ORIGINAL PAPER Crop yield model validation for Cameroon Munang Tingem & Mike Rivington & Gianni Bellocchi & Jeremy Colls Received: 5 October 2007 /Accepted: 7 April 2008 # Springer-Verlag 2008 Abstract A crop simulation model must first be capable of representing the actual performance of crops grown in any region before it can be applied to the prediction of climate variability and change impacts. A cropping systems model (CropSyst) simulations of crop productivity in the sub-Saharan Central African (using Cameroon as the case study) region, under the current climate were compared with observed yields of maize, sorghum, groundnut, bambara groundnut and soybean from eight sites. The model produced both over-and-under estimates, but with a mean percentage difference of only –2.8%, ranging from – 0.6% to –4.5%. Based on these results, we judged the CropSyst simulations sufficiently reliable to justify use of the model in assessing crop growth vulnerability to climatic changes in Cameroon and else where. 1 Introduction About 800 million people in the world are malnourished (World Bank 2007) and food production has to double in the next three decades to meet future needs (Thornton et al. 2006). These advances have to be achieved in the face of climate variability and change (IPCC 2007). The impacts on people’s livelihoods will be greatest in the tropics and subtropics, and particularly in the region of sub-Saharan Central Africa. Cameroon represents a good example of this region due to its bio-climatic range and socio-economic conditions. Here many poor small landholders depend on agriculture and have few alternative sources of income (Tingem et al. 2008). This strongly suggests the need for research and pilot studies to characterize and quantify the effects of climate variability and change at difference spatial scales, in order to assist farmers reduce the risks. Cameroon is chosen for this analysis due to its similarity to other countries in the region in terms of climate and crop diversity (Neba 1999; DeLancey and Mike 2000). About 45% of Cameroon’s gross domestic product originates from agriculture with close to 80% of the labour force employed in this sector (World Fact Book 2007). Most importantly, this sector is also responsible for providing food security to both the rural and urban populations from domestic production. However, this may not be true in the future. With a rapidly expanding population (ca. 17 million increasing by ca. 2%), the pressure on natural resources is mounting. Such considerations pose considerable challenges for food security and raise many questions: What will the impacts of climate change be at local as well as at national levels? Which research and development strategies are likely to help and how can they be appropriately targeted? Process-based crop models which are widely used by researchers from many different backgrounds can provide broad-brush approximations of the expected effects of climatic changes (Rosenzweig and Hillel 1998; Thornton and Jones 2003; Fischer et al. 2005). They use long-term weather data to account for weather variability in assessing risks involved with adopting alternative management strategies at a site of interest (i.e. Tingem et al. 2007). Nevertheless, before this is possible, models must be evaluated for each climatic region where they are intended for use in decision-making. Using them as supplied is dangerous, i.e. without sufficient validation being carried Theor Appl Climatol DOI 10.1007/s00704-008-0030-8 M. Tingem (*) : J. Colls Agriculture and Environmental Science Division, School of Biosciences, University of Nottingham, NG7 2RD Nottingham, UK e-mail: plxmrt@nottingham.ac.uk M. Rivington Macaulay Institute, Craigiebuckler, AB15 8QH Aberdeen, Scotland, UK G. Bellocchi Agrichiana Farming, Via di Sciarti 33/A, 53040 Montepulciano, Italyout for the sites at which they are applied. Testing and validation for locations other than those for which they were developed and validated is necessary. The cropping systems model CropSyst (Stöckle et al. 2003) was selected due to its robustness and relative ease of application, using commonly available information. It has been used to model the growth and development of several crops in many parts of the world, i.e. Mali (Badini et al. 2007), Burkina Faso (Badini et al. 1997), United Kingdom (Rivington et al. 2006) and Italy (Moriondo et al. 2007). Also CropSyst is credited with the capability