Global Change Biology (2015) 21, 4588–4601, doi: 10.1111/gcb.13061
Climate change and maize yield in southern Africa: what can farm management do? JAIROS RURINDA1,2,3, MARK T. VAN WIJK1,4, PAUL MAPFUMO2,3, KATRIEN D E S C H E E M A E K E R 1 , I W A N S U P I T 5 and K E N E . G I L L E R 1 1 Plant Production Systems, Wageningen University, PO Box 430, 6700AK Wageningen, The Netherlands, 2Department of Soil Science and Agricultural Engineering, University of Zimbabwe, PO Box MP 167, Mount Pleasant, Harare, Zimbabwe, 3Soil Fertility Consortium for Southern Africa (SOFECSA), University of Zimbabwe, PO Box MP 167, Mount Pleasant, Harare, Zimbabwe, 4International Livestock Research Institute (ILRI), Box 30709, Nairobi 00100, Kenya, 5Earth System Science & Climate Adaptive Land Management, Wageningen University and Research Centre, PO Box 47, 6700 AA Wageningen, The Netherlands
Abstract There is concern that food insecurity will increase in southern Africa due to climate change. We quantified the response of maize yield to projected climate change and to three key management options – planting date, fertilizer use and cultivar choice – using the crop simulation model, agricultural production systems simulator (APSIM), at two contrasting sites in Zimbabwe. Three climate periods up to 2100 were selected to cover both near- and long-term climates. Future climate data under two radiative forcing scenarios were generated from five global circulation models. The temperature is projected to increase significantly in Zimbabwe by 2100 with no significant change in mean annual total rainfall. When planting before mid-December with a high fertilizer rate, the simulated average grain yield for all three maize cultivars declined by 13% for the periods 2010–2039 and 2040–2069 and by 20% for 2070– 2099 compared with the baseline climate, under low radiative forcing. Larger declines in yield of up to 32% were predicted for 2070–2099 with high radiative forcing. Despite differences in annual rainfall, similar trends in yield changes were observed for the two sites studied, Hwedza and Makoni. The yield response to delay in planting was nonlinear. Fertilizer increased yield significantly under both baseline and future climates. The response of maize to mineral nitrogen decreased with progressing climate change, implying a decrease in the optimal fertilizer rate in the future. Our results suggest that in the near future, improved crop and soil fertility management will remain important for enhanced maize yield. Towards the end of the 21st century, however, none of the farm management options tested in the study can avoid large yield losses in southern Africa due to climate change. There is a need to transform the current cropping systems of southern Africa to offset the negative impacts of climate change. Keywords: adaptation, APSIM, fertilizer use, planting date, simulation modelling, Zea mays Received 23 November 2014 and accepted 13 July 2015
Introduction Concern of food insecurity has further increased in southern Africa because of a changing climate. Rising temperatures and changing rainfall patterns are already evident in southern Africa and threaten crop production (Lobell et al., 2008). In Zimbabwe, mean daily maximum temperature has increased by between 0.2 °C and 0.5 °C per decade, over the period of 1962–2000 (Rurinda et al., 2014), and is projected to further increase by 2–4 °C by 2100 (Unganai, 1996) similar to global projections (IPCC, 2013). A large area of southern Africa is projected to experience a decrease in rainfall by 2100 (Shongwe et al., 2009). Neukom et al. (2014) indicated that rainfall already declined in southern Africa between 1796 and 1996. Other studies have Correspondence: Jairos Rurinda, tel./fax +263 04 307304, e-mail: [email protected]
4588
indicated that total rainfall has not changed, but that rainfall patterns are more variable with delays in the onset of the rainy season and more frequent droughts (Tadross et al., 2009; Rurinda et al., 2013). Overall, although rainfall projections contain a large uncertainty and spatial variation, southern Africa is expected to become drier (IPCC, 2013). The increasing temperatures, in combination with more severe and frequent droughts, will profoundly reduce soil water available for plant uptake. Fraser et al. (2013) projected that soil moisture will decline by 25% in southern Africa due to more frequent droughts. Rising temperatures will shorten the crop growth period and increase plant water demand through higher transpiration rates, both potentially reducing plant production (Ludwig & Asseng, 2006; Springate & Kover, 2014). Furthermore, increasing temperatures will directly affect plants through heat waves, and this impact will be large when coupled with soil moisture © 2015 John Wiley & Sons Ltd
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4589 deficits. Lobell et al. (2011) reported that each degree day above 30 °C reduced maize yield by 1.7% under drought, compared with a decrease of 1% under favourable rain-fed conditions in Africa. Given that the impacts of higher temperatures are most pronounced on sandy soils of low water-holding capacity (Ludwig & Asseng, 2006), the predominant soils across smallholder cropping areas in southern Africa (World Soil Resource Base, 1998), smallholder farmers face a high risk of declining crop yields. Climate change may also have a positive influence on crop productivity. Crop yield is anticipated to improve due to increased concentration of carbon dioxide (CO2) in the atmosphere (Ludwig & Asseng, 2006). However, maize, the main staple cereal crop in southern Africa, is a C4 plant, which will benefit relatively little from increased CO2 concentrations (Taylor et al., 2014). Moreover, the poor soil nutrient availability on highly weathered sandy soils, which cover a large area of smallholder farming areas in the region, can reduce any yield benefit of elevated CO2 (Tubiello & Ewert, 2002). It is therefore highly likely that the impacts of increased temperatures coupled with soil moisture deficits will override the compensating effects of increased CO2 on crop yields in southern Africa (Rosenzweig & Parry, 1994). Efforts have been made to understand and quantify the impacts of increased temperatures and changing rainfall patterns on crop productivity in southern Africa (Fischer et al., 2005; Zinyengere et al., 2013). Crop production in Zimbabwe was considered to be determined mainly by the availability of soil water, with temperature not being a major limiting factor (Hussein, 1987). However, using a statistical model, Lobell et al. (2008) predicted that maize production will decline by between 20% and 40% in southern Africa due to a combination of warming temperatures and changing rainfall patterns. These studies quantified the possible effects of climate change on crop productivity, but did not analyse how these effects interact with the opportunities of adaptive farm management such as cultivar choice, timing of farming operations and adjusting soil nutrient inputs. Yet the net impact of climate change on crop yield depends strongly on the interactions between climate and such farm management factors (Reidsma et al., 2009). To provide a comprehensive assessment of climate change effects on crop yields, it is critical to understand the interactions between climate and possible adaptive farm management options. It has been argued that a relatively small change in farm management and selection of different crop varieties can significantly reduce any negative impact of moderate climate change (Howden et al., 2007). Through field experimentation, © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
Rurinda et al. (2013) demonstrated that improved timing of planting and adjusting soil nutrient inputs can stabilize maize yields under variable rainfall conditions in smallholder farming systems in Zimbabwe. By modelling the current cropping systems in sub-Saharan Africa (SSA), Folberth et al. (2013) showed that under irrigation, increasing soil nutrient supply in combination with improved cultivars would allow for a doubling of maize yields. However, the outcome for the same strategies is likely to be different under rain-fed conditions because of the interaction between crop fertilization and rainfall. In the few impact studies that have taken into account farmer adaptive management options (e.g. Waha et al., 2013), the broad scale of assessment makes their findings difficult to translate into knowledge that can drive local solutions. Although there is much debate about the appropriate scale to operate (Challinor et al., 2009), local studies with crop models allow for better calibration and validation (e.g. by taking into account soil nutrient and water conditions), compared with regional approaches (Fischer et al., 2005), and can also help farmers to make appropriate decisions. Further, the impacts of the changing climate on crop yields vary with location due to spatial variability in climate, particularly rainfall. For example, responses of maize in southern Africa to the changing climate can be as wide as 40% to +10% (Zinyengere et al., 2013). In addition, many adaptive farm management options have been identified mainly through a top-down approach (e.g. Phillips et al., 1998). Because of the uncertainties associated with the changing climate (Dessai & van der Sluijs, 2007), and the differences in endowments between farmers (Mtambanengwe & Mapfumo, 2005), a bottom-up approach can be useful for integrating local farmers’ and expert empirical knowledge as well as linking knowledge to action. We quantified the yield response of maize, the main staple crop in southern Africa, to projected climate change and to three key management options, namely adjusting planting dates, adjusting soil nutrient inputs and using different maize cultivars. Farmers believed that these adaptive management options could reduce the negative impacts of the changing climate on maize yields (Mapfumo et al., 2013). We focus especially on the interactions between these factors, to assess how the efficiency of interventions might change in future in southern Africa. We hypothesized that (i) the average yields for maize will decrease with increasing temperatures and changes in rainfall patterns for the future periods, 2010–2039, 2040–2069 and 2070–2099, compared with average yields for the baseline climate; (ii) the response of maize yield to fertilization and planting date will change with climate change; and (iii) this response will be affected by the choice of maize
4590 J . R U R I N D A et al. cultivar. The overall hypothesis was tested that improved management of planting date, fertilization and choice of crop cultivar have the potential to compensate for the predicted decreases in crop yields due to climate change in Zimbabwe.
