Future Global Soybean Yield Projections

This paper introduces the relationships between climate conditions and global scale soybean yields and their future projections until 2050. Using a combination of historical datasets and climate model projections, yield responses to temperature and precipitation change were discussed together with input of fertilizer. Future climate projection is based on downscale dataset from IPCC, A1 scenario. Datasets used in this study include spatial information but not yield data. To improve the accuracy of estimation, the specific vegetation types were used to obtain climate conditions on crop area. The relationships between soybean yield and climate conditions on crop area were conducted for soybean producing countries and future global soybean yield until 2050 are discussed.


Introduction
Economic development and population increase will result in shortage of food (Matsumura, 2011).Climate change influences crop yield and casts a shadow on food production.In 2010 summer, India and Nepal suffered from heavy rain, Italy was more humid and the USA was hot summer.Russia announced that crop production in 2010 went down 26 % from previous year and stopped exporting crop due to extreme climate event.These facts suggest that it is important to estimate the impact of changes in climatic condition on major crop yield such as rice, wheat, soybean, and maize.Table 1 suggests the top 5 major crop producing countries in the world.This paper introduces the development of methodology and the relationships between yield and climate conditions on crop for soybean producing countries.
Rice plays a crucial role in the supporting continued global population growth.World rice production in 2007 was approximately 645 million tons, with Asian farmers producing 90% and China and India producing 50% of the global rice supply (Kawashima, 2008).Animal nutritionists have observed that combining one part soybean meal with four parts grain dramatically boosts the efficiency of livestock and poultry; this converted grain into animal protein.Soybean occupies more U.S cropland than wheat (Lester R. Brown, 2009).
There are several researches related to climate conditions and crop yield.Data on national crop yields for 1961-2002 in the ten leading producers of wheat, rice and maize were combined with datasets on climate and crop locations to evaluate the empirical relationships between average temperature and the difference between daily maximum and minimum temperature and crop yields (David B, 2007).
The author's previous study analyzed the relationships between country based precipitation and crop yield (Matsumura, 2010).As a preliminary research, the author conducted the relationships between crop yield, precipitation and temperature for major top 5 crop producing countries using country averaged datasets.The multiple correlation coefficients were 0.245 for Maize (USA), 0.328 for Soybean (USA), 0.16 for Wheat (USA) and 0.185 for Rice (China).The correlation coefficient value is not so high.To improve the accuracy of estimation, the specific vegetation types were used to obtain climate conditions on crop area.
Figure 1 shows the comparison of country averaged precipitation and precipitation on crop area in USA.The comparison shows that the fluctuations of precipitation on crop area is less small compared with country averaged precipitation.This tendency can be seen in other countries.

Datasets
The datasets used to obtain the relationships between climate conditions and crop yield are introduced in 2.1 and future projections of down scaled climate conditions including data processing methodologies are introduced in 2.2.Datasets used in this study included spatial information but not yield data.

Combining data sets
Datasets with spatial information such as temperature, precipitation, vegetation, and cropping calendar were used in this study and country based major crop yields of wheat, soybean, rice, and maize were obtained from Food and Agricultural Organization (FAO-STAT, 2010).Crop yield dataset covers 236 countries and 5 areas such as Africa, Americas, Asia, Europe, and Oceania but not with spatial information.Crop yield dataset and fertilizer data set are provided b FAO and covers from 1961 to 2009, from 1961 to 2002 respectively.The vegetation data-set developed by a global mapping project (GSI, 2008) is used to highlight the vegetation type especially sparse vegetation, cropland, paddy field, and cropland/other vegetation mosaic.The grid size is 30 seconds by 30 seconds (often referred to as 1-km spatial resolution).
Monthly average temperature and precipitation datasets from January 1961 to December 2006 are obtained from the Climatic Research Unit, University of East Anglia (CRU, 2010) shown in Figure 4. Using vegetation dataset, temperature and precipitation on crop area are extracted and making use of boundary dataset (ESRI, 2008), country averaged precipitation and temperatures on crop area were obtained.
Multiplying highlighted crop area by temperature and precipitation datasets, temperature and precipitation on crop land can be extracted.Using of country boundary dataset (ESRI, 2008), precipitation and temperature on crop land are recalculated as an average value in for 186 countries Digitizing and geo-referencing the existing observations on crop planting and harvesting dates yielded the Cropping Calendar datasets (Cropping Calendar, 2009).The cropping calendar consists of 5 components: planting, vegetating, heading, filling, and harvesting.On the basis of the cropping calendar, the life cycle of major crops, including the five components in each country, is obtained.Overall conceptual framework to generate datasets for estimation is shown in Figure 6.