to simulate the growth of many crops from a uniform structure and a common set of parameters. This represents an advantage over separate model representations of crops in simulating the productivity of tropical agricultural systems in which multi-and inter-cropping rather than mono-cropping is dominant. It also provides for simultaneous modelling of changes in crop environment including plant and soil moisture and nutrients, which constitute constraints of productivity of tropical agricultural systems. The objective of this research was thus to test the performance of CropSyst (version 3.04.08) in Cameroon (as an analogous representation of sub-Saharan Central Africa) by comparison with reported data, for a range of climate and food crops carried out under non-limiting conditions. Validation of this model will offer the opportunity to evaluate the effects of climate variability and change on crop yields which are impossible to assess at present due to limited validation of such models in the region. 2 Study area Cameroon is a tropical country located in the sub-Saharan region of central-west Africa. The country displays highly contrasting physical and biogeographical features. The climate, reflecting the topography and latitudinal range is very diverse. It comprises two principal climate zones: the equatorial zone and the tropical zone. The equatorial zone stretches from 2 to 6°N covering the southern and the mountainous western part of the country. Its climate corresponds to the classical Guinean region, with the following subtypes: (1) the seaboard, e.g. Kribi and Tiko with abundant rainfall (2,634 and 3,198 mm year–1, respectively); (2) the inland areas, e.g. Yaounde with total rainfall <1,660 mm year–1, prevailing over the southern part of the southern Cameroon plateau and extending into the east of the country around Batouri; (3) north of 6° N, where the Sudanese-Sahelian subtype differs from the ‘inland’ with total rainfall decreasing from 1,513 to 834 mm year–1. Annual average temperature across the country varies between 20 and 29°C, and in the extreme north, daily temperatures are usually between 25 and 34°C. The humid equatorial zone in the south favours the cultivation of cash crops such as palm oil, bananas, cocoa, rubber, plantains, and coffee. Across Cameroon, the key food crops are maize, groundnut, sorghum, bambara groundnut and soybean. The semi-arid region to the north mostly favours the growth of millet, sorghum, maize and groundnuts. The growing season is related to the rainy season and planting is fine-tuned to specific times of the year. In the equatorial zone comprising Bamenda, Batouri, Kribi, Tiko and Yaounde, there are two rainy seasons: the first is the ‘long’ one, being March–July with planting in March; and the second is the ‘short’ one, from August to November. The tropical zone includes Garoua, Maroua and Ngaoundere, where there is only one rainy season, from May to October and planting starts in May (Ndemah 1999). Table 1 shows geo-location of the studied sites and in the Appendix, descriptions of the regional climatic differences within Cameroon identify the climatic spatial variation (see Appendix). 2.1 Crop growth model CropSyst (Stöckle et al. 2003), a multi-year, multi-crop, daily time-step cropping-system simulation model has been applied and used extensively to simulate crop growth and yield for a range of crops such as wheat, maize, soybean, sorghum, groundnut, and forage crops in diverse environments. The model simulates the soil-water budget, the soilplant nitrogen budget, crop canopy and root growth, crop phenology, dry matter production, yield, residue production and decomposition, and erosion. The main inputs are daily weather data (precipitation, maximum and minimum temperature, and solar radiation) with the model allowing the user to specify management options. These include the timing of events such as sowing, organic and inorganic nitrogen fertilizer applications (and rates), tillage, etc. Crop physiology is determined by cultivar specific coefficients controlling canopy and root growth and development (e.g. those determining leaf respiration, plant resource partitioning and phenological development). Soil profile properties Table 1 Geo-references of the eight agricultural study sites with annual rainfall Location Latitude Longitude Elevation (m) Annual rainfall (mm) Bamenda 6.05° 10.1° 1,239 2,378 Batouri 4.47° 14.37° 656 1,499 Garoua 9.33° 13.38° 244 