HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Lobell, D. B. Articles by Ortiz-Monasterio, J. I. Search for Related Content PubMed Articles by Lobell, D. B. Articles by Ortiz-Monasterio, J. I. Agricola Articles by Lobell, D. B. Articles by Ortiz-Monasterio, J. I. Related Collections Agroclimatology Wheat Crop Models

Wheat

Impacts of Day Versus Night Temperatures on Spring Wheat Yields

A Comparison of Empirical and CERES Model Predictions in Three Locations

^a Energy and Environment Directorate, Lawrence Livermore National Lab., Livermore, CA 94550
^b International Maize and Wheat Improvement Center (CIMMYT), Wheat Program, Apdo, Postal 6-641, 06600 Mexico D.F., Mexico

^* Corresponding author (dlobell{at}llnl.gov)

Received for publication July 14, 2006.

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Trends in recent temperature observations and model projections of the future are characterized by greater warming of daily minimum (tmin) relative to maximum (tmax) temperatures. To aid understanding of how tmin and tmax differentially affect crop yields, we analyzed variations of regional spring wheat yields and temperatures for three irrigated sites in western North America that were characterized by low correlations between tmin and tmax. The crop model CERES-Wheat v3.5 was evaluated in each site and used to project future response to temperature changes. Tmin and tmax exhibited distinct historical correlations with yields, with CERES successfully capturing the observed relationships in each region. In the Yaqui Valley of Mexico, historical yields were strongly correlated with tmin but not tmax. However, CERES projections of response to increased tmin or tmax (holding other variables constant) were similar (6% °C^–1), indicating that the apparent historical importance of tmin mainly results from covariation between temperatures and solar radiation and not greater direct effects of tmin on yields. In the San Luis-Mexicali Valley of Mexico and in the Imperial Valley of California, the opposite was observed: historical yield correlations with tmin and tmax were similar, but projected responses to tmax were roughly three times larger than tmin. The latter is explained by opposing effects of tmin and tmax on grain filling rates in CERES, with higher tmin increasing harvest indices. This model mechanism was not clearly supported by historical data and remains an area of uncertainty for projecting yield responses to climate change.

Abbreviations: CIMIS, California Irrigation Management Information System • University of East Anglia • DSSAT, decision support system for agrotechnology transfer • DTR, diurnal temperature range (tmax – tmin) • ET₀, reference evapotranspiration • HI, harvest index • IV, Imperial Valley, California • rad, solar radiation • SV, San Luis and Mexicali Valleys, Mexico • tmax, daily maximum temperature • tmean, daily mean temperature • tmin, daily minimum temperature • YV, Yaqui Valley, Mexico

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
A POTENTIALLY IMPORTANT TRAIT of anthropogenic climate change is asymmetric warming between day and night. On a global average, daily minimum temperature (tmin) has risen more than twice as fast as daily maximum temperature (tmax) over the past century (Easterling et al., 1997), and most models simulate a further reduction in the diurnal temperature range (DTR = tmax – tmin) in the next century (Dai et al., 2001).

For several reasons, the response of crop yields to temperature change may depend on the relative warming of tmin and tmax (Peters et al., 1971; Ziska and Manalo, 1996; Stone, 2001; Peng et al., 2004). Processes such as photosynthesis and transpiration are concentrated in daylight hours and therefore should be more responsive to tmax, whereas processes such as respiration occur throughout day and night. Crop development rates and the duration of critical phases such as grain filling may also be differentially sensitive to tmin and tmax. In addition, impacts of extreme hot or cold temperatures, such as winterkill in wheat or spikelet sterility in rice, depend on changes in daily extremes (Ziska and Manalo, 1996; Porter and Gawith, 1999).

Models capable of simulating the crop yield response to temperature and other environmental factors are useful tools to anticipate the impacts of climate change and to develop appropriate responses. Commonly used models of the major cereal crops, such as the models included in the DSSAT software (Jones et al., 2003), consider the effect of temperature on rates of several processes affecting crop yields. These processes include vernalization, crop phasic development, leaf appearance and expansion rates, photosynthesis and respiration, evapotranspiration, and grain filling (Ritchie and NeSmith, 1991; Wilkens and Singh, 2001).

