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 Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (2) Citing Articles via Google Scholar Google Scholar Articles by Jeutong, F. Articles by Smith, O.S. Search for Related Content PubMed Articles by Jeutong, F. Articles by Smith, O.S. Agricola Articles by Jeutong, F. Articles by Smith, O.S.

CROP BREEDING, GENETICS & CYTOLOGY

Comparison of Bioclimatic Indices for Prediction of Maize Yields

^a Dep. of Biometry, Univ. of Nebraska-Lincoln, 103 Miller Hall, Lincoln, NE 68583-0712 USA
^b Dept. of Computer Science and Engineering, Univ. of Nebraska-Lincoln, W205.7 Nebraska Hall, Lincoln, NE 68588-0115 USA
^c Pioneer Hi-bred International, 730 Northwest 62, P.O. Box 1004, Johnston, IA 50131 USA

keskridge1{at}unl.edu

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussions Conclusion REFERENCES

 
Yield prediction across a target production zone with varying strategies of agronomic practices has been a challenging problem to plant breeders when testing new genotypes for release. This study focused on comparing the importance of a new bioclimatic index called biological windows and six other traditional environmental indices as predictor variables of maize yields across sites (farmers' fields) and years, using a simple linear regression model. The yield data were collected for six hybrids evaluated in strip tests at 57 to 186 sites throughout Iowa during 1987–1994. The biological windows index was based on the Newhall Simulation Model and estimated the number of days the soil was moist and above 5°C. The environmental indices were July precipitation, temperature, the product of July precipitation and temperature, and the equivalent values for August. Because the actual values for the indices were not recorded at each site, all the indices were estimated for each site as the weighted averages of the data from 112 Iowa weather stations. Across years and within the Iowa sites, the mean percentiles of R-square distributions showed that biological windows had less predictive value for maize yields than the more traditional indicators such as August precipitation and temperature. For all indices, across years and within sites had much greater mean R-squares than across sites and within years, which had very low predictive values. For predicting yield across years within sites, there appeared to be an advantage in using August precipitation or the product of August precipitation and temperature over the five other indices. The R-square values for these two indices were at least 0.60 in 80% of the test sites for five hybrids.

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussions Conclusion REFERENCES

 
GENOTYPE x ENVIRONMENT INTERACTION (GEI) is a common problem that plant breeders face when releasing advanced plant materials for use in any relatively large target zone. Lin et al. (1986) reviewed alternative methods addressing this problem via stability analysis of genotypes and/or prediction over environments. A major limitation of such approaches in regression type analyses is that the use of the mean of the genotype in each environment as the independent environmental index variable often doesn't adequately characterize environments. Performance of a genotype in an environment relates not only to the type of environments, but also the interaction, intensity and timing of processes or events (Cooper and Byth, 1996). Crop response depends on a suite of bioclimatic variables that relate to particular characteristics of the genotypes being evaluated (Corbett, 1998). Consequently, there is considerable need for the identification and use of bioclimatic variables for the evaluation and comparison of advanced genotypes across large growing regions. The identification of such key bioclimatic variables coupled with a simple model with good yield prediction capability over large production zones in the presence of GEI would reduce the efforts invested at the early stage of genotypic evaluation. A good predictive model could help with minimizing the number of test sites needed within years and would aid with seed production decisions by predicting genotypic performance in future years.

Crop models that use direct or derived weather, agroclimatic, soil, and agricultural practice variables with varying degree of complexity have been developed for various crop prediction purposes. Bondavalli et al. (1970) and Nielsen et al. (1996) found that the most important weather variables that affect grain yield prediction of maize (Zea mays L.) in the U.S. Corn Belt region were precipitation and temperature occurring during the months of May and August. However, other studies stressed the importance of these two weather variables on maize yields during the critical period of July (tasseling and silking) and August (grain filling) (Teigen, 1991; Teigen and Thomas, 1995). This period corresponds to the critical growth and development stages, i.e. from initiation of maize floral parts to early stage of grain maturation (Runge, 1968; Duchon, 1986). Similar observations were reported in Canada (Dirks and Bolton, 1980, 1981). These two weather variables were thought to be a reflection of evapotranspiration (Runge, 1968) that can be roughly estimated by the product of the two variables (precipitation x temperature) (Teigen and Thomas, 1995). The significant effect of these two factors in determining maize yield variability was enhanced when other factors such as soil type and fertility (Liang et al., 1991) and crop intercepted solar radiation (Muchow and Sinclair, 1995) were considered. Other important factors were the sum of the ratios of actual to potential evapotranspiration (Dale and Daniels, 1995), soil compaction and structure (Dirks and Bolton, 1980), soil management, crop rotation and levels of N-P-K fertilizers (Bondavalli et al., 1970; Voss et al., 1970; Johnson et al., 1992).