Materials and methods
Study sites Two districts in eastern Zimbabwe with contrasting climates were studied: Makoni has a dry subhumid climate with annual rainfall ranging from 750 to 1000 mm, and Hwedza has a semi-arid tropical climate with annual rainfall ranging from 650 to 800 mm. The study sites were purposefully selected to represent many smallholder areas in southern Africa (see Fig. S1). Large areas in smallholder farming systems of southern Africa are typified by dry subhumid and semi-arid climates. More than 40% of smallholder agriculture in SSA is conducted in semi-arid regions (Rockstrom, 2004). The two study areas are also marked by increased climate variability and rising temperatures as in many parts of southern Africa (Tadross et al., 2009). Rainfall in Zimbabwe, as in most of southern Africa, falls between October and April with most rainfall in December and February. The farming system is maize-based with cattle providing draught power and manure for crop production, typical of smallholder farming areas in Zimbabwe. In both sites, average monthly temperatures are above 20 °C between October and April, and temperatures are highest in October before the start of the rains. The soils in both sites are granite-derived sandy soils, mainly Lixisols and Arenosols, with little capacity to store moisture (World Soil Resource Base, 1998). These sandy soils are representative for large areas in SSA and in particular southern Africa. Arenosols cover about 13% of SSA and over 6.5 million ha of cropland in southern Africa (Hartemink and Huting, 2008).
Description of the APSIM crop simulation model Maize yield responses to climate, planting date variation and amount of nitrogen applied were simulated using the agricultural production systems simulator (APSIM). APSIM is a process-based, daily time step model developed to simulate biophysical processes in farming systems in response to management decisions and in the face of climatic risk (Holzworth et al., 2014). APSIM has been tested against field experimental data in a wide range of growing conditions across the globe (Keating et al., 2003) and in semi-arid and subhumid regions of southern Africa (Chikowo et al., 2008). APSIM has recently been applied to evaluate risk of crop production in relation to climate variability and change across many areas in, for example, Europe, China and Africa (Ludwig & Asseng, 2006; Liu et al., 2013; Turner & Rao, 2013). We used APSIM version 7.5 (CSIRO Agriculture Flagship, The University of Queensland, Queensland, Australia) to quantify the sensitivity of maize yield to different adaptive farm management options under future predicted values for
daily weather variables, including rainfall, temperature and solar radiation. Daily weather data for baseline and future climates were retrieved from outputs of global circulation models (GCMs), described below. The main APSIM modules used in this study included the plant (maize), environment (meteorological input module, soil water, soil nitrogen and organic matter dynamics, soil phosphorus) and modules management. The soil phosphorus module was included because phosphorus is a limiting nutrient in soils of southern Africa including Zimbabwe. Each APSIM module demands a number of parameters. For the SOILWAT model, which simulates the dynamics of soil water, the input parameters included soil bulk density, soil water lower limit (LL15) and drained upper limit (DUL), and two parameters, U and CONA, which determine first- and second-stage soil evaporation. LL15 and DUL were derived based on soil classification using regression equations calculated by Hussein (1983). Soil moisture content at saturation was estimated from bulk density. The parameters, U and CONA, were set at 6.0 and 3 mm day 1, respectively, values acceptable for tropical conditions (Chikowo et al., 2008). A value of 0.7 was used for SWCON, a coefficient that specifies the proportion of the water in excess of field capacity that drains to the next layer in 1 day (Chikowo et al., 2008). The bare soil run-off curve number was set at 50 to take into account the low runoff associated with sandy soils because of high infiltration rates (Hussein, 1987). For the soil model, the organic matter content for each soil layer was measured in farmers’ fields during the experiments that were conducted in Makoni and Hwedza districts of Zimbabwe, between 2009 and 2012 (Rurinda et al., 2013). The initial soil N was set at 35 kg ha 1 (23 kg NO3 -N and 12 kg NH4+-N) based on field measurements conducted at the same study sites (Mtambanengwe & Mapfumo, 2006). The major soil parameters for one study site, Hwedza, are presented in Table S1. Soil data for Makoni is not shown because it was similar to that of Hwedza. APSIM crop parameters for SC401 (131 days to maturity), SC501 (137 days to maturity) and SC625 (142 days to maturity; see Supporting Information, Table S2) were selected to represent the very early, early and medium maturing cultivars used in the field experiments (Rurinda et al., 2013). The cultivars used in the model and field experiments are available on the market, and they are commonly grown in many areas of southern Africa including in the study areas. During simulations, soil organic matter, nitrogen, phosphorus and water were re-initialized for each growing season at the start of the planting window.
Description of the climate data A useful way of dealing with uncertainties in the future climate is to use a range of possible future climate change scenarios rather than a single projection (Challinor et al., 2009). Two representative concentration pathways (RCPs) – RCP4.5 and RCP8.5 – were the inputs for scenario modelling with an ensemble of five GCMs – CNRM-CM5, EC-EARTH, HADGEM2-ES, IPSL-CM5A-LR and MPI-ESM-LR. RCPs refer to the portion of the concentration pathway extending up to the
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4591 year 2100, for which integrated assessment models produced corresponding emission scenarios (IPCC, 2013). The RCP8.5 is a high concentration pathway for which radiative forcing reaches >8.5 W m 2 by 2100, and the RCP4.5 is an intermediate pathway in which radiative forcing is stabilized at about 4.5 W m 2 (Moss et al., 2010). The radiative forcing trajectories result from diverse combinations of socio-economic, technological, and policy and institutional development scenarios. Using multiple model outputs (ensemble) increases confidence in the projections compared with individual models (IPCC, 2013). The resolution of each GCM and main characteristics of each RCP are presented in Table S3. The climate data were obtained from the coupled model intercomparison project phase 5 (http://pcmdi3.llnl.gov/esgcet/home.htm, last accessed 4 January 2014). The GCM data were regridded to a spatial resolution of 0.5° 9 0.5°. The precipitation, minimum and maximum temperatures data from the 5 GCMs used in the study were biascorrected against the WATCH Forcing Data (see, http:// www.eu-watch.org/watermip/use-of-WATCH-forcing-data) using the method of Piani et al. (2010). The radiation data were bias-corrected with the method of Haddeland (2012). Both bias correction methods used the WATCH forcing data series (1971–2000) as a reference. Many studies have shown that bias-corrected climate data improve impact assessment results because of the adjustment of the simulated data to that of the observation (e.g. Hagemann et al., 2011). Climate model output, however, is affected by inconsistencies which cannot be addressed by bias correction such as the incorrect representations of dynamical and/or physical processes (Haerter et al., 2011). Inherent uncertainties due to internal variability of the climate system also limit the precision with which impacts are assessed (Challinor et al., 2009). Although in general the use of ensembles improves the modelling of climate data, aggregating outputs from several models can mask the effects of growing season dry spells. In this study, we therefore decided to use individual GCM data to run APSIM and afterwards calculate the average crop yields across the five GCMs. Because all five models were driven by the same radiative forcing scenarios, all runs represented an equally possible projection of the future change of the climate. Accordingly, the bias-corrected GCMs daily weather data for temperature, rainfall and solar radiation were used separately in APSIM to simulate maize growth and yield. The simulated yields from each of the five bias-corrected GCM data were then aggregated to provide average yield for each time period. The sensitivity of maize yield was assessed for three future climate periods – 2010–2039, 2040–2069 and 2070–2099 – and compared against a baseline period – 1976–2005. The baseline (1976–2005) maize yield was simulated both with the bias-corrected GCM data and with the observed meteorological point data for Hwedza. A comparison between the two allowed evaluating the appropriateness of using regridded bias-corrected GCM data. Longterm meteorological observed point data for Makoni were incomplete and inconsistent and hence could not be used in the study. The three future climate periods were selected to cover both near-term climate relevant for assessing relatively immediate benefits to agricultural investments (Lobell et al.,
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
2008) and long-term climate for sustainable crop production, illustrating the situation in which climate change is more clearly separated from natural climate variability (Ruane et al., 2013). The simulations did not extend beyond the year 2100 because model and scenario uncertainties increase with lead time (Hawkins and Sutton, 2009).