1,090 Kribi 2.95° 9.89° 16 2,634 Maroua 10.44° 14.25° 422 834 Ngaoundere 7.34° 13.57° 1,104 1,514 Tiko 4.08° 9.37° 52 3,198 Yaounde 3.83° 11.51° 760 1,655 Theor Appl Climatol(soil texture, thickness, initial water and nitrogen content) can be specified, with well-established pedo-transfer functions being utilized to derive values where observed data are not available. Crop growth is simulated for the whole canopy and roots by calculating unstressed (potential) biomass based on crop potential transpiration and on intercepted photosynthetically active radiation. This potential growth is then corrected by water and nitrogen limitations to determine actual daily biomass gain. The simulated yield is then obtained as the product between actual aboveground biomass accumulated at physiological maturity and a crop-specific harvest index (harvestable yield/aboveground biomass). The simulation of crop phenological development is based on the thermal time required to reach specific development stages. Thermal time is calculated as growing degree days (GDD, °C-days) accumulated throughout the growing season (starting from planting until physiological maturity). Average air temperature above a base and below a cut-off temperature is considered for GDD calculations. The accumulation of thermal time may be accelerated by water stress due to reduced cooling effects by transpiration from the canopy. Water-balance processes in CropSyst include rainfall input, runoff, and interception by the canopy and surface residue, infiltration and redistribution within the soil profile, crop transpiration and soil evaporation. Water at the bottom of the soil profile is available to be drained. In this study, reference evapotranspiration was estimated by the Priestley and Taylor (1972) method. A finite difference soil-water balance function, by which water moves up and down depending on the soil water potential of vertically adjacent layers, was used for the redistribution of water in the soil under non-limiting soil fertility (Richards 1931). CropSyst has data requirements that can be reasonably met and provides support utilities to fill in missing inputs based on well-established procedures (e.g. pedo-transfer functions to derive soil hydraulic parameters). For this reason, it provides a conceptually unified modelling system for many crops, minimizing the dangers of structural uncertainty in making both cross crop and inter-spatial comparisons (Rivington et al. 2006). As such, it is able to well represent the variation in yield determined by weather-driven environmental conditions and respond to specific management regimen. 3 Methods The study regions chosen represent the variety of agricultural landscapes of Cameroon. Changes in crop yields were evaluated for five crops: maize (Zea mays L.), sorghum (Sorghum bicolour L.), groundnut (Arachis hypogaea L.), bambara groundnut (Vigna subterranea L. Verdc) and soybean (Glycine max (L.) Merr). The choice of these crops was based on the availability of observed yield data for validation purposes, and the relative importance of these crops to the subsistence farmer/impoverished community. They give a representative view of crop production potential in Cameroon. Representative soil properties (thickness and texture) for each of the simulation points were extracted from the International Soil Reference and Information Center data base (http://www.isric.nl) (Batjes 1995). Agronomic data (e.g. yield, phenological observations) were obtained from the Central Bureau of Statistics published district reports (AGRISTAT 2001) and the Institute of Agricultural Research-Cameroon (through http://www.wisard.org). Daily observed values of maximum and minimum temperatures, and rainfall were obtained for 1979–2003 from the University Cooperation for Atmospheric Research (UCAR; http://dss.ucar.edu/datasets/) for each of the eight sites used in the study. For each region, the data from one of the major weather stations were chosen as representative of the climate of that region. For the purpose of evaluating long-term effects of climate variability on crop yields, the temporal range of the weather data for use in the crop model was expanded so as to allow a good estimation of the probability of extreme events. The weather generator ClimGen (version 4.1.05; http://www.bsyse.wsu.edu/clim gen) was used to produce daily climatic data to supplement the 1979–2003 UCAR data, creating a 50-year data set. Information