Several studies have used crop models to investigate the impacts of asymmetric warming on rainfed crops in the USA. Rosenzweig and Tubiello (1996) compared the effects of tmin and tmax on winter wheat in the central USA using CERES-Wheat and found that, for identical changes in average temperatures, increasing tmin more than tmax resulted in higher simulated yields than when increasing both by equal amounts. This disparity was attributed to reduced winterkill in experiments with greater tmin increases. Dhakhwa and Campbell (1998), using the EPIC model for maize, soybean, and wheat in the Southeastern USA, similarly report greater simulated yields for scenarios of greater nighttime warming than for scenarios of equal day–night warming. In their case, the main mechanism responsible was reduced water stress in the asymmetric warming scenario, where lower daytime temperatures led to reduced evaporative losses.

Whether these models properly represent the separate effects of tmin and tmax has not been specifically addressed. This is primarily because historical interannual variations in growing season tmin are highly correlated with tmax in most regions. Thus, "validation" of the model by comparing yield simulations with historical records does little to elucidate the model's ability to capture separately the responses to tmin and tmax.

However, there are some growing regions in the world for which tmin and tmax are poorly correlated historically. These provide a unique opportunity to develop understanding of tmin and tmax effects. Specifically, there are several regions where interannual correlations between tmin and tmax for 1981 through 2002 are not significant at p = 0.05 (Fig. 1 ). This is displayed for four different months for the 0.5° by 0.5° grid cells within the Climate Research Unit climate database (Mitchell and Jones, 2005). Although the mechanisms behind these correlations are beyond the scope of this paper, it is interesting to note that for most of the major cropping regions (e.g., Indo-Gangetic Plains, U.S. Corn Belt, Europe, and China), the correlations are typically greater than 0.8 in this dataset.

View larger version (27K): [in this window] [in a new window]   Fig. 1. Areas with nonsignificant (p > 0.05) correlation b/w daily minimum temperature and daily maximum temperature for 1981–2002 in Climate Research Unit dataset (shown in black) for four different months.

Here we examine the mechanisms behind the results reported in Lobell et al. (2005) for spring wheat in Northwest Mexico. The region of analysis has been expanded to include the Imperial Valley of California and two sites in Mexico. Specific goals were to (i) analyze the empirical relationships between tmin, tmax, radiation, and yields in the three sites; (ii) test the ability of CERES-Wheat to reproduce these empirical relationships; (iii) use CERES-Wheat to simulate the effect of changing tmin or tmax while holding other climatic variables constant; and (iv) compare estimates of climate change impacts from CERES to those from multiple regression models based on historical data. The latter objective provides a measure of how well past empirical relationships can be used to estimate impacts of future climate change.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Site Description and Data Sources
This study was conducted with data from three sites (Fig. 2 ). The Yaqui Valley (YV; 27.4° N, 109.9° W) is in the state of Sonora, Mexico. The San Luis/Mexicali Valley (SV; 32.5° N, 114.8° W) spans the border of Sonora and Baja California, Mexico. The Imperial Valley (IV; 33.0° N, 115.4° W) is at the southernmost edge of Imperial County, California. Total spring wheat production in 2002 was 1.08, 0.49, and 0.13 million t in YV, SV, and IV, respectively. Wheat is typically planted in late fall or early winter (November–January) and harvested in late spring (April–June). All regions have been irrigated throughout the study period (post-1980) and are heavily fertilized (>200 kg N ha^–1) and thus achieve yields that commonly approach genetic potential. As a result, temperature and radiation variations are among the major controls on average yields (Fischer, 1985).

View larger version (124K): [in this window] [in a new window]   Fig. 2. Study site locations shown on an enhanced vegetation index from the Moderate Resolution Imaging Spectroradiometer image from March 2006. Dark pixels indicate higher values of enhanced vegetative index corresponding to dense vegetation, which are mostly wheat crops in this region at this time of year.

All three sites were characterized by relatively low correlations between tmin and tmax, which allowed us to evaluate the potentially unique effects of each (Table 1). Correlations between temperatures and rad are more variable between regions. In YV, rad is negatively correlated with tmin and positively correlated with tmax, whereas tmin and tmax are negatively correlated with rad in SV and positively correlated with rad in IV. Most of these correlations were not statistically significant at p = 0.05 (two-sided t test), with the exceptions of rad and tmax in YV and IV.

View larger version (14K): [in this window] [in a new window]   Fig. 3. The response of (a) degree day, (b) photosynthesis rate, and (c) grain filling rate calculations in CERES v3.5 to changes in average temperature and diurnal temperature range. Equations are shown in Table 2.