Adaptive management options identified by farmers and field experimentation for testing APSIM Yield and biomass data used to test APSIM performance were derived from experiments conducted from 2009 to 2011 in farmers’ fields in Hwedza and Makoni districts of Zimbabwe (Rurinda et al., 2013). Three maize cultivars, three planting dates and three fertilization rates were laid out in a split-plot block design with three replications per treatment. Planting date was assigned to the main plot, and fertilization rate 9 maize cultivar subplots were randomized within the main plot. The three maize cultivars were SC 403 (131 days to maturity), SC 513 (137 days to maturity) and SC 635 (142 days to maturity). These represent the range of maize cultivars currently available to farmers and differ only by 11 days in time to maturity. The three planting windows defined together with farmers were 25 October–20 November, 21 November–15 December and 16 December–1 January. The three fertilization rates were a control treatment (unfertilized), low rate (35 kg N ha 1, 14 kg P ha 1, 3 t ha 1 manure, on a dry weight basis) and high rate (90 kg N ha 1, 26 kg P ha 1, 7 t ha 1 manure, on a dry weight basis). The basal fertilizer was applied as compound D (7% N, 14% P2O5 and 7% K2O) and top dressing as ammonium nitrate (34.5%). Rainfall for the 2009/2010 and 2010/2011 seasons was recorded on site. To evaluate the performance of the APSIM crop model, yield and biomass output from observed climate data simulations were compared to observed field experimental data for both Hwedza and Makoni study areas for the 2009/2010 and 2010/2011 seasons. Model error was expressed as the rootmean-squared error (RMSE) and the regression coefficient used to evaluate model precision. To assess maize yield response to planting date, the maize cultivars were planted in the model at an interval of 1 week from November 1st to January 10th, representative of the range of potential sowing dates (Rurinda et al., 2013). This was done for the three fertilizer rates as mentioned above. To assess maize yield response to N, increasing rates of ammonium nitrate were applied at 35 days after sowing from 0 up to 120 kg N ha 1 with an interval of 5 kg N ha 1. The maize yield–nitrogen response curves were simulated for each of the three planting windows defined by farmers, and phosphorus was applied at planting at 26 kg ha 1. In all scenarios, the maize cultivars were planted at 44 000 plants ha 1 in each season following 20 mm of rainfall in three consecutive days. If the rainfall condition was not met in a particular week (in the case maize response to planting date) or farmerdefined planting window (in the case maize response to N), the crop was forced to be planted at the end of the week or window.
4592 J . R U R I N D A et al. Fig. 1 Observed and predicted maize yield for (a) 2009/2010 season (root-mean-squared error, RMSE = 0.42, R2 = 0.88) and (b) 2010/ 2011 season (RMSE = 0.20, R2 = 0.70); biomass for (c) 2009/2010 season (RMSE = 0.75; R2 = 0.86) and (d) 2010/2011 season (RMSE = 0.91, R2 = 0.73), for all three adaptive management options. Three planting dates (early planting – represented by the green colour; normal planting – dark yellow; and late planting – dark red); three fertilization rates (zero fertilization – represented by crossed shapes; low rate: 14 kg P ha 1, 35 kg N ha 1, 3 t ha 1 manure – semi-filled shapes; and high rate: 26 kg P ha 1, 90 kg N ha 1, 7 t ha 1 manure – full-filled shapes) and three cultivars (SC401 – represented by a circle; SC501 – square; and SC625 – diamond), for (A) Makoni and (B) Hwedza.
Results
Performance of APSIM model APSIM performed well in predicting yields for maize planted early or during the normal window, for all fertilization rates and cultivars in the relatively good 2009/2010 rainfall season at both sites (RMSE < 0.5, R2 > 0.8; Fig. 1Aa, Be). Similarly, the model predicted the biomass reasonably well in the 2009/2010 season in both study areas (Fig. 1Ac, Bg). In the 2010/2011 season characterized by waterlogging conditions and prolonged dry spells, APSIM over-predicted both yield and biomass, especially for the nutrient-depleted soils in Hwedza (Fig. 1Bf, Bh). The discrepancies in the performance of APSIM with season’s rainfall quality suggest that the model does not perform well under waterlogging conditions.
Projected temperature and rainfall conditions by 2100 The temperature is projected to increase significantly in the future, and this increase will be stronger with the high radiative forcing scenario at both Makoni and Hwedza (Fig. 2Ad–f, Bj–l). Compared with the baseline climate, the minimum temperature is projected to increase by between 3.2 °C and 5.4 °C and maximum temperature by between 3.2 °C and 4.9 °C for the time horizon 2070–2099 for all five GCMs, under the high radiative forcing of 8.5 W m 2 for Hwedza (Fig. 2Bl). The greatest increase in temperature is projected for the time horizon 2070–2099, under the high radiative forcing of 8.5 W m 2 (Fig. 2Af, Bl). The direction of possible change in total rainfall is less clear for all three future time periods and for each radiative forcing (Fig. 3Aa– f, Bg–l). Coinciding trends such as increased length of the largest period of consecutive dry days and increased rainfall intensity (see Supporting Information, Fig. S2A–C), observed in different time horizons, can nullify signals in the mean total rainfall trend. Although the mean annual rainfall change signal is unclear by 2100, rainfall variability is highly likely to increase (Fig. S2A–C).
Consequences of the changing climate and its interaction with adaptive management options on maize yield Impact of maize cultivar choice on yield under varied scenarios of future climate. The baseline maize yield simulated with the (1976–2005) bias-corrected GCM data was similar to that simulated with on-site observed historical climate data (Fig. 4a, b). Yield responses for all three maize cultivars to future changes in climate as well as to adaptive management of planting date and soil nutrient input were similar (data not shown). In other words, the differences in yield between cultivars were negligible, even under future change in climate in both study areas. Compared with the baseline climate when planting before mid-December with a high fertilizer rate in Hwedza, the simulated average grain yield for all three maize cultivars declined by an average of 13% for the time horizons 2010–2039 and 2040–2069, and by 18% for 2070–2099 with radiative forcing of 4.5 W m 2 (Fig. 5Bg). With the same management, but under radiative forcing of 8.5 W m 2, the simulated grain yield in Hwedza declined by an average of 13% for the time horizon 2010–2039, 18% for 2040– 2069 and 32% for 2070–2099 (Fig. 5Bj). Climate change effects on maize were small when little fertilizer was used at both study areas (Fig. 5Bh,k). Thus, the greatest maize yield loss is projected towards the end of the 21st century mostly under the high radiative forcing of 8.5 W m 2 when a high rate of fertilizer is applied. Although a greater yield loss was simulated for Hwedza, the trends in yield changes were similar with that in Makoni, a region that generally receives more rainfall (Fig. 5Aa–f). Delayed planting effects on maize yield under varied scenarios of future climate. The simulated average maize yield increased gradually with planting date from early November to mid-December at both study sites (Fig. 5Aa,b,d,e, Bg,h,j,k). After mid-December, simulated yields declined drastically. Overall, the maize yield response to planting date was similar from November to mid-December for all three future climate periods for each radiative forcing (Fig. 5). Under zero
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4593
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
4594 J . R U R I N D A et al.
Fig. 2 Projected maximum and minimum temperatures for three climate change periods: near-term (2010–2039), mid-term (2040–2069) and long-term (2070–2099), against simulated baseline temperatures (solid lines), for two radiative forcings: RCP 4.5 and RCP 8.5, for (A) Makoni and (B) Hwedza. The box and whisker plots show the temperature variation based on the ensembles of five global circulation models. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4595
Fig. 3 Projected daily rainfall in each month for three climate change periods: near-term (2010–2039), mid-term (2040–2069) and longterm (2070–2099), against simulated baseline rainfall (solid lines), for two radiative forcings: RCP 4.5 and RCP 8.5, for (A) Makoni and (B) Hwedza. The box and whisker plots show the rainfall variation based on the ensembles of five global circulation models. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
4596 J . R U R I N D A et al.
Fig. 4 Baseline average maize yield (1976–2005) obtained with varying planting date for three maize cultivars (SC401, SC501 and SC625) under three fertilization rates (CR – zero fertilization; LR – low rate: 35 kg N ha 1, 14 kg P ha 1, 3 t ha 1 manure; and HR – high rate: 90 kg N ha 1, 26 kg P ha 1, 7 t ha 1 manure), for (a) on-site measured historical climate data and (b) bias-corrected global circulation model data, for Hwedza.
