation and partitioning, and resource efficiency, though such calibration is common in the literature (e.g., Hodges et al., 1987; ÓNeal et al., 2002; Rosenzweig et al., 2014). Examples of use of crop simulation models for estimating yields for specific site-years can be found in the literature (e.g., Bannayan and Crout, 1999; Baigorria et al., 2008; Quiring and Legates, 2008).
A Comprehensive Review of the CERES-Wheat, -Maize and -Rice Models' Performances
2016, Advances in AgronomyCitation Excerpt :A study using 50 years of experimental data reported that the grain yield simulations for maize in fertilized (side-dress nitrogen of 112 kg/ha) versus nonfertilized fields showed a much higher normalized RMSD in the nonfertilized field (82%) than in the fertilized field (39%) (Liu et al., 2010). O’Neal et al. (2002) pointed out that the low correlation (R2 = 0.33) between the observed and simulated grain yields given 127 ~ 227 kg N/ha application treatments was due to lower levels of fertilizer application in the treatments used. The normalized RMSDs found in the maize grain yield simulation across 0 ~ 400 kg/ha nitrogen application treatments in China were small as well (within 15%) (Liu et al., 2012).
Identifying irrigation and nitrogen best management practices for aerobic rice-maize cropping system for semi-arid tropics using CERES-rice and maize models
2015, Agricultural Water ManagementCitation Excerpt :CERES (Crop Estimation through Resource and Environment Synthesis)-Rice and -Maize are process-based models embedded in DSSAT simulate the main processes of crop growth and development such as phenological development, canopy leaf area growth, dry matter accumulation and grain yield. The CERES-Rice and -Maize models were evaluated by many researchers across locations (Sarkar and Kar, 2006; Timsina and Humphreys, 2006; O’Neal et al., 2002; Behera and Panda, 2009; Liu et al., 2011; He et al., 2012; Salmerón et al., 2012; Jeong et al., 2014; Ngwira et al., 2014) with good agreements between predicted and observed values. Even though simulation results generally will have some uncertainties associated with inputs and model parameters, but still the simulation models can be effectively utilized as a scientific tool to increase the resource use efficiency of cropping systems (Timsina and Connor, 2001; Sarkar and Kar, 2008; Timsina and Humphreys, 2006; Timsina et al., 2008).
Effects of El Niño Southern Oscillation on the space-time variability of Agricultural Reference Index for Drought in midlatitudes
2013, Agricultural and Forest MeteorologyCitation Excerpt :Secondly, they compared ARID to the water stress index for photosynthesis (WSPD) calculated by the Decision Support Systems for Agrotechnology Transfer (DSSAT) (Hoogenboom et al., 2004; Jones et al., 2003) CERES-Maize crop model (Jones and Kiniry, 1986) at 16 locations in Alabama, Florida, and Georgia. The WSPD was chosen because the DSSAT CERES-Maize has been extensively tested and used (Fraisse et al., 2001; Hodges et al., 1987; López-Cedrón et al., 2005; O’Neal et al., 2002). Also, for each of the 16 locations, seven other commonly used drought indices were compared with WSPD.
Precipitation variability in Northeast China from 1961 to 2008
2011, Journal of HydrologyCitation Excerpt :Precipitation is one of the most important variables in diagnosing climate change as well as revealing the eco-environmental response to climate change on a regional scale (Cannarozzo et al., 2006). Both temporal and spatial patterns of precipitation and evapotranspiration influence eco-hydrological processes (Oguntunde et al., 2006; Cannarozzo et al., 2006; McVicar et al., 2007) and, consequently, regional economies strongly tied to agriculture and raising livestock (O’Neal et al., 2002; Wei et al., 2005; Guo et al., 2006). Therefore, a better understanding of precipitation variability on a regional scale will assist in determining water management policies.
Parameterizing soil and weather inputs for crop simulation models using the VEMAP database
2010, Agriculture, Ecosystems and Environment