| Back to Home Page | Back to Institute Activities |
| Monday, July 19 | Overview and Orientation |
| Tuesday, July 20 | Analyze Historical Climate Data |
| Wednesday, July 21 | Analyze Historical Crop Data |
| Thursday, July 22 | Simulate Crop Yields and Analyze Results |
| Friday, July 23 | Identify Optimal Strategies |
| Monday, July 26 | Identify Optimal Strategies (continued) |
| Tuesday, July 27 | Explore Strategies with "Farmers" |
| Wednesday, July 28 | Summarize and Prepare for Presentation |
| Thursday, July 29 | Prepare presentation of results |
| Friday, July 30 | Present Results |
Monday, July 19
On a worldwide scale, the El Niño-Southern Oscillation (ENSO) is the most important single predictor of interannual climate variability. When properly analyzed and interpreted, ENSO phases are a simple form of seasonal climate forecast. They are also a useful proxy for other forms of climate forecast when developing and testing methodology for climate prediction application.
You are part of an international, collaborative research project that has the goal of applying seasonal climate forecasts to improve agricultural production in southeastern South America. Your strategy is to introduce ENSO-based field crop strategies to a pilot group of farmers near Pergamino with the hope that lessons learned there will be transferrable to other locations.
You will spend a relatively short time today forming teams and becoming familiar with the software and data that you will use in the mini-project.
a. Form teams of two or three. Team members will share a computer and work together on the computer exercises. Select a computer.
b. Review the exercises and schedule. Acquaint yourselves with the format and location of reading and supplemental materials. Begin to discuss whether your team will work on the field-scale crop management problem or the farm-scale crop mix problem on Friday and next Monday.
c. You will be working intensively with data manipulation, statistical analyses, and crop models throughout the week. Examine the program and software directories that will be used for the mini-project. We have provided more than one word processing, spreadsheet and statistics package to accommodate the preferences of different participants (and instructors).
d. Examine the spreadsheet files in the data directory. Select the statistical analysis software that you want to use. Learn how to transfer data between text files, spreadsheet software, and statistical software.
Hall, A. J., C. M. Rebella, C. M. Ghersa, and J.P. Culot. 1992. Field crop systems of the Pampas. P. 413-450. In C.J. Pearson (ed) Field Crop Ecosystems of the World. Elsevier, Exeter, UK.
Podestá, G., D. Letson, J. Jones, J. Hansen, J. O'Brien, and D. Legler. 1999. Regional application of ENSO-based climate forecasts to agriculture in the Americas.
Tuesday, July 20
Distributions of historical climate data associated with ENSO phases (i.e., warm or El Niño, neutral, and cold or La Niña) represent one useful form of seasonal climate forecast. ENSO phases are also a useful proxy for other forms of climate forecast; tools and methods for using ENSO phase information can be adapted for more refined forecasts in other formats as they become available.
Exploratory and statistical analyses of historical climate data can indicate the influence of ENSO phases on agriculturally-relevant climate variables. You will use simple graphical and statistical methods to characterize the influence of ENSO on historical monthly climate data for Pergamino, Argentina. To provide a complete data series, monthly mean temperatures and precipitation totals extracted from Pergamino daily data were averaged with monthly data from the Global Historical Climatology Network for nearby Junin. Although we will use ENSO phases based on the Japan Meteorological Agency (JMA), you may wish to compare results with other indices of ENSO or other climatic influences.
Experience with simple, relevant univariate statistical tests and graphical displays of association between historical climate data and categorical indices of ENSO.
Understanding of the influence of ENSO on Pergamino's climate as preparation for the remainder of the mini-project.
a. Calculate and plot long-term monthly means by ENSO phase for each climate variable. Monthly data are provided in \DATA\CLIMATEDATA.___ (WB3 or XLS).
b. Based on the crop calendar, identify critical climate variables and periods for crop response.
c. Analyze ENSO influence on the important climate variables and periods identified in step B.
1. Use analysis of variance (ANOVA) to test the influence of ENSO phases on climate variables.
2. Identify which ENSO phases are different in their effects based on Duncan's Multiple Range Test (DMRT).
3. Test that the assumptions of normality of residuals and homogeneity of variances required by ANOVA are reasonably met.
4. If tests show serious violations of ANOVA assumptions, then analyze the same data using the non-parametric Kruskall-Wallace (K-W) rank ANOVA.
5. When you find significant ENSO phase effects, prepare quartile box plots for each phase.
Test another climatic classification of years. The spreadsheet, ENSO.___ (WB3 or XLS) contains a list of years classified by each of several definitions of ENSO phases. If time permits, compare results obtained by the JMA definition of ENSO phases with ENSO phases by one of the other definitions.
