Evaluating the Fidelity of Downscaled Climate Data on Simulated Wheat and Maize Production

Evaluating the fidelity of downscaled climate data on simulated wheat and maize production in the southeastern US

DavideCammaranoa, Lydia Stefanovab, Brenda V. Ortizc, Melissa Ramirez-Rodriguesa, SentholdAssenga, VasubandhuMisrab,d, e, Gail Wilkersonf, Bruno Bassog, James W. Jonese, Kenneth J. Booteaand Steven DiNapolib

a Department of Agricultural & Biological Engineering, University of Florida, Gainesville, FL, 32611, U.S.A.

b Center for Ocean-Atmospheric Prediction Studies, College of Arts and Sciences, Florida State University, Tallahassee, 32306, U.S.A.

c Agronomy & Soils, Auburn University, Auburn, AL, 36849, U.S.A.

d Earth, Ocean & Atmospheric Sci & Center for Ocean-Atmospheric Prediction Studies, College of Arts and Sciences, Florida State University, Tallahassee, 32306, U.S.A.

e Florida Climate Institute, Florida State University, Tallahassee, 32306, U. S. A.

f Department of Crop Science, North Carolina State University, Raleigh, NC, 27695, U.S.A.

g Department of Geological Sciences and WK Kellogg Biological Station, Michigan State University, East Lansing MI 48824, U.S.A.

Corresponding author: DavideCammarano ()

1

Online Resources(OR) 1: Materials and Methods

OR 1.1 Data sources

Historical observed daily weather data were obtained from the National Climatic Data Center (NCDC). The data consist of daily maximum temperature (Tmax), minimum temperature (Tmin), and rainfall (Rain) for the period 1979-2000. Solar radiation (Srad) was estimated using the empirical relationship of Hargreaves and Samani (1982).

The corresponding meteorological datasets from the regional climate models for the same time period were obtained from The two types of regional climate model data used are: (a) dynamically downscaled reanalysis (data product CLARReS10; Stefanova et al. 2012a), and (b) dynamically downscaled climate model simulations (data product CLAREnCE10; Stefanova et al. 2012b). Both regional data sets are based on dynamical downscaling using the Florida State University-Florida Climate Institute (FSU-FCI) Regional Spectral Model (RSM; Misra et al. 2011) at a 10-km horizontal grid resolution over the Southeastern United States.

For the downscaled reanalysis, boundary conditions for the RSM were provided by the European Centre for Medium Range Weather Prediction (ECMWF) 40-year Reanalysis (ERA-40; Uppala et al. 2005). For the downscaling of the GCM, boundary conditions were provided by two global climate models: the Community Climate System Model (CCSM; Collins et al. 2006) and the Hadley Centre Coupled Model, version 3 (HadCM3; Gordon et al. 2000).

OR 1.2 Bias correction

In addition to dynamical downscaling of the global reanalysis/climate model data with a regional climate model (RCM), a bias-correction was applied to all relevant model fields (Rain, Tmax, Tmin). This bias correction becomes necessary especially for application studies like crop modeling as the RCM’s are imperfect climate models, which generate errors even when they are forced with reliable large scale forcing such as the global reanalysis (Stefanova et al. 2012a; DiNapoli and Misra 2012).

In the case of precipitation, the bias correction entailed a quantile matching, whereby the cumulative distribution functions of the downscaled data and gridded observations (Higgins et al. 1996) for each day within a given calendar month were matched, and the magnitude of each day’s simulated precipitation was rescaled accordingly (Wood et al. 2002). The CLARReS10 data (i.e., the downscaled reanalysis) are available for 1979-2000, while the CLAREnCE10 data (i.e., downscaled GCMs) are available for 1969-1999. Although in this study we only used a common period (1979-1999) bias correction of precipitation was done using the maximum available period for each data set. In other words, the downscaled reanalysis precipitation was bias-corrected using observations for 1979-2000, while the downscaled GCM precipitation was bias corrected using observations for 1969-1999.