on ClimGen is given in Stöckle et al. (2003). ClimGen had previously been tested in Cameroon (Tingem et al. 2007) and found to perform well in simulating the range of precipitation and temperature values at all sites. Results of this evaluation are given in Table 2 for the probability levels (p-value) calculated by the t-test and Ftest for the yearly means, variance and percent difference (negative values show model under-estimation). In the current work, CropSyst was applied to estimate final potential crop (maize, sorghum, groundnut, bambara groundnut and soybeans) yields using the 50-year combined observed and generated weather data. Parameterization of CropSyst was based on typical values obtained from the literature, default values from the CropSyst user manual and based on the authors personal experience. A number of parameters accounting for cultivar-specific differences were calibrated based on generic observed development and growth characteristics. Further parameterization was achieved by minimization of differences between actual and simulated yields for a limited number of simulation trials using available field reported data. The calibration of the phenological parameters (e.g. GDD) was made using data provided by the Institute of Agricultural Research (IRA-Cameroon). Parameterization was conducted on the assumption that if the model could be shown to perform well under limited parameterization effort, then an increase in calibration data availability and parameterization effort will improve performance. Overall, the parameterization effort was restricted reflecting the limited Theor Appl Climatoldata resources that are typically available for model applications in the sub-Saharan region. Simulations were run with sowing dates set to 15 March, corresponding to the 74th day of the year (DOY), in Bamenda, Batouri, Kribi Tiko, and Yaounde. In Garoua, Maroua and Ngaoundere, the sowing date was set to 15 May (day of year 135) which agrees with traditional crop management in each area (Molua 2003; Ndemah 1999). A 1-m soil depth was used to simulate the soil-water balance, because it corresponds to typical observed maximum crop root length (Farre 1998). Forty per cent of crop residue was assumed to remain in the field after harvest for recycling purposes (Abraha and Savage 2006). No irrigation was used as this is not a common practice in Cameroon. In order to verify the applicability of the CropSyst model to the selected agricultural regions, and to ensure the reliability of its results, an evaluation of the simulated crop yields was conducted. This was done for each crop and agricultural region by comparing the averages of the simulated and observed (AGRISTAT 2001) yields. Evaluation was based on the relative difference (RD, %) between observed and simulated yields with baseline data (see Table 3). 4 Results 4.1 Evaluation of modelling performance The results, presented in Table 3, are expressed as the relative percentage difference between average simulated and observed yields. The model produced both over-and Table 3 Relative difference (RD; %) between observed yields and yields simulated with baseline climate data for five crops; RD = [(Observed– Simulated)/Observed] × 100, average over 50 years. NA no data available Region Crops Bambara Groundnut Maize Sorghum Soybean RD (%) RD (%) RD (%) RD (%) RD (%) Bamenda –5.4 –0.5 0.1 NA –2.1 Batouri NA –0.1 –1.6 NA NA Garoua –0.1 –0.03 –0.1 0.2 NA Kribi NA –0.7 –0.9 NA NA Maroua –0.8 –9.3 0.4 –6.0 NA Ngaoundere –4.6 –13.0 –0.7 –1.7 2.2 Tiko –5.5 0.6 –1.9 NA –10.5 Yaounde –2.1 –4.7 –1.5 NA NA Table 2 Results of the statistical tests showing the comparison of the observed precipitation (mm) and maximum temperature (°C) annual means and variances with those of 25-year synthetic data generated by ClimGen at eight sites Bamenda Batouri Garoua Kribi Maroua Ngaoundere Tiko Yaounde Precipitation Obs. mean 195.8 123.3 83.1 219.5 65.9 124.7 266.5 135.7 Obs. variance 2,5185 5,376 8,426 22,020 6,843 12,638 38,782 8,083 Gen. mean 176.1 134.7 88.3 72.3 72.3 126.1 260.3 153.6 Gen. variance 1,1202 8,506 7,601 7,647 7,646 11,875 41,799 7,594 % difference –10.1 9.3 6.3 –67.1 9.7 1.1 –2.3 13.2 p-value for t-test 0.724 0.74 0.888 0.838 0.854 0.976 0.94 0.624 p-value for F-test 0.097 0.229 0.434 0.417 0.429 0.46 0.452 0.46 Tmax Obs. mean 23.8 29.6 34.9 30.0 34.5 29.0 30.0 28.4 Obs. variance 2.93 2.50 9.27 2.05 8.81 4.00 3.25 2.47 Gen. mean 24.7 29.1 34.3 30.0 33.5 