As with nearly all process-based models, the use of CERES requires that one first calibrate several model parameters (Table 3). Here, crop parameters for YV were taken from Bell and Fischer (1994) and for SV from Lobell et al. (2005). The parameters for SV were also used in IV because the two regions are close and because data for an independent calibration of IV were not available. Crop growth was simulated without water or nutrient stress because crops in these regions were irrigated and heavily fertilized.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

View larger version (21K): [in this window] [in a new window]   Fig. 4. January–April average daily minimum temperature (tmin), daily maximum temperature (tmax), and solar radiation (rad) vs. observed (top) and CERES-simulated (bottom) yields in YV.

View larger version (26K): [in this window] [in a new window]   Fig. 5. CERES-simulated biomass (a) and yield (b) response to 1°C increase in daily minimum temperature (tmin) or daily maximum temperatue (tmax) in Yaqui Valley, Mexico.

San Luis/Mexicali Valley
Historical relationships between weather and yields in SV revealed similar effects of tmin and tmax (Fig. 6 ). The empirical effects of tmin (–8.0% °C^–1) and tmax (–10.1% °C^–1) were well reproduced by CERES simulations (Table 4). The empirical effect of solar radiation (8.8% MJ^–1) was also successfully modeled by CERES (9.8% MJ^–1). Thus, similar to the case of YV, CERES was able to capture historical yield variations and their correlation with climatic variables. Unlike YV, where the empirical effect of tmin dominated tmax, the effects of each were apparently similar in SV. Correlations between CERES variables indicated that yield reductions were less associated with days to maturity than for YV (Table 5). Yield variations in SV were also only weakly associated with changes in HI, which reflected changes in grain filling rates (r = 0.54). Thus, of the three processes in Fig. 3, variations in photosynthesis rates seemed to be most important for yields in SV, rather than crop development or grain filling rates. Because CERES was run without water or nutrient stress, simulated yield responses to temperatures did not involve changes in vapor pressure deficits or water stress.

View larger version (17K): [in this window] [in a new window]   Fig. 6. January–April average daily minimum temperature (tmin), daily maximum temperature (tmax), and solar radiation (rad) vs. observed (top) and CERES-simulated (bottom) yields in San Luis and Mexicali Valleys, Mexico.

View larger version (24K): [in this window] [in a new window]   Fig. 7. CERES-simulated biomass (a) and yield (b) response to 1°C increase in daily minimum temperature (tmin) or daily maximum temperature (tmax) in San Luis and Mexicali Valleys, Mexico.

View larger version (18K): [in this window] [in a new window]   Fig. 8. January–April average daily minimum temperature (tmin), daily maximum temperature (tmax), and solar radiation (rad) vs. observed (top) and CERES-simulated (bottom) yields in Imperial Valley, California.

View larger version (24K): [in this window] [in a new window]   Fig. 9. CERES-simulated biomass (a) and yield (b) response to 1°C increase in daily minimum temperature (tmin) or daily maximum temperature (tmax) in Imperial Valley, California.

CERES Variations in Harvest Index
Harvest index is a diagnostic variable in CERES, determined only after final biomass and grain yields have been computed. There is therefore no single equation used to compute HI that can be evaluated directly. Instead, HI combines the effects of several model variables, including grain number, the temperature-dependent grain filling rate (Table 2), the genetic coefficient G2 (Table 3), and photosynthesis rates during grain filling. To evaluate the simulation of HI in CERES, historical yields were compared with CERES-simulated values of yield and biomass (Table 6). In cases where actual yields were more highly correlated with simulated yields than biomass (YV and IV), this indicated that variations in HI simulated by CERES helped to improve model agreement with observations compared with the use of a constant HI. This improvement thus provided some support of the treatment of HI in CERES. In contrast, in SV, agreement with historical yields was higher for simulated biomass than yields, indicating that simulation of HI changes decreased model performance at this site.

[1]

where RSS₁ is the residual sum of square errors for a linear model with observed yields as the response variable and CERES simulated yields as the predictor variable, and RSS₂ is the corresponding value for a model with CERES simulated biomass as the predictor variable. The p values of the F statistic were higher than 0.1 in all three sites. Thus, although observations in two of the three sites suggest that CERES treatment of HI improved model performance, this improvement was not statistically significant at p = 0.10.

A recent synthesis of CERES modeling studies (Timsina and Humphreys, 2006) showed that predictions of grain yield across several studies were slightly better than for biomass yield (RMSE = 13% vs. 16%). However, the evaluation of biomass yields was based on just two studies, one of which used CERES-Wheat version 2.1 (Hundal and Kaur, 1997). Wilkens and Singh (2001) questioned the temperature response of grain filling rates in CERES, given the expectation that lower night temperatures should reduce night respiration and therefore enhance grain filling. Overall, it seems that CERES simulations of HI are critical for projecting response to changes in DTR and that this aspect of model performance requires further testing.