For the temperature variables a simple “delta” bias-correction approach was used, whereby the difference between observed station climatology and downscaled modeled climatology for each calendar day was added to the downscaled anomalies. For the bias correction of temperature variables, a common period (1979-1999) was used. Solar radiation values were estimated using the bias-corrected downscaled Tmax and Tmin following Hargreaves and Samani (1982).

OR 1.3 Crop simulation study

Crop simulations were performed using the DSSAT 4.5 CERES-Maize model (Jones and Kiniry 1986; Decision Support System for Agrotechnology Transfer) (Hoogenboom et al. 2010), and the N-Wheat model (APSIM-Nwheat version 1.55s; Asseng 2004). These models have been extensively tested with various field experimental data sets across the world (Carberry et al. 1989; Fraisse et al. 2001; Asseng 2004).

The models were calibrated and evaluated using experimental data from three locations in Alabama for wheat, and one in Georgia for maize obtained from Jones et al. (2012) (Online Resources 2). Genetic coefficients for wheat and maize are shown in the supplemental material (Online Resources 3). The Root Mean Square Error (RMSE) for calibration and validation was 1.89 t ha-1 for maize and 0.74 t ha-1 for wheat (Online Resources2). Six locations were selected for crop model simulations within the SE US, northern and southern Georgia (N-GA, S-GA), northern and southern Alabama (N-AL, S-AL), North Carolina (N-NC), and Florida (N-FL) (Online Resources 4). Two different soil types were used for each location, a silty-clay and a sandy soil (details of the physical and chemical properties are given in Online Resources 5).

Two different levels of irrigation regime were simulated, a non-irrigated regime (rainfed) and a fully irrigated regime. For the irrigated scenario the water was applied as soon as the soil water content fell below 40% of soil water field capacity and re-filling to 85% field capacity in the top 40cm. Both models were run with no nitrogen (N) stress by switching N limitations off (DSSAT) and supplying plenty of N fertilizer (APSIM-Nwheat). The planting density was 6 and 350 plants m-2 for maize and wheat, respectively. For both soils we initialized the crop models with soil water content 50% higher than the lower limit while initial N was not important since N stress was eliminated during simulated growth. Harvest for both crops happened automatically at maturity. The wheat growing season encompasses two years (for example it was sown in 1979 and harvested in 1980) and the maize was sown and harvested in the same year. Both crops were simulated for 20 years starting from 1979.

For average growing season (GS) temperature (AV T) and rainfall (GS Rain) calculations, the growing season was considered to be from 1 March to 1 August for a summer crop (maize) and from 1 October to 1 June for a winter crop (wheat).

Low and high temperatures can be critical for crop growth. For example, Tmin < 8ºC for maize (Cedron et al. 2005) and <0 ºC for wheat (Porter and Gawith 1999) stop development and growth. Maximum temperatures >32ºC causes heat stress (Porter and Gawith 1999) and reduced yield due to accelerated senescence in the wheat model (Asseng et al. 2011). The maize model has no direct heat stress effect other than accelerating the life cycle up to 34 ºC (shortening grain-fill duration) (Cedron et al. 2005). In addition, in-season temperature differences can affect evapotranspiration demand, and therefore water use and the timing of water availability in the soil during the growing season. Numbers of days below and above these thresholds were used as weather data characteristics when comparing the different weather sources with the observed data. Similarly, the average number of high rainfall (>30 mm), and dry days at anthesis (10 days before and after anthesis) were quantified to characterize rainfall variability in the different data sources.

OR 1.4 Data analysis

We compared the multi-year average yields simulated by using observed meteorological data with dynamically downscaled reanalysis and dynamically downscaled GCM. A two-tailed Student’s t-test was used to determine the statistical significance of the differences between the observed and the dynamically downscaled climate variables and between the simulated yields for each station. In addition, the intra-seasonal variability of Tmin, Tmax, precipitation and solar radiation was diagnosed from the standard deviation of these variables within each season. Similarly, the difference between the observed and the dynamically downscaled reanalysis/GCM was also tested for significance using the same approach.