28.1 30.1 27.8 Gen. variance 2.96 3.35 8.89 2.00 8.08 3.83 3.30 2.33 % difference 3.8 –1.7 –1.7 0.0 –2.9 –3.1 0.3 –2.1 p-value for t-test 0.188 0.457 0.617 0.933 0.447 0.287 0.937 0.329 p-value for F-test 0.493 0.318 0.473 0.484 0.445 0.472 0.490 0.463 Probability levels (p-value) calculated by the t-test and F-test for the yearly means and variance, and percent difference (negative values show model under-estimation) are shown. A probability of 0.05 or lower indicates a departure from the observation that is significant at 5% level (Tingem et al. 2007) Theor Appl Climatolunder-estimates, but with a mean percentage difference of only –2.8%, ranging from –0.6% to –4.5%. Further evidence of good model performance comes when one compares observed and simulated yields for individual crops across all regions. Relative percentage difference oscillate between – 0.1% to –4.6% for bambara, +0.6% to –13% for groundnut, +0.1% to –1.9% for maize (Fig. 1), +0.2% to –1.7% for sorghum and +2.2% to –10.5%. Relative differences between observed and simulated crop yield are a vital part of overall validation, but also provides valuable information on the behaviour (or characteristics) of the yield data, such as when and where it is able to perform well or not, i.e. under-or over-estimating yield. This serves as a reminder of the need for model evaluation prior to application and the development of an understanding as to how uncertainties may be introduced from input data and parameterization. 5 Discussion and conclusion CropSyst simulation of crop yieldswasmade for Cameroon (as a representation of the sub-Saharan central African region) using information on current agricultural practices with a combined observed and generated currentweather data set. The findings showthat, despite the limited availability of calibration data, the model is capable of producing good estimates. According to Ritchie et al. (1998) and Brassard (2003), a difference between observed and simulated yields of up to ±15% is judged acceptable. For all crops and regions, the validation results are within this range, hence the model can be seen as robust under the diverse range of environmental conditions found within Cameroon and could be satisfactorily employed in the assessment of impacts of and adaptations to climate variability and change. This finding has particular relevance for agricultural modelling in Cameroon and elsewhere, where crop models have not been tested, making it difficult to evaluate long-term effects of weather and climate on ecosystem responses. The climate change simulations are the subject of the papers that will follow in this series. It is necessary, however, to consider these results in relative and not in absolute terms. The final yield of a crop is the result of complex interactions between many factors in the soilplant-atmosphere system and also occurrence of the nonsimulated factors such as harvest losses, pests and diseases, which often cause real values to be different from those of the model. However, knowing the scale of errors (i.e. relative percent difference) is important when interpreting outputs from simulation crop models. As with all modelling studies, the estimates made must also be interpreted with consideration of the uncertainties arising from the quality of input data and parameterization effort. We urge more improvement in the amount of available data and rigorous field experimentation which could enhance our ability to assess the impacts of future climate scenarios on cropping systems dynamics. Acknowledgements We acknowledge the help and assistance provided by Claudio O. Stöckle and Roger L. Nelson (Biological Systems Engineering Department, PullmanWA, USA) in using CropSyst. Appendix 1000 1200 1400 1600 1800 2000 2200 2400 Bamenda Batouri Garoua Kribi Maroua Ngaoundere Tiko Yaounde Maize grain yield (Kgha-1) Fig. 1 Calibration simulations (black) vs. observed (grey) crop yields (AGRISTAT 2001) under current climate. Results are grouped by province for clarity 0 100 200 300 400 500 600 700 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Rainfall (mm) Rainfall (mm) Rainfall (mm) 0510 15 20 25 30 35 Bamenda Mean Ppt mean temp 0 100 200 300 400 500 600 700 0510 15 20 25 30 35 Temperature (oC) Temperature (oC) Temperature (oC) Garoua 0 100 200 300 400 500 600 700 0510 15 20 25 30 35 Yaounde Fig. 2 Mean monthly rainfall and temperature distribution at studied sites Theor Appl ClimatolReference Abraha