Trenberth, K. 1997. The definition of El Niño. Bull. Amer. Meteor. Soc. 78:2771-2777.
Wednesday, July 21
Where they are available, historical crop data at a reporting district scale can provide a valuable perspective of the possible influence of ENSO (or other indices of predictable components of climate variability) on crop production. However, historical crop data integrate the effects of climate variability, changes in soil types associated with changes in land use, improvements in technology, and price cycles. Some processing is needed to separate the effects of interannual climate variability from the effects of these other factors that tend to change more slowly. Hansen et al. (1988) discusses this as a signal filtering problem. You will apply similar analyses to the de-trended yields that you applied to historic climate data yesterday. We will focus on maize yields, although other crops and variables are provided.
a. Plot the historic time series of Pergamino maize yields (CROPDATA.WB3 or CROPDATA.XLS), with different symbols for each ENSO phase. The spreadsheet file, ENSODATA.___ (WB3 or XLS), contains a template for such graphs. Look for any obvious patterns.
b. Detrend the Pergamino maize yield series.
1. Fit the observed yields (yO,t) to a smoothing trend line (yS,t). Loess smoothing in S-Plus and a spectral smoothing filter (\SMOOTH\SMOOTH.EXE) are available. Use a filtering period of > 7 years to avoid removing substantial ENSO-related variability with the trend. Add the trend to the graph developed in step A.
2. Select a response variable yt. If yield variability (i.e., the standard deviation about the trend) seems to increase in proportion to changes in the trend, then use the ratio of observed to smoothed (i.e., yt = yO,t / yS,t). Otherwise, use anomalies (i.e., yt = yO,t - yS,t). What is the standard deviation of the response variable?
3. Analyze the response of yt to ENSO phases by ANOVA and/or K-W, and DMRT as you did for climate data yesterday. Test assumptions of ANOVA. Print box plots if results are significant.
c. Repeat steps A and B for national maize yields. Use the same response variable that you used in step B2 to allow comparison with Pergamino district yields. Do both show the same trend? Which shows greater interannual variability? Which is more sensitive to ENSO-related climate variability?
a. Propose and evaluate an alternative method of accounting for time trends.
b. Analyze another maize variable (e.g., production, or the ratio of harvested to planted area) or another crop.
Bell, M.A. and R.A. Fischer. 1994. Using yield prediction models to assess yield gains: a case study for wheat. Field Crops Research 36:161-166.
Hansen, J.W., A. Hodges, and J.W. Jones. 1998. ENSO influences on agriculture in the southeastern United States. Journal of Climate 11:404-411.
Hansen, J.W., J.W. Jones, A. Irmak, and F.S. Royce. ENSO impacts on crop production in the Southeast US. Invited paper presented at the American Society of Agronomy Symposium, Impacts of Climate Variability on Agriculture.
Podestá, G.P., C.D. Messina, M.O. Grondona, and G.O. Magrin. Associations between grain crop yields in central-eastern Argentina and El Niño-Southern Oscillation. Journal of Applied Meteorology (In press).
Thursday, July 22
Simulation models are an essential tool for analyzing the interactive effects of climate variability and management decisions on agricultural production. Crop models can reveal the impacts of climate variability on yields at a given site under given management without confounding effects such as technology trends. To be used with confidence, crop model results need to be evaluated carefully against real yield data in the region of interest. Although proper validation is beyond the scope of this mini-project, you will look at how simulated maize yields compare with reporting district-scale yields.
IMPORTANT NOTE: The daily weather data that you have were filled-in with a stochastic generator to provide a complete series to simplify running the crop models. Before applying any graphical or statistical analyses, you must discard simulated crop data for years (1955-1957, 1964-1967 harvests) with substantial gaps in the raw weather data during the growing season.
a. Examine the maize model input file (\DSSAT35\MAIZE\ARPE3201.MZX). List the import management decisions specified in this file.
b. Simulate maize yields using the default management.
c. Import output file (\DSSAT35\MAIZE\SUMMARY.OUT) into a spreadsheet. Remember to discard simulated crop data for years with weather data problems. Plot yield time series, with different symbols for each ENSO phase.
d. Compare simulated yields with reported Pergamino district maize yields. How do the patterns of central tendency and variability compare? Calculate and compare means, standard deviations and CVs of historical and simulated yields for the last 10 years. Plot the last 10 years of both on the same graph. Calculate RMSE:
[ RMSE formula to be inserted here]
(subscripts indicate yields predicted by the model, and observed in the district). Based on these results, would you promote the use of the CERES maize model to derive recommendations tailored to ENSO phases?
e. Analyze simulated yield response to ENSO phases. Use the same procedure that you used for historical climate data.
a. Apply the same smoothing technique and transformation that you applied to historical maize yields yesterday. Prepare a predicted vs. observed scatter plot. Calculate the linear correlation, slope and intercept of the relationship. Select and apply an appropriate test of the hypothesis that the crop model responds to interannual climate variability in the same way as district maize yields.
b. Test sensitivity of the mean and standard deviation of yields to either planting date, planting density, N fertilizer strategy or irrigation. Does the default, representative management strategy appear to be optimal on the average?