fertilization, no effect of planting date was observed (Fig. 5Ac,f, Bi,l). Responses of maize yield to adjustments in the amount of nitrogen under a changing climate. Fertilizer increased yield significantly under both the baseline and future climates particularly when planting before mid-December compared to the control (zero fertilization; e.g. Fig. 5Aa, Bg). However, after mid-December, the impact of fertilizer on yield was drastically decreased (e.g. Fig. 5Aa, Bg). The response of maize to increased nitrogen applied decreased for all three climate periods for each radiative forcing, compared with the baseline climate. For example, at 80 kg N ha 1, a maize yield of about 4.5 t ha 1 was simulated for the baseline climate, against maize yields of about 3.5 t ha 1 for 2010–2069 and 3 t ha 1 for 2070–2099 under the RCP4.5 at both study sites (Fig. 6Aa–d, Bi–l). Maize yield responses to nitrogen were comparable between climate periods 2010–2039 and 2040–2069 for each radiative forcing (Fig. 6Ab,c, Bj,k). Maize yield responses to nitrogen further decreased with increasing climate change with time. The smallest yield response to nitrogen was simulated for the climate period 2070–2099 and for the high radiative forcing of 8.5 W m 2. For example, applying 60 kg N ha 1 resulted in a maximum yield benefit of only about 2.5 t ha 1 for the period 2070–2099 for RCP8.5 (Fig. 6Bp).
Discussion Our analyses suggest that the average yields for maize in Zimbabwe will decrease with the increasing temper-
atures predicted for the current century, although the yield decline will be relatively small ( 13%) until the middle of the century particularly for the low greenhouse gas emission scenario. Larger yield losses ( 32%) are projected for the more distant future, that is towards the end of the century, and under high fertilization rate for the high radiative forcing scenario of 8.5 W m 2. Although the yield trend for the two sites studied was similar, the impacts of climate change on maize yield were more pronounced for Hwedza, a semi-arid region characterized by higher temperatures and increased rainfall variability (Rurinda et al., 2014). The decline in future yields was driven mainly by increasing temperatures that increased the crop maturation rate and hence shortened the crop growing period (Springate & Kover, 2014). This means less crop biomass can be assimilated, and hence, total crop biomass is reduced, which directly results in yield reduction. Using a statistical model and for a shorter time horizon, that is 2030, Lobell et al. (2008) projected much larger losses of maize yield (< 30%) in southern Africa, mainly due to increased temperatures and decreased rainfall. Suboptimal crop management, studied here in relation to application of mineral fertilizer and planting date of maize, has much larger negative impacts on maize production in the near future. In particular, the nonlinear effect of delay in planting date on maize production is important, because it suggests that farmers have quite some flexibility in choosing their planting date up to mid-December. This finding is contrary to the existing literature in which delays in planting date are often directly associated with lower yields (e.g.
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4597
Fig. 5 Average seasonal yield distribution with planting date for maize cultivar SC401 under three fertilization rates (CR – zero fertilization; LR – low rate: 35 kg N ha 1, 14 kg P ha 1, 3 t ha 1 manure; HR – high rate: 90 kg N ha 1, 26 kg P ha 1, 7 t ha 1 manure), for the simulated baseline climate (1976–2005) and in response to climate change for three periods: near-term (2010–2039), mid-term (2040–2069) and long-term (2070–2099) and for two radiative forcings: RCP 4.5 and RCP 8.5, for (A) Makoni and (B) Hwedza. Data for the other two cultivars (SC501 and SC625) are not presented because the yields for all three cultivars were similar. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
4598 J . R U R I N D A et al.
Fig. 6 Simulated average maize yield for three maize cultivars planted early: SC401, SC501 and SC625, in response to nitrogen fertilization for three future climate periods: 2010–2039, 2040–2069 and 2070–2099, against the simulated baseline climate, 1976–2005, for two radiative forcings: RCP 4.5 and RCP 8.5, for (A) Makoni and (B) Hwedza. The standard deviation (SD) presented is for one cultivar (SC401) because the standard deviation for all three cultivars was similar.
Waddington & Hlatshwayo, 1991) and widely applied ‘rules-of-thumb’ which suggest a 2.3% decline in grain yield per day delay in planting (Shumba, 1989) over the
period of October to mid-December. Our results even suggest a slight yield increase with small delays in planting, which would also hold under a changing © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4599 climate (see e.g. Fig. 5Aa). Delays in planting are caused mainly by lack of draught power, farm labour and fertilizer or farmers’ perceptions of rainfall patterns (Shumba et al., 1992). The similarity in simulated yields until planting dates after mid-December reinforced the findings from our earlier experimental studies that there is no yield difference between crops planted early (25 October–20 November) or during what farmers considered to be a normal planting window (21 November–15 December; see Fig. S3; Rurinda et al., 2013). Pushing for very early planting will not bring much yield benefit under a changing climate, while the risk of losing the crop might increase due to increased rainfall variability during the start of the growing season (Raes et al., 2004). Similarly, in other regions of Africa, changing sowing dates during the crop growing period is projected to be ineffective in counteracting adverse climatic effects because of the narrow window of rainfall that affects the timing of farm operations (Tingem et al., 2009). Towards the end of the 21st century, none of the farm management options considered in this study can avoid large yield losses of more than 30%. In particular, the maize response to mineral nitrogen is projected to decline significantly as the climate changes. Although fertilizer use will remain a key strategy to enhance crop production also under climate change, if crops are not planted too late (see Fig. 5; Rurinda et al., 2013), maize yields might plateau at smaller mineral nitrogen rates than under current climate (see e.g. Fig. 6Bp). Our simulation results indicate that farmers who can afford to buy more fertilizers could reduce their application rate to about 80 kg N ha 1 by mid-century and 60 kg N ha 1 towards the end of the century, against the current recommendation rate of 120 kg N ha 1 in Zimbabwe to increase economic returns. Rurinda et al. (2013) demonstrated that the use of smaller amounts of fertilizers would not compromise household food selfsufficiency in smallholder farming systems. However, if fertilizer use is reduced, farmers will not be able to maximize production suggesting that investment outside crop production will be critical to enhance household income and food security. The yields of all three cultivars, which represent the maturity range of the cultivars currently available in much of southern Africa (http://seeds.seedco.co/zimbabwe, accessed 20 July 2015), are likely to respond similarly to the predicted future climates. The range in time to maturity of these cultivars is small, with a difference between the relatively short (SC403) and long (SC635) duration cultivars of only 11 days. The yield potential of these cultivars is claimed to vary between 4 and 9 t ha 1 (see, © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
Table S2), but we found no yield difference tested in the field under farmers management (Rurinda et al., 2013). Given that farmers are now experiencing delayed onset of rainfall with no change in the date of the end of rainy season (Rurinda et al., 2013), breeding for shorter maturing cultivars with greater tolerance to more frequent droughts could help to stabilize future crop production. By contrast, Traore (2014) found that short-duration varieties seemed more vulnerable to predicted increases in temperature than long-duration varieties in West Africa. The model analysis of Traore (2014) suggests that the growing period of the short-duration maize varieties is shortened due to increased temperatures, so that overall biomass and grain production is reduced. This could mean that although shortduration varieties are perceived to be more climate robust because they are not dependent on a long growing season with ample rainfall, the temperature response of such varieties might counteract this positive characteristic. More detailed crop experimentation is needed to investigate how these interactions work out in reality compared with model analyses. However, new cultivars will only help smallholder farmers to adapt to climate change if other binding constraints to crop production such as poor soil fertility are first removed (Tittonell & Giller, 2013). Overall, our results indicate that in the short-term improved farm management practices can buffer the negative impacts of climate change. Yet in the longer term, climate change will further increase food insecurity of smallholder farmers in southern Africa given that none of the current farm management practices, including fertilization, planting date adjustments and cultivar choice, can avoid large yield loss. The limitation of the current farm management options to compensate for the huge yield loss due to long-term climate change suggests that there is need to transform the cropping systems in southern Africa to offset the negative impacts of climate change and build a climate resilient livelihood system. Opportunities are also needed for farmers to explore livelihood options outside agriculture.
Acknowledgements We thank the International Development Research Centre (IDRC) and Department for International Development (DFID) for funding through the Climate Change Adaptation in Africa (CCAA) Grant #104140 to the University of Zimbabwe. Climate data were provided through the Combine project (European Commission’s 7th Framework Programme Grant Agreement No. 226520) and the IPOP Sustainable and Smart Food Supply programme (Wageningen University, the Netherlands). We would also like to thank the two anonymous reviewers whose comments greatly improved the manuscript.