OR 1.5 References

Asseng S (2004) Wheat crop systems – A simulation analysis. CSIRO Publisher, Melbourne, Australia, ISBN 064309119X: pp 275

Asseng S, Foster I and Turner N C (2011) The impact of temperature variability on wheat yields. Global Change Biology 17: 997-1012

Carberry PS, Muchow RC, McCown RL (1989) Testing the CERES-Maize Simulation Model in a Semi-Arid Tropical Environment. Field Crop Res 20:297-315

Cedron FXL, Boote KJ, Nogueira BR, Sau F (2005) Testing CERES-Maize versions to estimate maizeproduction in a cool environment. Eur J Agron 23:89–102

Collins WD, Bitz CM, Blackmon ML, Bonan GB, Bretherton CS, Carton JA, Chang P, Doney SC, Hack JJ, Henderson TB, Kiehl JT, Large WG, McKenna DS, Santer BD, Smith RD (2006) The Community Climate System Model (CCSM3). J Clim 19:2122–2143

DiNapoli S,Misra V (2012)Reconstructing the 20th century high-resoluitonclimateof the southeastern United StatesJ Geophys Res doi:10.1029/2012JD018303

Fraisse CW, Sudduth KA, Kitchen NR (2004) Calibration of the CERES-Maize model for simulating site-specific crop development and yield on claypan soils. Applied EngAgric 17: 547–556

Gordon C, Cooper C, Senior CA, Banks H, Gregory JM, Johns TC, Mitchell JFB and Wood RA (2000) The simulation of SST, sea ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. ClimDyn 16:147-168

Hargreaves GH, Samani ZA (1982) Estimating potential evapotranspiration. J Irrig Drain Eng 108:225-230

Higgins RW, Janowiak JE, Yao Y-P (1996) A gridded hourly precipitation data base for the United States (1963-1993). NCEP/Climate Prediction Center Atlas 1, National Centers for Environmental Prediction, 46pp

Hoogenboom G, Jones JW,WilkensPW, PorterCH, Boote KJ, Hunt LA, SinghU, LizasoJL, WhiteJW, UryasevO, RoyceFS, OgoshiR, GijsmanAJ,Tsuji GY(2010) Decision Support System for Agrotechnology Transfer (DSSAT) Version 4.5 [CD‐ROM]. University of Hawaii, Honolulu, Hawaii

Jones CA, Kiniry JR(1986) CERES-Maize: a simulation model of maize growth and development. Texas A&M University Press, College Station, TX

Jones JW, Boote KJ, Baigorria G (2012) Annual report – climate change in agriculture.

Misra V, Moeller L, Stefanova L, Chan S, O'Brien JJ, Smith TJ III, Plant N (2011) The influence of the Atlantic Warm Pool on the Florida panhandle sea breeze. J Geophys Res 116:D00Q06, doi:10.1029/2010JD015367

Porter J R and Gawith M (1999) Temperatures and the growth and development of wheat: a review. Eur J Agron 10:23-36

Stefanova L, Misra V, Chan SC, Griffin M, O’Brien J, Smith TJ III (2012a) A proxy for high-resolution regional reanalysis for the Southeast United States: assessment of precipitation variability in dynamically downscaled reanalyses. ClimDyn 38:2449-2466

Stefanova L, Misra V, Smith TJ III (2012b) Climate means, trends and extremes in the Everglades: historical data and future projections. 9th INTECOL, Orlando FL, USA, Jun 3-8

Uppala SM, Kallberg PW, Simmons AJ, Andrae U, et al (2005) The ERA-40 re-analysis. Q J R MeteorolSoc 131:2961–3012

Wood AW, Maurer EP, Kumar A, Lettenmaier DP (2002) Long-range experimental hydrologic forecasting for the eastern United States. J Geophys Res 107:Art No 4429

Online Resources2.(a)Simulated vs. observed Maize grain yield for irrigated and rainfed Mid-Season maize hybrid in Tifton (Ga) from maize yield trial data from 1997-2008and ; (b)simulated vs. observed wheat yield for Baldwin cultivar in Tennessee Valley Research and Extension Center (tvs) at Limestone County located in Belle Mina, northern Alabama (34°41'22.24"N, 86°53'10.66"W), Wiregrass Research and Extension Center (wgs) at Henry County Headland, southern Alabama (31°22'39.97"N, 85°18'51.74"W), and E.V. Smith Research and Extension Center (evs) at Macon County, Shorter, Central Alabama (32°25'20.46"N, 85°53'20.76"W)using the 2009/10 and 2010/2011wheat trial data.