Another approach to predicting genotype performance and characterizing target environments can be based on soil biological windows and soil climate regimes obtained from the Newhall Simulation Model (Van Wambeke et al., 1992; Van Wambeke, 1981; 1982; and 1985). The concept of biological windows is largely from Soil Taxonomy (Soil Survey Staff, 1975; USDA NRCS, 1999) where the soil biological windows value is the cumulative days throughout the year when soils are moist and above 5°C. The soil biological windows value underlies the classification of soil moisture regimes and can be used to portray shifts in moisture regimes through the cropping seasons and individual years (Waltman et al., 1997). Yamoah et al. (1998a; 1998b) showed that soil biological windows at 5°C had good within site predictive value for corn yields in multiyear rotation experiments in the rainfed portion of eastern Nebraska. Given that the Newhall Simulation Model may be used to estimate bioclimatic conditions which can be based on data collected by the cooperative weather station network, biological windows is a potential predictor of hybrid corn performance across large regions of the corn belt.

Although the above cited studies were successful in identifying relevant factors that influence maize yields, a major limitation was that results were based on only one or a few locations, and usually for only a few years. Maize yields from seed companies' strip trials carried out in many years and frequently in hundreds of locations per year can be used to encompass most of the constraints encountered in the previous studies.

The objectives of this study were to use strip trial hybrid yields from Pioneer Hi-Bred, Intl. to compare the maize yield predictive ability of biological windows to traditional temperature and precipitation variables when predictions were made both across sites within years and across years within sites. Outcomes from this study will help focus attention on the impact of bioclimatic variables in varietal testing and can help with planning corporate seed production activities with regard to multiple-site-year field trials.

	Materials and methods

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussions Conclusion REFERENCES

 
Choice of Hybrids, Years, and Testing Sites
The choice of hybrids was based on their substitution rates over sites and years as follows. Six hybrids (3379, 3394, 3417, 3475, 3525, and 3563) whose yield records were available for at least 3 out of 8 yr (1987–1994) of testing at the same site (farmer's field) and with at least an average of 50 sites and a minimum of 30 sites per year were retained in the available data sets. The number of years and average number of test sites per year for these hybrids are included in Table 1 . The minimum and maximum average number of sites that met the requirement of at least 3 yr of testing were 57 (3–5 yr) for 3475 and 186 (3–5 yr) for 3417, respectively. The sites were identified by geographical coordinates approximated by zip code centroids across Iowa, where most of the hybrid trials were evaluated under rainfed conditions.

The soil biological windows value, which is defined as the cumulative number of days on the year basis that the soil is moist (-0.3 to -1.5 MPa) and above 5°C (Newhall and Berdanier, 1992; Waltman et al., 1997). A software program, the Newhall Simulation Model, was used to estimate the biological windows values for each of the year and station combinations on the basis of mean monthly air temperature and precipitation (Van Wambeke et al., 1992). The concept of biological windows may serve as a coupled bioclimatic indicator of maize yield since it calculates a water balance (precipitation-potential evapotranspiration), in relationship to a soil temperature calendar (when a soil reaches 5 and 8°C). For Iowa, a root zone water-holding capacity of 200 mm was assumed across all weather stations to simplify the model runs. Biological windows may serve as a reasonable predictor of yield across years and sites since biological windows relates to basic soil processes during the growing season, such as mineralization, nitrification, and microbial activity, and integrates water balance and soil temperature. The biological windows for the sites were estimated in the same way as the site temperature and precipitation.

Method for Comparing Environmental Indices as Yield Predictor Variables
Simple linear regressions of each hybrid yield on each of the seven environmental indices were computed. July precipitation, temperature, precipitation x temperature product used as a surrogate evapotranspiration index, August precipitation, temperature, precipitation x temperature, and biological windows were used as independent variables for each site–year combination. For each site, regressions were computed across years for each hybrid. For each year, regressions were computed across sites for each hybrid. Our main reason for using simple linear regression was that even though the nature of maize yield response to weather variables was traditionally nonlinear (Teigen, 1991), it may be assumed that within the two critical month period for maize, the relationship between maize yields and important weather variables was strongly linear. Within a threshold evapotranspiration (250–300 mm), a strong linear trend of maize yield with varying amounts of water applied was found (Hillel and Gubon, 1973). Similar relationships were reported for maize (Nielsen et al., 1996), wheat (Triticum aestivum L.), millet [Pennisetum glaucum (L.) R. Br.], and oat (Avena sativa L.) in the northern plains of the USA (Hanks et al., 1969). In addition for the over-year analyses, many sites only had a few years so the number of independent variables that could be fit was very limited.