Boote, K.J., J.W. Jones, G. Hoogenboom, and N.B. Pickering. 1998. The CROPGRO model for grain legumes. p. 99-128. In G.Y. Tsuji, G. Hoogenboom, and P.K. Thornton (ed.) Understanding options for agricultural production. Kluwer, Dordrecht, The Netherlands.
Ritchie, J.T. 1998. Soil water balance and plant water stress. p. 41-54. In G.Y. Tsuji, G. Hoogenboom, and P.K. Thornton (ed.) Understanding options for agricultural production. Kluwer, Dordrecht, The Netherlands.
Ritchie, J.T., U. Singh, D.C. Godwin, and W.T. Bowen. 1998. Cereal growth, development and yield. p. 79-98. In G.Y. Tsuji, G. Hoogenboom, and P.K. Thornton (ed.) Understanding options for agricultural production. Kluwer, Dordrecht, The Netherlands.
Friday, July 23 and Monday, July 26
Seasonal climate forecasts will benefit individuals and society only when they serve as a basis for improved decisions. You can chose to work with either field-scale maize crop management decisions, or with farm-scale land allocation decisions.
Crop models allow a range of options to be examined relatively quickly for a given set of past or expected weather conditions. However, the number of combinations quickly becomes unmanageable when considering several management variables simultaneously. An adaptive simulated annealing (ASA) optimization algorithm linked to the DSSAT family of crop models provides a robust tool for selecting management strategies that maximize expected economic returns. An alternative tool that performs a uniform grid search is useful for preparing data to plot yield or economic return response surfaces to two management variables at a time. Be forewarned: both procedures are slow. Plan on using batch files to set up optimizations to run overnight. You will use crop yields simulated with historical daily weather data for all available years and for each ENSO phase to calculated expected returns for each optimization iteration.
Although we hope to develop these optimization tools, into user-friendly software, the current versions are somewhat challenging. Please read the ASA-DSSAT Guide (Royce, 1998). Jim Hansen is familiar with the software, files and procedure.
a. Select either the ASA (MAIZE.EXE) or grid (MAIZE-GRID.EXE) optimization method. Grid optimization can be used to prepare response surface graphs for two decision variables (e.g., planting date and N amount). Adaptive simulated annealing is more efficient for optimizing more than two decision variables, or when required resolution is high (i.e., more than about 40 ¥ 40 combinations in the two-dimensional search space).
b. Examine the input files (IBSNAT35.INP, ASA_OPT, GRID_OPT, ALL_YRS.INP, WARM.INP, NEUTRAL.INP, COLD.INP) in \ASA-MAIZE. Examine RUNALL.BAT to see how files can be manipulated to run optimizations for each category of years in batch mode. One way to prepare for batch runs is to make multiple copies of ASA_OPT or GRID_OPT with the appropriate ENSO phases or category of years selected, and saved under other names.
c. Select management variables to optimize. Specify the range, initial values (ASA only) and resolution of each that you wish to optimize. Make sure that the range and initial values of all other management variables are fixed at the desired constant value.
d. Run the optimizations for all years and each ENSO phase.
e. Tabulate and summarize results. A template spreadsheet (\ASA-MAIZE\OPTVALUE.___) is provided. You will need the number of years in each ENSO phase, and mean yields and gross margins for all years and for each phase, for the fixed (i.e., optimized for all years) and flexible management (i.e., optimized for each ENSO phase) strategies. How do El Niño and La Niña shift optimal management? What is the average value (yield and financial) of tailoring optimal management to ENSO phases?
a. Repeat the optimization procedure, tailoring optimal management to another ENSO or climate index. You will not need to re-run the all-year optimization. Is the selected ENSO or climate index more or less valuable than the JMA ENSO index?
b. Use the grid search optimizer to optimize two decision variables (holding the others constant) at a time. Plot the response surfaces for each ENSO phase. Describe the response surfaces. How do their shapes change in El Niño and in La Niña years?
Royce, F.S. 1998. ASA-DSSAT Guide. University of Florida. Unpublished report.
Hansen, J.W., J.W. Jones, A. Irmak, and F.S. Royce. ENSO impacts on crop production in the Southeast US. Invited paper presented at the American Society of Agronomy Symposium, Impacts of Climate Variability on Agriculture.