Empirical Versus CERES Model Predictions of Climate Change Impacts
In all three sites, the empirical effect of tmin and tmax inferred from historical yield variations differed from the responses simulated by CERES when changing one variable while holding all others constant. Indeed, an often-cited strength of process-based models is their ability to simulate yields for combinations of temperatures and radiation that have not previously been observed. However, in cases where process models cannot be used (e.g., because of lack of data required for model calibration), purely empirical (i.e., statistical) models can provide a simple and potentially useful alternative. For comparison with CERES predictions of yield response to temperature increases, we computed multiple linear regression models of yield response, with January–April average tmin, tmax, and rad comprising the three predictor variables. The models are summarized in Table 7. For all models, interaction terms and a term for ET₀ were considered but did not significantly improve model R² and were therefore not used.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

The CERES model successfully simulated historical yield variations and their correlations with tmin, tmax, and solar radiation. The model thus seems to be a useful tool to assess the impact of change in these climatic variables at these sites.

Historical correlations between yields and tmin (or tmax) were often at odds with the modeled sensitivity to tmin (or tmax), with the latter measured as the average change in CERES-simulated yields when holding all other variables constant. This finding reflected the important role of covariation of tmin and tmax with each other and especially with solar radiation in historical temperature–yield relationships.

CERES sensitivities of simulated yields to tmin were substantially different than for tmax in two of the three sites, with tmax increases resulting in 3 times greater yield losses than similar increases in tmin. In contrast, biomass sensitivities were similar for tmin and tmax, indicating that simulated changes in HI explain the difference in model sensitivities. Further testing of HI simulations in CERES is therefore critical to improved understanding of tmin and tmax effects on wheat yields.

Predictions of average yield response to climate change using multiple regression models based on historical data differed from CERES predictions by less than 3% for most sites. Thus, although correlations between individual temperature variables and yields may differ with CERES model sensitivities, multiple regression models that include solar radiation agree well with CERES predictions. Process-based models such as CERES offer distinct advantages, such as an ability to simulate adaptation responses or effects of rising CO₂. However, in the absence of calibrated process-based models, empirical models can provide a useful means to assess climate change impacts.

For the regions studied here, the effects of higher temperatures were consistently negative. Modeled sensitivities for symmetric warming ranged from –10% °C^–1 at the warmest site (YV) to –7% °C^–1 at the coolest site (IV). Maintaining current yields in the face of 1 to 4°C temperature increases expected over the next 50 yr (Ruosteenoja et al., 2003) therefore represents a considerable challenge to the farmers of these regions. Crop breeders will also be challenged to generate new varieties that maintain or improve current yields in a future warmer climate.

	ACKNOWLEDGMENTS

 
We thank Tony Hunt for helpful discussions and three anonymous reviewers for comments that improved the manuscript. This work was performed under the auspices of the U.S. Department of Energy by University of California, Lawrence Livermore National Laboratory under Contract W-7405-Eng-48. Support was also provided by the David and Lucille Packard Foundation.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 