Developing downscaled projection data
The outputs of Global Climate Model (GCM) are quoted in IPCC, AR4 report and used for future global scale soybean projection.The outputs are based on IPCC SRES scenarios A1.To apply the future results for estimated equation listed in Table 2, it is required to downscale the future projection outputs.In another world, the grid size of GCM is 100 km and downscaled projection data will be required.Each GCM uses a scenario defined "20C3M" (20th Century Climate Coupled Model) to verify the reappearance.The parameters in each model are different and bias should be extracted to obtain downscaled projection.Following procedures are conducted to obtain downscaled monthly average temperature and precipitation.
(STEP1) The average temperature and precipitation datasets of CRU TS 2.10 in each month from 1971 to 2000 are calculated.
(STEP2) The GCM output's average from 1971 to 2000 is calculated and imposed in 0.5 degree spatial dataset.
(STEP3) The GCM outputs based on SRES scenarios in 2010, 2020, 2030, 2040 and 2050 are obtained and imposed in 0.5 degree spatial dataset.(STEP4) The ratio of precipitation and temperature processed in STEP2 and STEP3 are calculated.The relationship "Kelvin=Celsius+273.15" is used.Bilinear Interpolation methodology is used.The value in Grid T (x, y) is calculated by nearby 4 points following equation.
x y y The bilinear interpolation methodology, original GCM outputs, imposed values are shown in Figure 7.The example of imposed datasets is shown in Figure 8.The figure expresses the average monthly temperature in June from 2041 to 2050.

Multiple regression analysis
Country based major crop yields of wheat, soybean, rice, and maize were obtained from Food and Multiplying highlighted crop area by temperature and precipitation datasets, temperature and precipitation on crop land can be extracted.Using of country boundary dataset (ESRI, 2008).Precipitation and temperature on crop land are recalculated as a country based average value.Aggregating temperature and precipitation data based on cropping calendar, the aggregated precipitation and temperature datasets are obtained.Figure 9 shows the structure of a generated dataset in China.The generated datasets cover 186 countries and four major crops.Each country covers four major crop yield, monthly temperature, precipitation, fertilizer per area and cropping calendar information.Cropping calendar in includes plant_start, plant_range, plant_end, harvest_start, harvest_range, harvest_end, total days with spatial information.Using of country boundary dataset, cropping calendar information is also recalculated as a country based average value.If plant starts in July and harvest ends in September, temperature and precipitation datasets are aggregated from July to September.If plant starts in October and harvest ends in March, temperature and precipitation datasets are aggregated from October to next year's March.
The relationships among soybean yield, aggregated temperature, aggregated precipitation, fertilizer per agricultural area were obtained by using multiple regression analysis.Case A used temperature and precipitation.Case B used temperature, precipitation, and fertilizer.
Applying actual monthly temperature and precipitation dataset for obtained equation listed in Table 2, the actual value and calculated value of yield in each country are then obtained.Figure 10 shows the soybean yield of actual, calculated value in case A and case B in USA as an example.
In general, as for crop yield, if temperature and precipitation increase, crop yield is expected to increase.Table 2 shows the opposite result in some countries.
IPCC reports that precipitation changes occur more spatially and seasonally in the eastern parts of North and South America, northern Europe, and northern and central Asia become significantly wet.However, the Sahel, the Mediterranean, southern Africa, and parts of southern Asia become drier (IPCC, 2007) Attention should be paid to some places where the temperature drops below zero, the frost formed provides water for the soil.For those countries where infrastructures are not well equipped, temperature plays an important role for providing water on crop.

Future projections of global soybean yield
Future monthly temperature and precipitation datasets are obtained from 2010 to 2050 by every 10 years.Using same methodology mentioned above, future monthly temperature and precipitation are applied for obtained equations in Case A. As for using future projections of fertilizer is an issue of concern.Adjusting actual value in 2009 and calculated value in 2010 in same value and global scale future projections of soybean yields are obtained."+" suggests that soybean yield will be increased and "-" suggests that soybean yield will be decreased.

Conclusion
The relationships among yield and climate conditions on crop area were conducted for soybean producing countries and the influence of temperature, precipitation, and fertilizer were obtained.It is well known that temperature and precipitation increase result in increase of yield.
There are several countries that conflicts with well known fact.There are some countries where temperature increase results in decrease of yield.Also, attention should be paid to some places where the temperature drops below zero, the frost formed provides water for the soil.For those countries where infrastructures are not well equipped, temperature plays an important role for providing water on crop.
According to William C. ( 2007), country-level climate projections were conducted, and a consensus that warming would continue in a similar pattern up to the 2080s was obtained in the general circulation model (GCM) climate projections.
IPCC reports that precipitation changes occur more spatially and seasonally in the eastern parts of North and South America, northern Europe, and northern and central Asia become significantly wet.However, the Sahel, the Mediterranean, southern Africa, and parts of southern Asia become drier (IPCC, 2007).
Those data used in this study have spatial information but yield data.If yield information at targeted area can be obtained, the relationships among yield, temperature and precipitation can be obtained at targeted area.The authors are working on developing database which can provide time series temperature and precipitation data at targeted area.
The authors are working on estimating the relationships for other major crops such as wheat, rice, and maize and future projections of major crop yield.In this paper, linear multiple regression analysis is simply used in this paper, but for the future challenges, non linear multiple regression analysis is expected to conduct.
The methodology introduce in this paper can be useful educational tools and authors are planning to organize workshop for foreign students from all over the world.

Table 2 .
The results of world soybean producing countries in Case A and Case B

Table 2 .
The results of world soybean producing countries in Case A and Case B (Continued)