4600 J . R U R I N D A et al.
References
Mtambanengwe F, Mapfumo P (2005) Organic matter management as an underlying cause for soil fertility gradients on smallholder farms in Zimbabwe. Nutrient
Challinor AJ, Ewert F, Arnold S, Simelton E, Fraser E (2009) Crops and climate
Cycling in Agroecosystems, 73, 227–243. Mtambanengwe F, Mapfumo P (2006) Effects of organic resource quality on soil profile N dynamics and maize yields on sandy soils in Zimbabwe. Plant and Soil, 281, 173–191. Neukom R, Nash DJ, Endfield GH et al. (2014) Multi-proxy summer and winter precipitation reconstruction for southern Africa over the last 200 years. Climate Dynamics, 42, 2713–2726. Phillips JG, Cane MA, Rosenzweig C (1998) ENSO, seasonal rainfall patterns and sim-
change: progress, trends, and challenges in simulating impacts and informing adaptation. Journal of Experimental Botany, 60, 2775–2789. Chikowo R, Corbeels M, Tittonell P, Vanlauwe B, Whitbread A, Giller KE (2008) Aggregating field-scale knowledge into farm-scale models of African smallholder systems: summary functions to simulate crop production using APSIM. Agricultural Systems, 97, 151–166. Dessai S, van der Sluijs J (2007) Uncertainty and Climate Change Adaptation – A Scoping Study. Netherlands Environmental Assessment Agency, Utrecht. Fischer G, Shah M, Tubiello FN, Van Velhuizen H (2005) Socio-economic and climate change impacts on agriculture: an integrated assessment, 1990–2080. Philosophical Transactions of the Royal Society B: Biological Sciences, 360, 2067–2083. Folberth C, Yang H, Gaiser T, Abbaspour KC, Schulin R (2013) Modeling maize yield responses to improvement in nutrient, water and cultivar inputs in sub-Saharan Africa. Agricultural Systems, 119, 22–34. Fraser EDG, Simelton E, Termansen M, Gosling SN, South A (2013) “Vulnerability hotspots”: integrating socio-economic and hydrological models to identify where cereal production may decline in the future due to climate change induced drought. Agricultural and Forest Meteorology, 170, 195–205. Haddeland I, Heinke J, Voß F, Eisner S, Chen C, Hagemann S, Ludwig F (2012) Effects of climate model radiation, humidity and wind estimates on hydrological simulations. Hydrology and Earth System Sciences, 16, 305–318. Haerter JO, Hagemann S, Moseley C, Piani C (2011) Climate model bias correction and the role of timescales. Hydrology and Earth System Sciences, 15, 1065–1079. Hagemann S, Chen S, Haerter JO, Heinke J, Gerten D, Piani C (2011) Impact of a statistical bias correction on the projected hydrological changes obtained from three GCMs and two hydrology models. Journal of Hydrometeorology, 12, 556–578. Hartemink AE, Huting J (2008) Land cover, extent, and properties of Arenosols in Southern Africa. Arid Land Research and Management, 22, 134–147. Hawkins E, Sutton R (2009) The Potential to Narrow Uncertainty in Regional Climate Predictions. Bulletin of the American Meteorological Society, 90, 1095–1107. Holzworth DP, Huth NI, deVoil PG et al. (2014) APSIM – evolution towards a new generation of agricultural systems simulation. Environmental Modelling & Software, 62, 327–350. Howden SM, Soussana JF, Tubiello FN, Chhetri N, Dunlop M, Meinke H (2007) Adapting agriculture to climate change. Proceedings of the National Academy of Sciences, USA, 104, 19691–19696. Hussein J (1983) A review of methods for determining available water capacities of soils and description of an improved method for estimating field capacity. Zimbabwe Journal of Agricultural Research, 21, 73–87. Hussein J (1987) Agro-climatological analysis of growing season in natural regions lll, lV and V of Zimbabwe. Cropping in the semi-arid areas of Zimbabwe. Proceedings of a workshop held in Harare 24th to 28th August 1987. (ed. Agritex), pp. 25–189. Agritex/DR&SS/GTZ-GART, Harare, Zimbabwe. IPCC (2013) Summary for policymakers. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (eds Stocker TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM), pp. 1–27. Cambridge University Press, Cambridge. Keating BA, Carberry PS, Hammer GL et al. (2003) An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy, 18, 267–288. Liu Z, Hubbard KG, Lin X, Yang X (2013) Negative effects of climate warming on maize yield are reversed by the changing of sowing date and cultivar selection in Northeast China. Global Change Biology, 19, 3481–3492. Lobell DB, Burke MB, Tebaldi C, Mastrandrea MD, Falcon WP, Naylor RL (2008) Prioritizing climate change adaptation needs for food security in 2030. Science, 319, 607–610. Lobell DB, B€anziger M, Magorokosho C, Vivek B (2011) Nonlinear heat effects on African maize as evidenced by historical yield trials. Nature Climate Change, 1, 42–45. Ludwig F, Asseng S (2006) Climate change impacts on wheat production in a Mediterranean environment in Western Australia. Agricultural Systems, 90, 159–179. Mapfumo P, Adjei-Nsiah S, Mtambanengwe F, Chikowo R, Giller KE (2013) Participatory action research (PAR) as an entry point for supporting climate change adaptation by smallholder farmers in Africa. Environmental Development, 5, 6–22. Moss RH, Edmonds JA, Hibbard KA et al. (2010) The next generation of scenarios for climate change research and assessment. Nature, 463, 747–756.
ulated maize yield variability in Zimbabwe. Agricultural and Forest Meteorology, 90, 39–50. Piani C, Weedon GP, Best M, Gomes SM, Viterbo P, Hagemann S, Haerter JO (2010) Statistical bias correction of global simulated daily precipitation and temperature for the application of hydrological models. Journal of Hydrology, 395, 199–215. Raes D, Sithole A, Makarau A, Milford J (2004) Evaluation of first planting dates recommended by criteria currently used in Zimbabwe. Agricultural and Forest Meteorology, 125, 177–185. Reidsma P, Ewert F, Boogaard H, Diepen KV (2009) Regional crop modelling in Europe: the impact of climatic conditions and farm characteristics on maize yields. Agricultural Systems, 100, 51–60. Rockstro¨m J (2004) Making the best of climatic variability: options for upgrading rainfed farming in water scarce regions. Water Science and Technology, 49, 151–156. Rosenzweig C, Parry ML (1994) Potential impact of climate change on world food supply. Nature, 367, 133–138. Ruane AC, Cecil LD, Horton RM et al. (2013) Climate change impact uncertainties for maize in Panama: farm information, climate projections, and yield sensitivities. Agricultural and Forest Meteorology, 170, 132–145. Rurinda J, Mapfumo P, van Wijk MT, Mtambanengwe F, Rufino MC, Chikowo R, Giller KE (2013) Managing soil fertility to adapt to rainfall variability in smallholder cropping systems in Zimbabwe. Field Crops Research, 154, 211–225. Rurinda J, Mapfumo P, van Wijk MT, Mtambanengwe F, Rufino MC, Chikowo R, Giller KE (2014) Sources of vulnerability to a variable and changing climate among smallholder households in Zimbabwe: a participatory analysis. Climate Risk Management, 3, 65–78. Shongwe ME, Van Oldenborgh GJ, Van Den Hurk BJJM, De Boer B, Coelho CAS, Van Aalst MK (2009) Projected changes in mean and extreme precipitation in Africa under global warming. Part I: Southern Africa. Journal of Climate, 22, 3819–3837. Shumba EM (1989) An agronomic study of appropriate maize (Zea mays) tillage and weed control technologies in a communal area of Zimbabwe. DPhil. Thesis. University of Zimbabwe, Zimbabwe. Shumba EM, Waddington SR, Rukuni M (1992) Use of tine-tillage, with atrazine weed control, to permit earlier planting of maize by smallholder farmers in Zimbabwe. Experimental Agriculture, 28, 443–452. Springate DA, Kover PX (2014) Plant responses to elevated temperatures: a field study on phenological sensitivity and fitness responses to simulated climate warming. Global Change Biology, 20, 456–465. Tadross M, Suarez P, Lotsch A et al. (2009) Growing-season rainfall and scenarios of future change in southeast Africa: implications for cultivating maize. Climate Research, 40, 147–161. Taylor SH, Ripley BS, Martin T, De-Wet L-A, Woodward FI, Osborne CP (2014) Physiological advantages of C4 grasses in the field: a comparative experiment demonstrating the importance of drought. Global Change Biology, 20, 1992–2003. Tingem M, Rivington M, Bellocchi G (2009) Adaptation assessments for crop production in response to climate change in Cameroon. Agronomy for Sustainable Development, 29, 247–256. Tittonell P, Giller KE (2013) When yield gaps are poverty traps: the paradigm of ecological intensification in African smallholder agriculture. Field Crops Research, 143, 76–90. Traore B (2014) Climate change, climate variability and adaptation options in smallholder cropping systems of the Sudano-Sahel region in West Africa. PhD Thesis. Wageningen University, Wageningen, The Netherlands. Tubiello FN, Ewert F (2002) Simulating the effects of elevated CO2 on crops: approaches and applications for climate change. European Journal of Agronomy, 18, 57–74. Turner NC, Rao KPC (2013) Simulation analysis of factors affecting sorghum yield at selected areas in eastern and southern Africa, with emphasis on increasing temperatures. Agricultural Systems, 121, 53–62. Unganai LS (1996) Historic and future climatic change in Zimbabwe. Climate Research, 6, 137–145. Waddington SR, Hlatshwayo M (1991) Agronomic Monitoring to Determine Causes of Low Yield in Later Planted Maize: Mangwende Communal Area, Zimbabwe: 1987/88 and 1988/89 Seasons. CIMMYT, Harare, Zimbabwe.