Online Resources 3.Genetic coefficients used to describe a mid-season maize hybrid for DSSAT crop model and the wheat cultivar Baldwin for the N-Wheat model.

Maize Genetic parameters
P1 / P2 / P5 / G2 / G3 / PHINT
GDD / - / GDD / # per plant / mg d-1 / GDD
Mid-Cycle / 300 / 0.3 / 990 / 795 / 8.1 / 39
Wheat Genetic parameters
P1V / P1D / P5 / Grno / Fillrate / Stmwt / phylo
- / - / GDD / Kernel g-1stem-1 / mg kernel-1day-1 / g stem-1 / GDD
Baldwin / 3.1 / 4.2 / 750 / 25.7 / 2.2 / 3 / 123

Maize Coefficients:

P1: Thermal time from seedling emergence to the end of the juvenile phase (expressed in degree days above a base temperature of 8ºC) during which the plant is not responsive to changes in photoperiod.

P2: Extent to which development (expressed as days) is delayed for each hour increase in photoperiod above the longest photoperiod at which development proceeds at a maximum rate (which is considered to be 12.5 hours).

P5: Thermal time from silking to physiological maturity (expressed in degree days above a base temperature of 8 ºC).

G2: Maximum possible number of kernels per plant.

G3: Kernel filling rate during the linear grain filling stage and under optimum conditions (mg day-1).

PHINT: Phylochron interval; the interval in thermal time (degree days) between successive leaf tip appearances.

Wheat Coefficients:

P1V: Sensitivity to vernalization.

P1D: Sensitivity to photoperiod.

P5: Thermal time from start of grain filling to maturity.

Grno: Coefficient of kernel number per stem weight at the beginning of grain filling.

Fillrate: Maximum kernel growth rate

Stmwt: Potential final dry weight of a single stem (excluding grain).

Phylo: Phyllocron interval.

1

Online Resources 4.Locations, wheat cultivar used, maize hybrid used, wheat and maize planting date and locations coordinates.

Representative Site / Wheat Cultivar / Maize Hybrid / Wheat Planting Date / Maize Planting Date / Latitude / Longitude
Clarke County, GA (N-GA) / Baldwin / Medium Season / Nov 15 / Apr 15 / 30.86 / -83.29
Tifton, GA (S-GA) / Baldwin / Medium Season / Nov 15 / Mar 20 / 31.43 / -83.58
Belle Mina, AL (N-AL) / Baldwin / Medium Season / 15-Oct / Apr 15 / 34.41 / -86.53
Wiregrass, AL (S-AL) / Baldwin / Medium Season / 29-Oct / Mar 20 / 31.22 / -85.18
Robeson County, NC (N-NC) / Baldwin / Medium Season / Oct 25 / Apr 15 / 34.62 / -79.02
Panhandle Region, FL (N-FL) / Baldwin / Medium Season / Nov 30 / Mar 20 / 30.27 / -83.31

Online Resources 5. Physical and chemical properties of the two soil profiles used for the simulations.