Thornton, P.K. 1994. The value of information concerning near-optimal nitrogen fertilizer scheduling. Agricultural Systems 45:315-330.
Wilks, D.S. 1998. Forecast value: prescriptive decision studies. p. 109-145. In: R.W. Katz and A.H. Murphy (ed) Economic value of weather and climate forecasts. Cambridge University Press, Cambridge, UK.
Friday, July 23 and Monday, July 26
Seasonal climate forecasts will benefit individuals and society only when they serve as a basis for improved decisions. You can chose to work with either field-scale maize crop management decisions, or with farm-scale land allocation decisions.
Farmers in the region of Pergamino indicated that they could adjust allocation of land among crops in response to expected seasonal rainfall anomalies. You will use the nonlinear optimization feature in spreadsheet software to identify the crop mix (i.e., allocation of available land among crops) that maximizes either expected wealth or, for the risk-averse farmer, expected utility of wealth at the end of the growing season. As with the field-scale crop management optimization example, you will use crop yields simulated with historical daily weather data for all available years and for each ENSO phase to calculated expected wealth or utility. Carlos Messina is familiar with the crop mix optimization spreadsheet and procedure.
a. Examine farm size, production cost and crop price data in the spreadsheet.
b. In DSSAT, simulate yields of maize, soybean and sunflower using the crop management files provided. Copy yield simulation results from SUMMARY.OUT in each crop's data directory into the spreadsheet.
c. Using default settings, optimize crop mix for all years. Use a stacked bar graph to plot the optimized crop mix with the average relative areas planted to these three crops in the Pergamino district in recent years.
d. Optimize crop mix for each ENSO phase. Plot the optimal crop mix for each ENSO phase and for all years. Calculate the value of conditioning crop mix on ENSO phase.
e. Repeat steps D and E using both higher and lower coefficients of relative risk aversion.
f. Summarize results in a format that can be presented as recommendations to farmers.
a. Repeat the optimizations using crop prices from each of the last five years. How sensitive are optimal crop mix and potential forecast value to price fluctuations?
b. Repeat the optimizations, tailoring optimal crop mix to another ENSO or climate index. You will not need to repeat the all-year optimization. Is the selected ENSO or climate index more or less valuable than the JMA ENSO index?
c. Try replacing yields simulated with the default management with yields simulated with management optimized for each ENSO phase. What is the impact of optimizing both crop mix and crop management?
Messina, C.M., J.W. Hansen, and A.J. Hall. 1999. Land allocation conditioned on ENSO phases in the Pampas of Argentina. Agricultural Systems 60:197-212.
Wilks, D.S. 1998. Forecast value: prescriptive decision studies. p. 109-145. In: R.W. Katz and A.H. Murphy (ed) Economic value of weather and climate forecasts. Cambridge University Press, Cambridge, UK.
Tuesday, July 27
Models are never perfect representations of reality. Crops may respond to climate variability quite differently than crop models predict due to factors that the models do not consider, such as diseases or delayed harvest due to wet field conditions. Part of your task as a researcher is to recognize when models fail to capture, or misrepresent, aspects of system response that are critical to either interpretation or decision support application. Good farmers generally understand their own environments and production systems much better than researchers. Like all of us, they make decisions for reasons that may sometimes have little to do with technical or financial considerations.
Incorrect model predictions, technical constraints imposed by other components of the overall farming system, and subjective, sociological factors are all potential barriers to adoption of ENSO-based crop strategies. However, these barriers are not necessarily insurmountable. Clearly, interaction between forecast providers, forecast application researchers, and the farmer end users is essential to identifying and overcoming obstacles to the beneficial use of seasonal climate forecasts.
Three individuals who are familiar with cropping systems in the Pampas region of Argentina will role play as farmers. You will interview these farmers with the objective of evaluating the practical feasibility and acceptability of model-derived recommendations tailored to ENSO phases. Present whatever background information about the project or your results that you think is essential to the farmers. What is their perception of ENSO, climate prediction, your project, and your use of models? Present your model-derived crop management and crop mix recommendations to the farmers. What is their perception of the recommendations? Do they identify practical or sociological constraints to adopting the recommended strategies? Based on the farmer interactions, outline a plan for the remainder of the project.
Royce, F., S. Meira, and J. Jones. 1998. Current perceptions and potential uses of climate predictions by farmers of the humid Pampas. Unpublished report, University of Florida.
Wednesday, July 28
Thursday, July 29
Friday July 30
[ Return to top of page | Return to Institute Activities | Return to Home Page ]
Page Last Updated: July 18, 1999 Contact Information: Guillermo Podestá, Institute Science Coordinator e-mail: gpodesta@rsmas.miami.edu Telephone: 1.305.361.4142 FAX: 1.305.361.4622