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4601 Waha K, M€ uller C, Bondeau A, Dietrich JP, Kurukulasuriya P, Heinke J, Lotze-Campen H (2013) Adaptation to climate change through the choice of cropping
World Soil Resource Base (1998) World Soil Resources Report No. 84. Food and Agriculture Organization, Rome.
system and sowing date in sub-Saharan Africa. Global Environmental Change, 23, 130–143.
Zinyengere N, Crespo O, Hachigonta S (2013) Crop response to climate change in southern Africa: a comprehensive review. Global and Planetary Change, 111, 118–126.
Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. A map showing the location of Makoni and Hwedza in Zimbabwe under the respective dry sub-humid and semi-arid climatic zones of Africa. Map adapted from World Meteorological Organization (WMO), United Nations Environment Programme (UNEP), Climate Change 2001: Impacts, Adaptation, and Vulnerability, Contribution of working group II to the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC). Figure S2. (A) Average change in number of periods of at least five consecutive dry days (cdd) for three climate periods, against the baseline simulated climate data, for two radiative forcings: representaive concentration pathways (rcp) of 4.5 and 8.5 W m 2, in Southern Africa. (B) Average change in length of the largest period of consecutive dry days (cdd) for three climate periods, against the baseline simulated climate data, for two radiative forcings: representative concentration pathways (rcp) of 4.5 and 8.5 W m 2, in Southern Africa. (C) Average change in rainfall intensity (mm) for three climate periods, against the baseline simulated climate data, for two radiative forcings: representative concentration pathways (rcp) of 4.5 and 8.5 W m 2, in Southern Africa. Figure S3. Simulated average maize yields for on-site observed climate data, and five GCMs (CNRM-CM5: atmospheric resolution – 1.9 9 1.9; EC-EARTH: atmospheric resolution-, HADGEM2-ES: atmospheric resolution – 2.5 9 3.75), IPSL-CM5A-LR: atmospheric resolution – 2.5 9 3.75 and MPI-ESM-LR: atmospheric resolution – 1.9 9 1.9) for the baseline climate (1976–2005) and for two radiative forcings: RCP 4.5 and RCP 8.5, for climate period, 2010–2039, for Hwedza. Error bars represent standard deviation. Figure S4. Maize grain yield in response to cultivar, planting date, and fertilization rate for (a) 2009/10 and (b) 2010/11 seasons in Makoni; and for (c) 2009/10 and (d) 2010/11 seasons in Hwedza. Error bars represent SED for a = time of planting, b = fertilization rate, c = crop cultivar. Source: Rurinda et al. (2013). Table S1. Soil physical and chemical properties used for the simulations in APSIM for Hwedza. Soil data for Makoni is not shown because it was similar to that of Hwedza. Table S2. Crop parameters for three maize cultivars used for the simulations in APSIM. Table S3. Description of the two Representative Concentration Pathways (RCPs): RCP8.5 and RCP4.5, and resolutions of the five Global Circulation Models (GCMs), used in the study.
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
Climate change and maize yield in southern Africa: what can farm management do? JAIROS RURINDA1,2,3, MARK T. VAN WIJK1,4, PAUL MAPFUMO2,3, KATRIEN D E S C H E E M A E K E R 1 , I W A N S U P I T 5 and K E N E . G I L L E R 1 1 Plant Production Systems, Wageningen University, PO Box 430, 6700AK Wageningen, The Netherlands, 2Department of Soil Science and Agricultural Engineering, University of Zimbabwe, PO Box MP 167, Mount Pleasant, Harare, Zimbabwe, 3Soil Fertility Consortium for Southern Africa (SOFECSA), University of Zimbabwe, PO Box MP 167, Mount Pleasant, Harare, Zimbabwe, 4International Livestock Research Institute (ILRI), Box 30709, Nairobi 00100, Kenya, 5Earth System Science & Climate Adaptive Land Management, Wageningen University and Research Centre, PO Box 47, 6700 AA Wageningen, The Netherlands
Abstract There is concern that food insecurity will increase in southern Africa due to climate change. We quantified the response of maize yield to projected climate change and to three key management options – planting date, fertilizer use and cultivar choice – using the crop simulation model, agricultural production systems simulator (APSIM), at two contrasting sites in Zimbabwe. Three climate periods up to 2100 were selected to cover both near- and long-term climates. Future climate data under two radiative forcing scenarios were generated from five global circulation models. The temperature is projected to increase significantly in Zimbabwe by 2100 with no significant change in mean annual total rainfall. When planting before mid-December with a high fertilizer rate, the simulated average grain yield for all three maize cultivars declined by 13% for the periods 2010–2039 and 2040–2069 and by 20% for 2070– 2099 compared with the baseline climate, under low radiative forcing. Larger declines in yield of up to 32% were predicted for 2070–2099 with high radiative forcing. Despite differences in annual rainfall, similar trends in yield changes were observed for the two sites studied, Hwedza and Makoni. The yield response to delay in planting was nonlinear. Fertilizer increased yield significantly under both baseline and future climates. The response of maize to mineral nitrogen decreased with progressing climate change, implying a decrease in the optimal fertilizer rate in the future. Our results suggest that in the near future, improved crop and soil fertility management will remain important for enhanced maize yield. Towards the end of the 21st century, however, none of the farm management options tested in the study can avoid large yield losses in southern Africa due to climate change. There is a need to transform the current cropping systems of southern Africa to offset the negative impacts of climate change. Keywords: adaptation, APSIM, fertilizer use, planting date, simulation modelling, Zea mays Received 23 November 2014 and accepted 13 July 2015
Introduction Concern of food insecurity has further increased in southern Africa because of a changing climate. Rising temperatures and changing rainfall patterns are already evident in southern Africa and threaten crop production (Lobell et al., 2008). In Zimbabwe, mean daily maximum temperature has increased by between 0.2 °C and 0.5 °C per decade, over the period of 1962–2000 (Rurinda et al., 2014), and is projected to further increase by 2–4 °C by 2100 (Unganai, 1996) similar to global projections (IPCC, 2013). A large area of southern Africa is projected to experience a decrease in rainfall by 2100 (Shongwe et al., 2009). Neukom et al. (2014) indicated that rainfall already declined in southern Africa between 1796 and 1996. Other studies have Correspondence: Jairos Rurinda, tel./fax +263 04 307304, e-mail: [email protected]
4588
indicated that total rainfall has not changed, but that rainfall patterns are more variable with delays in the onset of the rainy season and more frequent droughts (Tadross et al., 2009; Rurinda et al., 2013). Overall, although rainfall projections contain a large uncertainty and spatial variation, southern Africa is expected to become drier (IPCC, 2013). The increasing temperatures, in combination with more severe and frequent droughts, will profoundly reduce soil water available for plant uptake. Fraser et al. (2013) projected that soil moisture will decline by 25% in southern Africa due to more frequent droughts. Rising temperatures will shorten the crop growth period and increase plant water demand through higher transpiration rates, both potentially reducing plant production (Ludwig & Asseng, 2006; Springate & Kover, 2014). Furthermore, increasing temperatures will directly affect plants through heat waves, and this impact will be large when coupled with soil moisture © 2015 John Wiley & Sons Ltd
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4589 deficits. Lobell et al. (2011) reported that each degree day above 30 °C reduced maize yield by 1.7% under drought, compared with a decrease of 1% under favourable rain-fed conditions in Africa. Given that the impacts of higher temperatures are most pronounced on sandy soils of low water-holding capacity (Ludwig & Asseng, 2006), the predominant soils across smallholder cropping areas in southern Africa (World Soil Resource Base, 1998), smallholder farmers face a high risk of declining crop yields. Climate change may also have a positive influence on crop productivity. Crop yield is anticipated to improve due to increased concentration of carbon dioxide (CO2) in the atmosphere (Ludwig & Asseng, 2006). However, maize, the main staple cereal crop in southern Africa, is a C4 plant, which will benefit relatively little from increased CO2 concentrations (Taylor et al., 2014). Moreover, the poor soil nutrient availability on highly weathered sandy soils, which cover a large area of smallholder farming areas in the region, can reduce any yield benefit of elevated CO2 (Tubiello & Ewert, 2002). It is therefore highly likely that the impacts of increased temperatures coupled with soil moisture deficits will override the compensating effects of increased CO2 on crop yields in southern Africa (Rosenzweig & Parry, 1994). Efforts have been made to understand and quantify the impacts of increased temperatures and changing rainfall patterns on crop productivity in southern Africa (Fischer et al., 2005; Zinyengere et al., 2013). Crop production in Zimbabwe was considered to be determined mainly by the availability of soil water, with temperature not being a major limiting factor (Hussein, 1987). However, using a statistical model, Lobell et al. (2008) predicted that maize production will decline by between 20% and 40% in southern Africa due to a combination of warming temperatures and changing rainfall patterns. These studies quantified the possible effects of climate change on crop productivity, but did not analyse how these effects interact with the opportunities of adaptive farm management such as cultivar choice, timing of farming operations and adjusting soil nutrient inputs. Yet the net impact of climate change on crop yield depends strongly on the interactions between climate and such farm management factors (Reidsma et al., 2009). To provide a comprehensive assessment of climate change effects on crop yields, it is critical to understand the interactions between climate and possible adaptive farm management options. It has been argued that a relatively small change in farm management and selection of different crop varieties can significantly reduce any negative impact of moderate climate change (Howden et al., 2007). Through field experimentation, © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