The R-square estimates from the regressions were then used as criteria for evaluating and comparing the indices as yield predictor variables within and across sites and years. The rationale was the higher the frequency of larger (explaining more of the variability) R-square values associated with an index, the better the index. This could be achieved by inspection of empirical cumulative distribution functions (CDFs) or more easily by graphical representation of the CDFs for the indices.

	Results and discussions

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussions Conclusion REFERENCES

 
Empirical distributions of R-square values (quartiles, mean, and median) for each hybrid under the seven bioclimatic indices and the two regression methods (across years within sites and across sites within years) are shown in Table 1. The graphs for the empirical CDFs for across years within sites R-squares are presented in Fig. 1 , using five indices and 13 R-square percentiles values: the minimum, the first and third quartiles, and the other percentiles generated by increment of 10 starting from the 10th percentile. The July and August precipitation x temperature variables, which were very similar to the precipitation variables, were not included in Fig. 1.

View larger version (50K): [in this window] [in a new window]   Fig. 1 Relative frequencies of R-square values across years within sites for five bioclimatic indices and six maize hybrids (BW = Biological Windows; JP = July precipitation; JT = July temperature; AP = August precipitation; AT = August temperature)

In the across years within site method, large R-square values were indicative of a strong linear relationship between yield and bioclimatic indices and suggested that previous yield information of hybrids in a given site can serve as a simple empirical model (with limited input data) for predicting hybrid yield performance at the same site if future bioclimatic index values are presumed. The best fit was obtained with either August precipitation or August precipitation x temperature. For these two indices, the average R-square and median values were above 0.60 for all hybrids, except hybrid 3475. Although the variances of the estimated parameters were not directly computed, approximately 40 to 50% reduction in the interquartile ranges of R-squares under these two indices for hybrids 3394, 3417, 3525, and 3563, is indicative of relatively more stable estimates of the R-squares of the two indices. The relationship between July-August rainfall and temperatures on maize yields have been reported at many U.S. Cornbelt research stations sites (Runge, 1968; Dirks and Bolton, 1981; Teigen, 1991; Teigen and Thomas, 1995). In the across years within site method with all hybrids except 3475, biological windows performed worse than August precipitation. For 3475, August temperature had a higher mean R-square than biological windows. For hybrids 3394, 3417, 3525, and 3563, there appeared to be a net advantage using with August precipitation or August precipitation x temperature as independent variables over the five other indices for yield prediction. For these two indices and four hybrids, the relative frequency of low R-square values (low R-square sites) was small. With these two variables, the number of sites where the R-square values were at least equal to 0.60 approximated 80% of the total number of test sites for each hybrid, which translates into 131 out of 164, 149 out of 186, 105 out of 131 and 70 out of 87 sites, respectively, for hybrids 3394, 3417, 3563, and 3525.

These results for the across years regression method indicated that simpler weather variables chosen at the proper time in the growing season are preferred to biological windows.

It is possible that the predictability of biological windows could be improved by obtaining biological windows values on a per month basis instead of over the entire year. Also, the threshold of the soil temperature being above 5°C may be too low for maize. Increasing the threshold to 10°C could also improve predictability of biological windows. Nevertheless, the use of biological windows as defined by Newhall and Berdanier (1992) will likely not predict as well as August precipitation when predicting maize yields over years for particular sites.

In the across sites within years regression method, the consistently poor R-square values associated with any of the seven bioclimatic indices suggested that when used alone as an independent variable, major sources of yield variability across locations were not accounted for, and therefore, would fail to predict yield based on across sites information. These results confirm previous findings (Voss et al., 1970; Dirks and Bolton, 1980). In N-P-K fertilizer trials conducted in a relatively large number of locations and in long-term rotation experiments at one site, the R-square values from polynomial regressions of maize yields on weather variables tended to be high within individual treatment or rotation, with their magnitudes depending on other factors such as soil fertility levels or effective rooting depth and root zone available water holding capacity. The empirical distributions of R-squares under the across sites regression method were worthless for any prediction purpose, regardless of the hybrid and environmental index used and were not plotted. Even in the best situation (hybrid 3475 in this case), the maximum R-square value still falls below 0.60.