Depth / LL / DUL / SAT / RGF / SBD / SOC / Clay / Silt
(cm) / (cm cm-3) / (cm cm-3) / (cm cm-3) / (g cm-3) / (%) / (%) / (%)
Silty-Clay
15 / 0.194 / 0.400 / 0.540 / 1 / 1.39 / 1.1 / 27.6 / 65.6
46 / 0.208 / 0.391 / 0.535 / 0.543 / 1.4 / 0.46 / 33.8 / 60
74 / 0.242 / 0.410 / 0.542 / 0.301 / 1.39 / 0.27 / 41.8 / 52
135 / 0.251 / 0.413 / 0.542 / 0.124 / 1.38 / 0.22 / 43.9 / 49.3
163 / 0.265 / 0.419 / 0.544 / 0.051 / 1.36 / 0.19 / 46.7 / 45
180 / 0.263 / 0.415 / 0.542 / 0.032 / 1.35 / 0.15 / 46.6 / 44.2
203 / 0.263 / 0.415 / 0.542 / 0 / 1.35 / 0.15 / 46.6 / 44.2
Sandy
10 / 0.079 / 0.198 / 0.381 / 1 / 1.35 / 1.04 / 1.1 / 0.9
15 / 0.060 / 0.156 / 0.361 / 1 / 1.5 / 0.32 / 2 / 0
45 / 0.060 / 0.156 / 0.361 / 0.6 / 1.5 / 0.32 / 2 / 0
119 / 0.055 / 0.144 / 0.355 / 0.262 / 1.48 / 0.1 / 1.6 / 0
180 / 0.055 / 0.143 / 0.355 / 0.062 / 1.56 / 0.09 / 1.5 / 0
203 / 0.055 / 0.143 / 0.355 / 0 / 1.56 / 0.09 / 1.5 / 0

LL: Lower Limit;

DUL: Drain Upper Limit;

SAT: Saturation;

RGF: Root Growth Factor;

SBD: Soil Bulk Density;

SOC: Soil Organic Carbon.

1

Online Resources 6. 20-year means of simulated wheat and maize yield for two different soils and two different management regimes (rainfed and irrigated). Differences with observation-driven simulations significant above the 99% level are indicated with ****, above 95% with ***, above 90% with **, and above 80% with *.