Rurinda et al. (2013) demonstrated that improved timing of planting and adjusting soil nutrient inputs can stabilize maize yields under variable rainfall conditions in smallholder farming systems in Zimbabwe. By modelling the current cropping systems in sub-Saharan Africa (SSA), Folberth et al. (2013) showed that under irrigation, increasing soil nutrient supply in combination with improved cultivars would allow for a doubling of maize yields. However, the outcome for the same strategies is likely to be different under rain-fed conditions because of the interaction between crop fertilization and rainfall. In the few impact studies that have taken into account farmer adaptive management options (e.g. Waha et al., 2013), the broad scale of assessment makes their findings difficult to translate into knowledge that can drive local solutions. Although there is much debate about the appropriate scale to operate (Challinor et al., 2009), local studies with crop models allow for better calibration and validation (e.g. by taking into account soil nutrient and water conditions), compared with regional approaches (Fischer et al., 2005), and can also help farmers to make appropriate decisions. Further, the impacts of the changing climate on crop yields vary with location due to spatial variability in climate, particularly rainfall. For example, responses of maize in southern Africa to the changing climate can be as wide as 40% to +10% (Zinyengere et al., 2013). In addition, many adaptive farm management options have been identified mainly through a top-down approach (e.g. Phillips et al., 1998). Because of the uncertainties associated with the changing climate (Dessai & van der Sluijs, 2007), and the differences in endowments between farmers (Mtambanengwe & Mapfumo, 2005), a bottom-up approach can be useful for integrating local farmers’ and expert empirical knowledge as well as linking knowledge to action. We quantified the yield response of maize, the main staple crop in southern Africa, to projected climate change and to three key management options, namely adjusting planting dates, adjusting soil nutrient inputs and using different maize cultivars. Farmers believed that these adaptive management options could reduce the negative impacts of the changing climate on maize yields (Mapfumo et al., 2013). We focus especially on the interactions between these factors, to assess how the efficiency of interventions might change in future in southern Africa. We hypothesized that (i) the average yields for maize will decrease with increasing temperatures and changes in rainfall patterns for the future periods, 2010–2039, 2040–2069 and 2070–2099, compared with average yields for the baseline climate; (ii) the response of maize yield to fertilization and planting date will change with climate change; and (iii) this response will be affected by the choice of maize
4590 J . R U R I N D A et al. cultivar. The overall hypothesis was tested that improved management of planting date, fertilization and choice of crop cultivar have the potential to compensate for the predicted decreases in crop yields due to climate change in Zimbabwe.
Materials and methods
Study sites Two districts in eastern Zimbabwe with contrasting climates were studied: Makoni has a dry subhumid climate with annual rainfall ranging from 750 to 1000 mm, and Hwedza has a semi-arid tropical climate with annual rainfall ranging from 650 to 800 mm. The study sites were purposefully selected to represent many smallholder areas in southern Africa (see Fig. S1). Large areas in smallholder farming systems of southern Africa are typified by dry subhumid and semi-arid climates. More than 40% of smallholder agriculture in SSA is conducted in semi-arid regions (Rockstrom, 2004). The two study areas are also marked by increased climate variability and rising temperatures as in many parts of southern Africa (Tadross et al., 2009). Rainfall in Zimbabwe, as in most of southern Africa, falls between October and April with most rainfall in December and February. The farming system is maize-based with cattle providing draught power and manure for crop production, typical of smallholder farming areas in Zimbabwe. In both sites, average monthly temperatures are above 20 °C between October and April, and temperatures are highest in October before the start of the rains. The soils in both sites are granite-derived sandy soils, mainly Lixisols and Arenosols, with little capacity to store moisture (World Soil Resource Base, 1998). These sandy soils are representative for large areas in SSA and in particular southern Africa. Arenosols cover about 13% of SSA and over 6.5 million ha of cropland in southern Africa (Hartemink and Huting, 2008).
Description of the APSIM crop simulation model Maize yield responses to climate, planting date variation and amount of nitrogen applied were simulated using the agricultural production systems simulator (APSIM). APSIM is a process-based, daily time step model developed to simulate biophysical processes in farming systems in response to management decisions and in the face of climatic risk (Holzworth et al., 2014). APSIM has been tested against field experimental data in a wide range of growing conditions across the globe (Keating et al., 2003) and in semi-arid and subhumid regions of southern Africa (Chikowo et al., 2008). APSIM has recently been applied to evaluate risk of crop production in relation to climate variability and change across many areas in, for example, Europe, China and Africa (Ludwig & Asseng, 2006; Liu et al., 2013; Turner & Rao, 2013). We used APSIM version 7.5 (CSIRO Agriculture Flagship, The University of Queensland, Queensland, Australia) to quantify the sensitivity of maize yield to different adaptive farm management options under future predicted values for
daily weather variables, including rainfall, temperature and solar radiation. Daily weather data for baseline and future climates were retrieved from outputs of global circulation models (GCMs), described below. The main APSIM modules used in this study included the plant (maize), environment (meteorological input module, soil water, soil nitrogen and organic matter dynamics, soil phosphorus) and modules management. The soil phosphorus module was included because phosphorus is a limiting nutrient in soils of southern Africa including Zimbabwe. Each APSIM module demands a number of parameters. For the SOILWAT model, which simulates the dynamics of soil water, the input parameters included soil bulk density, soil water lower limit (LL15) and drained upper limit (DUL), and two parameters, U and CONA, which determine first- and second-stage soil evaporation. LL15 and DUL were derived based on soil classification using regression equations calculated by Hussein (1983). Soil moisture content at saturation was estimated from bulk density. The parameters, U and CONA, were set at 6.0 and 3 mm day 1, respectively, values acceptable for tropical conditions (Chikowo et al., 2008). A value of 0.7 was used for SWCON, a coefficient that specifies the proportion of the water in excess of field capacity that drains to the next layer in 1 day (Chikowo et al., 2008). The bare soil run-off curve number was set at 50 to take into account the low runoff associated with sandy soils because of high infiltration rates (Hussein, 1987). For the soil model, the organic matter content for each soil layer was measured in farmers’ fields during the experiments that were conducted in Makoni and Hwedza districts of Zimbabwe, between 2009 and 2012 (Rurinda et al., 2013). The initial soil N was set at 35 kg ha 1 (23 kg NO3 -N and 12 kg NH4+-N) based on field measurements conducted at the same study sites (Mtambanengwe & Mapfumo, 2006). The major soil parameters for one study site, Hwedza, are presented in Table S1. Soil data for Makoni is not shown because it was similar to that of Hwedza. APSIM crop parameters for SC401 (131 days to maturity), SC501 (137 days to maturity) and SC625 (142 days to maturity; see Supporting Information, Table S2) were selected to represent the very early, early and medium maturing cultivars used in the field experiments (Rurinda et al., 2013). The cultivars used in the model and field experiments are available on the market, and they are commonly grown in many areas of southern Africa including in the study areas. During simulations, soil organic matter, nitrogen, phosphorus and water were re-initialized for each growing season at the start of the planting window.