There could be several factors in this study that limited the ability of any of the indices to predict yields over sites within a year. Improved yield prediction across sites within years would likely require improved site precipitation estimates, better site information (soil type and fertility, management practices, root zone water-holding capacity, elevation, aspect, and slope), and additional precision in georeferencing. However, with these data, zip code was the only available site variable other than yield. The field test sites were identified by their geographical coordinates of the zip code centroids which likely resulted in substantial errors in georeferencing and precipitation predictions. Similarly, the inverse-distance weighting method possibly gave excessive influence to individual weather stations giving a "bull's-eye" effect, resulting in imprecise precipitation predictions for the test plots. This outcome is often common with precipitation data during a growing season in the Midwest because of thunderstorm events and their inherent variability. Yet, even with improved site information on fertility and management practices, better georeferencing and a more precise method of predicting precipitation at the test sites, substantial improvement of predicting yields across sites within years is unlikely without precipitation measured at each of the sites.

	Conclusion

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussions Conclusion REFERENCES

 
In summary, this study showed that across year, within site maize yield predictive ability of biological windows does not compare favorably with other simpler bioclimatic variables. Maize yield prediction across years, within sites using August precipitation and August precipitation by temperature variables can be achieved with a relatively good level of precision, for Iowa landscapes, using a simple linear regression method. Yield prediction within year across sites failed to produce useful results for any of the evaluated bioclimatic variables. Better yield predictions across sites within years should be achieved with improved precipitation estimates, more accurate georeferencing of test sites, knowledge of soil properties such as fertility, soil type and root zone available water holding capacity, and farmer management practices.

	ACKNOWLEDGMENTS

 
We thank the numerous cooperators for obtaining the data, the reviewers for their helpful comments and Daryl Travnicek for help with making the figure.

	NOTES

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussions Conclusion REFERENCES

 
Univ. of Nebraska, Agricultural Research Division. Publication Series No. 12178.

Received for publication January 5, 2000.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussions Conclusion REFERENCES

 