Winter Growing Season (1 Oct - 1 Jun) / Summer Growing Season (1 Mar - 1 Aug)
Clay Soil / Sandy Soil / Clay Soil / Sandy Soil / Clay Soil / Sandy Soil / Clay Soil / Sandy Soil
Raifed / Rainfed / Irrigated / Irrigated / Raifed / Rainfed / Irrigated / Irrigated
(t ha-1) / (t ha-1)
N-GA / OBS / 7.8(0.9) / 7.8(0.9) / 7.9(0.9) / 7.9(0.9) / 8.0(2.7) / 5.9(3.3) / 10.0(1.5) / 10.0(1.5)
ERA40 / 7.7(1.1) / 7.7(1.1) / 7.7(1.1) / 7.7(1.1) / 8.4(2.3) / 6.9(3.0) / 9.8(1.1) / 9.8(1.1)
CCSM / 7.5(0.9) / 7.5(0.9) / 7.5(0.9)* / 7.5(0.9)* / 7.9(2.4) / 5.7(2.7) / 9.4(1.5)* / 9.4(1.5)*
HadCM3 / 7.6(0.7) / 7.6(0.7) / 7.7(0.8) / 7.7(0.8) / 8.5(2.1) / 6.6(2.5) / 9.5(1.3) / 9.5(1.3)
S-GA / OBS / 8.3(0.8) / 8.3(0.8) / 8.3(0.8) / 8.3(0.8) / 8.0(2.3) / 5.8(2.4) / 10.0(1.3) / 10.0(1.3)
ERA40 / 8.3(0.7) / 8.3(0.7) / 8.3(0.7) / 8.3(0.7) / 8.3(2.3) / 6.4(2.9) / 9.7(1.5) / 9.7(1.5)
CCSM / 8.1(0.9) / 8.0(0.9) / 8.1(0.9) / 8.0(0.9) / 8.3(2.3) / 6.4(2.7) / 9.4(1.3)* / 9.4(1.3)*
HadCM3 / 8.2(0.7) / 8.0(0.7) / 8.2(0.7) / 8.1(0.7) / 8.9(2.2) / 6.9(2.8)* / 9.7(1.5) / 9.7(1.5)
N-AL / OBS / 8.4(1.0) / 8.4(1.0) / 8.5(1.1) / 8.5(1.1) / 8.9(3.0) / 6.5(3.7) / 10.8(1.4) / 10.8(1.4)
ERA40 / 7.9(1.3)* / 7.9(1.3)* / 7.9(1.3)* / 7.9(1.3)* / 9.4(1.9) / 7.3(2.6) / 10.0(1.3)** / 10.0(1.3)**
CCSM / 7.7(0.9)*** / 7.7(0.9)*** / 7.8(1.0)*** / 7.8(1.0)*** / 9.5(1.9) / 7.3(3.2) / 10.3(1.3) / 10.3(1.3)
HadCM3 / 8.3(1.1) / 8.3(1.1) / 8.4(1.1) / 8.4(1.1) / 8.7(2.1) / 7.3(2.5) / 9.5(1.1)**** / 9.5(1.1)****
S-AL / OBS / 8.9(1.0) / 8.9(1.0) / 9.0(1.0) / 9.0(1.0) / 9.0(2.5) / 7.0(2.6) / 10.8(1.3) / 10.8(1.3)
ERA40 / 8.5(0.9) / 8.6(0.9) / 8.6(0.9)* / 8.6(0.9)* / 8.9(2.8) / 7.2(3.2) / 10.5(1.4) / 10.5(1.4)
CCSM / 8.5(1.2) / 8.5(1.2) / 8.6(1.1) / 8.6(1.1) / 9.3(2.7) / 7.4(3.4) / 10.5(1.5) / 10.5(1.5)
HadCM3 / 8.5(1.2) / 8.6(1.2) / 8.7(1.1) / 8.7(1.1) / 9.2(2.2) / 7.0(3.2) / 10.3(1.4) / 10.3(1.4)
N-NC / OBS / 8.7(0.9) / 8.7(0.9) / 8.8(1.0) / 8.8(1.0) / 10.1(2.3) / 8.1(3.3) / 11.7(1.2) / 11.7(1.2)
ERA40 / 8.8(1.0) / 8.8(1.0) / 8.9(1.0) / 8.9(1.0) / 10.8(1.5) / 9.4(2.7) / 11.5(0.9) / 11.5(0.9)
CCSM / 7.9(1.3)*** / 7.9(1.3)*** / 8.1(1.4)** / 8.1(1.4)** / 9.4(2.8) / 8.2(3.1) / 10.8(1.3)*** / 10.8(1.3)***
HadCM3 / 8.8(1.0) / 8.8(1.0) / 8.9(1.0) / 8.9(1.0) / 10.8(2.1) / 9.5(2.6)* / 11.5(1.3) / 11.5(1.3)
N-FL / OBS / 8.0(0.7) / 8.1(0.7) / 8.2(0.7) / 8.2(0.7) / 8.0(2.3) / 5.5(2.5) / 10.1(1.0) / 10.1(1.0)
ERA40 / 7.8(0.7) / 7.8(0.7) / 7.9(0.7) / 7.9(0.7) / 8.5(2.2) / 6.4(2.8) / 9.9(1.4) / 9.9(1.4)
CCSM / 7.6(0.9)* / 7.6(0.9)* / 7.8(0.9)* / 7.8(0.9)* / 8.3(2.3) / 6.6(2.9) / 9.8(1.0) / 9.8(1.0)
HadCM3 / 7.6(0.7)** / 7.6(0.7)*** / 7.7(0.7)*** / 7.7(0.7)*** / 8.8(2.6) / 7.2(2.9)** / 10.0(1.4) / 10.0(1.4)

1

Online Resources 7.(a) Relationship between simulated maturity dates using observed weather data and using the CCSM model for N-AL, (b) numbers of dry days at flowering for the Observed weather (full line) data and the CCSM model (dotted line) for N-AL, and (c)numbers of days with Tmin<0 °C for the Observed weather (full line) data and the CCSM model (dotted line) for N-AL.

1