• Bell, M.A., and R.A. Fischer. 1994. Using yield prediction models to assess yield gains: A case study for wheat. Field Crops Res. 36:161–166.
• Bristow, K.L., and G.S. Campbell. 1984. On the relationship between incoming solar-radiation and daily maximum and minimum temperature. Agric. For. Meteorol. 31:159–166.[CrossRef]
• Dai, A., T.M.L. Wigley, B.A. Boville, J.T. Kiehl, and L.E. Buja. 2001. Climates of the twentieth and twenty-first centuries simulated by the NCAR climate system model. J. Clim. 14:485–519.[CrossRef]
• Dhakhwa, G.B., and C.L. Campbell. 1998. Potential effects of differential day-night warming in global climate change on crop production. Clim. Change 40:647–667.
• Easterling, D.R., B. Horton, P.D. Jones, T.C. Peterson, T.R. Karl, D.E. Parker, M.J. Salinger, V. Razuvayev, N. Plummer, P. Jamason, and C.K. Folland. 1997. Maximum and minimum temperature trends for the globe. Science 277:364–367.[Abstract/Free Full Text]
• Fischer, R.A. 1985. Number of kernels in wheat crops and the influence of solar-radiation and temperature. J. Agric. Sci. 105:447–461.
• Hundal, S.S., and P. Kaur. 1997. Application of the CERES-Wheat model to yield predictions in the irrigated plains of the Indian Punjab. J. Agric. Sci. 129:13–18.
• Jones, J.W., G. Hoogenboom, C.H. Porter, K.J. Boote, W.D. Batchelor, L.A. Hunt, P.W. Wilkens, U. Singh, A.J. Gijsman, and J.T. Ritchie. 2003. The DSSAT cropping system model. Eur. J. Agron. 18:235–265.[CrossRef]
• Lobell, D.B., J.I. Ortiz-Monasterio, G.P. Asner, P.A. Matson, R.L. Naylor, and W.P. Falcon. 2005. Analysis of wheat yield and climatic trends in Mexico. Field Crops Res. 94:250–256.
• Mitchell, T.D., and P.D. Jones. 2005. An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Clim. 25:693–712.
• National Agriculture Statistics Service. 2006. Published estimates data base. Available at http://www.nass.usda.gov/. NASS, USDA, Washington, DC (accessed 1 June 2006; verified 5 Dec. 2006).
• Peng, S., J. Huang, J. Sheehy, R. Laza, R. Visperas, X. Zhong, G. Centeno, G. Khush, and K. Cassman. 2004. Rice yields decline with higher night temperature from global warming. Proc. Natl. Acad. Sci. USA 101:9971–9975.[Abstract/Free Full Text]
• Peters, D.B., J.W. Pendelton, R.H. Hageman, and C.M. Brown. 1971. Effect of night air temperature on grain yield of corn, wheat, and soybeans. Agron. J. 63:80–93.
• Porter, J.R., and M. Gawith. 1999. Temperatures and the growth and development of wheat: A review. Eur. J. Agron. 10:23–36.
• Ritchie, J.T., and D.S. NeSmith. 1991. Temperature and crop development. p. 5–29. In J.A.J.T.R. Hanks (ed.) Modeling plant and soil systems. Vol. 31. ASA, Madison, WI.
• Rosenzweig, C., and M.L. Parry. 1994. Potential impact of climate-change on world food-supply. Nature 367:133–138.[CrossRef][Web of Science]
• Rosenzweig, C., and F.N. Tubiello. 1996. Effects of changes in minimum and maximum temperature on wheat yields in the central US: A simulation study. Agric. For. Meteorol. 80:215–230.[CrossRef]
• Ruosteenoja, K., T. Carter, K. Jytha, and H. Tuomenvirta. 2003. Future climate in world regions: An intercomparison of model-based projections for the new ipcc emissions scenarios. Finnish Environment Institute, Helsinki, Finland.
• Secretaría de Agricultura, G. Desarrollo Rural, Pesca y Alimentación (SAGARPA). 2005. Sistema integral de informacion agroalimentaria y pesquera. Available at www.siap.sagarpa.gob.mx/ (accessed 1 June 2006; verified 5 Dec. 2006).
• Sheehy, J.E., P.L. Mitchell, and A.B. Ferrer. 2006. Decline in rice grain yields with temperature: Models and correlations can give different estimates. Field Crops Res. 98:151–156.
• Stone, P. 2001. The effects of heat stress on cereal yield and quality. p. 302. In A.S. Basra (ed.) Crop responses and adaptations to temperature stress. Food Products Press, Binghamton, NY.
• Timsina, J., and E. Humphreys. 2006. Performance of CERES-Rice and CERES-Wheat models in rice–wheat systems: A review. Agric. Syst. 90:5–31.
• Wilkens, P., and U. Singh. 2001. A code-level analysis for temperature effects in the CERES models. p. 1–7. In J. White (ed.) Modeling temperature response in wheat and maize. CIMMYT, El Batan, Mexico.
• Ziska, L.H., and P.A. Manalo. 1996. Increasing night temperature can reduce seed set and potential yield of tropical rice. Aust. J. Plant Physiol. 23:791–794.

This article has been cited by other articles:

				P. V. V. Prasad, S. R. Pisipati, Z. Ristic, U. Bukovnik, and A. K. Fritz Impact of Nighttime Temperature on Physiology and Growth of Spring Wheat Crop Sci., November 24, 2008; 48(6): 2372 - 2380. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Lobell, D. B. Articles by Ortiz-Monasterio, J. I. Search for Related Content PubMed Articles by Lobell, D. B. Articles by Ortiz-Monasterio, J. I. Agricola Articles by Lobell, D. B. Articles by Ortiz-Monasterio, J. I. Related Collections Agroclimatology Wheat Crop Models