MG, Savage MJ (2006) Potential impacts of climate change on the grain yield of maize for the midland of Kwazulu-Natal, South Africa. Agric Ecosys Environ 115:150–160 AGRISTAT (2001) Semi-annual bulletin of the statistics of agricultural sector 2000/2001, DEPA, Ministry of Agriculture. Yaounde, Cameroon Badini O, Stöckle CO, Franz EH (1997) Application of crop simulation modelling and GIS to agroclimatic assessment in Burkina Faso. Agric Ecosys Environ 64:233–244 Badini O, Stöckle CO, Jones JW, Nelson R, Kodio A, Keita M (2007) A simulation-based analysis of productivity and soil carbon in response to time-controlled rotational grazing in the West African Sahel region. Agric Syst 94:87–96 Batjes N (1995) A homogenised soil data file for global environmental research: a subset of FAO, ISRIC and NRCS profiles (version 1.0). Working paper 95/10, International Soil Reference Information Center (ISRIC), Wageningen, The Netherlands Brassard JE (2003) Valuation des impacts de la hausse de la concentration atmospherique du CO2 et des changements climatiques sur la production agricole du Quebec. Me‘moire de Maitrise, Departement de Geographie, Universite de Montreal, QC, 193 pp DeLancey M, Mike D (2000) Historical dictionary of the Republic of Cameroon, 3rd edn. Scarecrow, Lanham, MD Farre I (1998) Maize (Zea mays L.) and sorghum (Sorghum bicolor L. Moench) response to deficit irrigation. Agronomy and Modelling, PhD Thesis, University of Lieida, Spain, 150 pp Fischer G, Shah M, Tubiello F, Van Velthuizen HT (2005) Socioeconomic and climate change impacts on agriculture: an integrated assessment, 1990–2080. Philos Trans Roy Soc B 360:2067–2083 IPCC (2007) Working Group II Contribution to the Intergovernmental Panel on Climate Change Fourth Assessment Report: Climate Change 2007: impacts, adaptation and vulnerability. Brussels, Belgium Molua EL (2003) Global climate change and Cameroon’s Agriculture: evaluating the economic impacts. PhD Thesis, Institute of Agricultural Economics, Georg-August University, Goettingen, Germany, 94 pp Moriondo M, Maselli F, Bindi M (2007) A simple model of regional wheat yield based on NDVI data. Eur J Agron 26:266–274 Ndemah RN (1999) Towards an integrated crop management strategy for the African stalk borer Busseola fusca (Fuller) (Lepidoptera: Noctuidae) in maize systems in Cameroon. PhD Thesis, University of Hannover, Hannover, Germany, 145 pp Neba A (1999) Modern geography of the Republic of Cameroon, 3rd edn. Neba, Bamenda, Cameroon Priestly CHB, Taylor RJ (1972) On the assessment of surface heat flux and evaporation using large-scale parameters. Mon Weather Rev 100:81–82 Richards LA (1931) Capillary conduction of liquids in porous mediums. Physics 1:318–333 Ritchie JT, Singh U, Godwin DC, Bowen WT (1998) Cereal growth, development and yield. In: Tsuji GY, Hoogenboom G, Thornton PK (eds) Understanding options for agricultural production. Kluwer, Dordrecht, pp 79–98 Rivington M, Matthews KB, Bellocchi G, Buchan K, Stöckle CO, Donatelli M (2006) An integrated assessment approach to conduct analyses of climate change impacts on whole-farm systems. Environ Model Softw 22:202–210 Rosenzweig C, Hillel D (1998) Climate change and the global harvest. PBD: 1998. United States, Oxford University Press, New York, 323 pp Stöckle CO, Donatelli M, Nelson R (2003) CropSyst, a cropping systems simulation model. Eur J Agron 18:289–307 Thornton PK, Jones PG (2003) The potential impacts of climate change on maize production in Africa and Latin America in 2055. Glob Environ Change 13:51–59 Thornton PK, Jones PG, Owiyo TM, Kruska RL, Herero M, Kristjanson P, Notenbaert A, Bekele N (2006) Mapping climate vulnerability and poverty in Africa. Report to the Department for International Development, ILRI, Nairobi, Kenya, 200 pp Tingem M, Rivington M, Azam Ali SN, Colls JJ (2007) Assessment of the ClimGen stochastic weather generator at Cameroon sites. Afr J Environ Sci Technol 1:86–92 Tingem M, Rivington M, Azam Ali SN, Colls JJ (2008) Climate variability and maize production in Cameroon: simulating the effects of extreme dry and wet years. Singapore J Trop Geogr (in press) Worldbank (2007) World Development Indicators Database. Avaliable at http://devdata.worldbank.org/. Cited 24 December 2007 World Fact Book (2007) (Cameroon): United States Central Intelligence Agency (CIA). Aavailable at https://www.cia.gov/library/publications/the-world factbook/geos/cm. Cited 9 December 2007 Theor Appl Climatol