• Bondavalli B., Colyer D., Kroth E.M. Effects of weather, nitrogen and population on corn yield response. Agron. J. 1970;62:669-672.[Abstract/Free Full Text]
• Cooper, M., and D.E. Byth. 1996. Understanding plant adaptation to achieve systematic applied crop improvement-A fundamental challenge. In M. Cooper and G.L. Hammer (ed.) Plant adaptation and crop improvement. CAB International, Wallingford, UK.
• Corbett J.D. Classifying maize production zone in Kenya through multivariate cluster analysis. In: Hassan R.M., ed. Maize technology development and transfer: A GIS application for research planning in Kenya. Wallingford, UK: CAB International, 1998.
• Dale R.F., Daniels J.A. A weather-soil variable for estimating soil moisture stress and corn yield probabilities. Agron. J. 1995;87:1115-1121.[Abstract/Free Full Text]
• Dirks V.A., Bolton E.F. Regression analyses of grain yield of corn, level of leaf NPK and soil conditions in a long-time rotation experiment on Brookston clay. Can. J. Soil Sci. 1980;60:599-611.
• Dirks V.A., Bolton E.F. Climatic factors contributing to year-to-year variation in grain yield of corn on Brookston clay. Can. J. Plant Sci. 1981;61:293-305.
• Duchon C.E. Corn yield prediction using climatology. J. Clim. Appl. Meteorol. 1986;25(5):581-590.
• Hanks R.J., Gardner H.R., Florian R.L. Plant growth-evapotranspiration relations for several crops in the Central Great Plains. Agron. J. 1969;61:30-34.[Abstract/Free Full Text]
• Hillel D., Gubon Y. Relation between evapotranspiration rate and maize yield. Water Res. 1973;9:743-748.
• Johnson N.C., Copeland P.J., Crookston R.K., Pfleger F.L. Mycorrhizae: Possible explanation for yield decline with continuous corn and soybean. Agron. J. 1992;84:387-390.[Abstract/Free Full Text]
• Liang B.C., MacKenzie A.F., Kirby P.C., Remillard M. Corn production in relation to water inputs and heat units. Agron. J. 1991;83:794-799.[Abstract/Free Full Text]
• Lin C.S., Binns M.R., Lefkovitch L.P. Stability Analysis: Where do we stand?. Crop Sci. 1986;26:894-900.[Abstract/Free Full Text]
• Muchow R.C., Sinclair T.R. Effect of nitrogen supply on maize yield: II. Field and model analysis. Agron. J. 1995;87:642-648.[Abstract/Free Full Text]
• Newhall F., Berdanier C.R. Calculation of soil moisture regimes from the climatic record. Washington, DC: U.S. Gov. Print. Office, 1992 USDA-SCS.
• Nielsen D., Peterson G., Anderson R., Ferreira V., Shawcroft W., Remington K. Estimating corn yields from precipitation records. A method for assessing adapted environments and production risk. Conservation Tillage Fact Sheets No 2-96. Sterling, CO: USDA-ARS, USDA-NRCS, and Colorado Conservation Tillage Association, 1996.
• Runge E.C.A. Effects of rainfall and temperature interactions during the growing season on corn yield. Agron. J. 1968;60:503-507.[Abstract/Free Full Text]
• Soil Survey Staff. Soil Taxonomy. Washington, DC: U.S. Gov. Print. Office, 1975 USDA Soil Conservation Service, Agricultural Handbook No. 436.
• Teigen, L.D. 1991. Weather, climate and variability of U.S. corn yield. Feed Sit. Outlook Rpt. Fds. 320. U.S. Dept. Agric., Econ. Res. Serv. pp. 28–37.
• Teigen, L.D., and M. Thomas, Jr. 1995. Weather and yield, 1950–94 relationships, distributions, and data. ERS Staff Paper. Econ. Res. Serv. Com. Agric Div. NO 9527.
• USDA Natural Resources Conservation Service. Soil Taxonomy: A basic system of soil classification for making and interpreting soil surveys. Second Edition, USDA Agriculture Handbook 436. Washington, DC: US Gov. Print. Office, 1999.
• Van Wambeke A. Calculated soil moisture and temperature regimes of South America. Washington, DC: Soil Management Support Services Technical Monograph No. 2, USDA-SCS, 1981.
• Van Wambeke A. Calculated soil moisture and temperature regimes of Africa. Washington, DC: Soil Management Support Services Technical Monograph No. 3, USDA-SCS, 1982.
• Van Wambeke A. Calculated soil moisture and temperature regimes of Asia. Washington, DC: Soil Management Support Services Technical Monograph No. 9, USDA-SCS, 1985.
• Van Wambeke A., Hastings P., Tolomeo M. Newhall Simulation Model-A basic program for the IBM PC (DOS 2.0 or Later). Ithaca, NY: Dep. of Agronomy, Cornell University, 1992.
• Voss R.E., Hanway J.J., Fuller W.A. Influence of soil, management, and climatic factors on the yield response by corn (Zea mays L.) to N, P, and K fertilizer. Agron. J. 1970;62:736-740.[Abstract/Free Full Text]
• Waltman W.J., Ciolkosz E.J., Mausbach M.J., Svoboda M.D., Miller D.A., Kolb P.J. Soil climate regimes of Pennsylvania. Bulletin No. 873. University Park, PA: Pennsylvania State University Agricultural Expt. Station, 1997.
• Yamoah C.F., Varvel G.E., Francis C.A., Waltman W.J. Weather and management impact on crop yield variability in rotations. J. Prod. Agric. 1998;11:219-225 a.
• Yamoah C.F., Varvel G.E., Waltman W.J., Francis C.A. Long-term nitrogen use and nitrogen-removal index in continuous crops and rotations. Field Crops Res. 1998;57:15-27 b.

This article has been cited by other articles:

				A. J. Miller and J. H. Knouft GIS-based characterization of the geographic distributions of wild and cultivated populations of the Mesoamerican fruit tree Spondias purpurea (Anacardiaceae) Am. J. Botany, December 1, 2006; 93(12): 1757 - 1767. [Abstract] [Full Text] [PDF]

				E. A. Lee, T. K. Doerksen, and L. W. Kannenberg Genetic Components of Yield Stability in Maize Breeding Populations Crop Sci., November 1, 2003; 43(6): 2018 - 2027. [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 Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (2) Citing Articles via Google Scholar Google Scholar Articles by Jeutong, F. Articles by Smith, O.S. Search for Related Content PubMed Articles by Jeutong, F. Articles by Smith, O.S. Agricola Articles by Jeutong, F. Articles by Smith, O.S.