Description of the climate data A useful way of dealing with uncertainties in the future climate is to use a range of possible future climate change scenarios rather than a single projection (Challinor et al., 2009). Two representative concentration pathways (RCPs) – RCP4.5 and RCP8.5 – were the inputs for scenario modelling with an ensemble of five GCMs – CNRM-CM5, EC-EARTH, HADGEM2-ES, IPSL-CM5A-LR and MPI-ESM-LR. RCPs refer to the portion of the concentration pathway extending up to the
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4591 year 2100, for which integrated assessment models produced corresponding emission scenarios (IPCC, 2013). The RCP8.5 is a high concentration pathway for which radiative forcing reaches >8.5 W m 2 by 2100, and the RCP4.5 is an intermediate pathway in which radiative forcing is stabilized at about 4.5 W m 2 (Moss et al., 2010). The radiative forcing trajectories result from diverse combinations of socio-economic, technological, and policy and institutional development scenarios. Using multiple model outputs (ensemble) increases confidence in the projections compared with individual models (IPCC, 2013). The resolution of each GCM and main characteristics of each RCP are presented in Table S3. The climate data were obtained from the coupled model intercomparison project phase 5 (http://pcmdi3.llnl.gov/esgcet/home.htm, last accessed 4 January 2014). The GCM data were regridded to a spatial resolution of 0.5° 9 0.5°. The precipitation, minimum and maximum temperatures data from the 5 GCMs used in the study were biascorrected against the WATCH Forcing Data (see, http:// www.eu-watch.org/watermip/use-of-WATCH-forcing-data) using the method of Piani et al. (2010). The radiation data were bias-corrected with the method of Haddeland (2012). Both bias correction methods used the WATCH forcing data series (1971–2000) as a reference. Many studies have shown that bias-corrected climate data improve impact assessment results because of the adjustment of the simulated data to that of the observation (e.g. Hagemann et al., 2011). Climate model output, however, is affected by inconsistencies which cannot be addressed by bias correction such as the incorrect representations of dynamical and/or physical processes (Haerter et al., 2011). Inherent uncertainties due to internal variability of the climate system also limit the precision with which impacts are assessed (Challinor et al., 2009). Although in general the use of ensembles improves the modelling of climate data, aggregating outputs from several models can mask the effects of growing season dry spells. In this study, we therefore decided to use individual GCM data to run APSIM and afterwards calculate the average crop yields across the five GCMs. Because all five models were driven by the same radiative forcing scenarios, all runs represented an equally possible projection of the future change of the climate. Accordingly, the bias-corrected GCMs daily weather data for temperature, rainfall and solar radiation were used separately in APSIM to simulate maize growth and yield. The simulated yields from each of the five bias-corrected GCM data were then aggregated to provide average yield for each time period. The sensitivity of maize yield was assessed for three future climate periods – 2010–2039, 2040–2069 and 2070–2099 – and compared against a baseline period – 1976–2005. The baseline (1976–2005) maize yield was simulated both with the bias-corrected GCM data and with the observed meteorological point data for Hwedza. A comparison between the two allowed evaluating the appropriateness of using regridded bias-corrected GCM data. Longterm meteorological observed point data for Makoni were incomplete and inconsistent and hence could not be used in the study. The three future climate periods were selected to cover both near-term climate relevant for assessing relatively immediate benefits to agricultural investments (Lobell et al.,
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
2008) and long-term climate for sustainable crop production, illustrating the situation in which climate change is more clearly separated from natural climate variability (Ruane et al., 2013). The simulations did not extend beyond the year 2100 because model and scenario uncertainties increase with lead time (Hawkins and Sutton, 2009).
Adaptive management options identified by farmers and field experimentation for testing APSIM Yield and biomass data used to test APSIM performance were derived from experiments conducted from 2009 to 2011 in farmers’ fields in Hwedza and Makoni districts of Zimbabwe (Rurinda et al., 2013). Three maize cultivars, three planting dates and three fertilization rates were laid out in a split-plot block design with three replications per treatment. Planting date was assigned to the main plot, and fertilization rate 9 maize cultivar subplots were randomized within the main plot. The three maize cultivars were SC 403 (131 days to maturity), SC 513 (137 days to maturity) and SC 635 (142 days to maturity). These represent the range of maize cultivars currently available to farmers and differ only by 11 days in time to maturity. The three planting windows defined together with farmers were 25 October–20 November, 21 November–15 December and 16 December–1 January. The three fertilization rates were a control treatment (unfertilized), low rate (35 kg N ha 1, 14 kg P ha 1, 3 t ha 1 manure, on a dry weight basis) and high rate (90 kg N ha 1, 26 kg P ha 1, 7 t ha 1 manure, on a dry weight basis). The basal fertilizer was applied as compound D (7% N, 14% P2O5 and 7% K2O) and top dressing as ammonium nitrate (34.5%). Rainfall for the 2009/2010 and 2010/2011 seasons was recorded on site. To evaluate the performance of the APSIM crop model, yield and biomass output from observed climate data simulations were compared to observed field experimental data for both Hwedza and Makoni study areas for the 2009/2010 and 2010/2011 seasons. Model error was expressed as the rootmean-squared error (RMSE) and the regression coefficient used to evaluate model precision. To assess maize yield response to planting date, the maize cultivars were planted in the model at an interval of 1 week from November 1st to January 10th, representative of the range of potential sowing dates (Rurinda et al., 2013). This was done for the three fertilizer rates as mentioned above. To assess maize yield response to N, increasing rates of ammonium nitrate were applied at 35 days after sowing from 0 up to 120 kg N ha 1 with an interval of 5 kg N ha 1. The maize yield–nitrogen response curves were simulated for each of the three planting windows defined by farmers, and phosphorus was applied at planting at 26 kg ha 1. In all scenarios, the maize cultivars were planted at 44 000 plants ha 1 in each season following 20 mm of rainfall in three consecutive days. If the rainfall condition was not met in a particular week (in the case maize response to planting date) or farmerdefined planting window (in the case maize response to N), the crop was forced to be planted at the end of the week or window.
4592 J . R U R I N D A et al. Fig. 1 Observed and predicted maize yield for (a) 2009/2010 season (root-mean-squared error, RMSE = 0.42, R2 = 0.88) and (b) 2010/ 2011 season (RMSE = 0.20, R2 = 0.70); biomass for (c) 2009/2010 season (RMSE = 0.75; R2 = 0.86) and (d) 2010/2011 season (RMSE = 0.91, R2 = 0.73), for all three adaptive management options. Three planting dates (early planting – represented by the green colour; normal planting – dark yellow; and late planting – dark red); three fertilization rates (zero fertilization – represented by crossed shapes; low rate: 14 kg P ha 1, 35 kg N ha 1, 3 t ha 1 manure – semi-filled shapes; and high rate: 26 kg P ha 1, 90 kg N ha 1, 7 t ha 1 manure – full-filled shapes) and three cultivars (SC401 – represented by a circle; SC501 – square; and SC625 – diamond), for (A) Makoni and (B) Hwedza.
Results
Performance of APSIM model APSIM performed well in predicting yields for maize planted early or during the normal window, for all fertilization rates and cultivars in the relatively good 2009/2010 rainfall season at both sites (RMSE < 0.5, R2 > 0.8; Fig. 1Aa, Be). Similarly, the model predicted the biomass reasonably well in the 2009/2010 season in both study areas (Fig. 1Ac, Bg). In the 2010/2011 season characterized by waterlogging conditions and prolonged dry spells, APSIM over-predicted both yield and biomass, especially for the nutrient-depleted soils in Hwedza (Fig. 1Bf, Bh). The discrepancies in the performance of APSIM with season’s rainfall quality suggest that the model does not perform well under waterlogging conditions.
Projected temperature and rainfall conditions by 2100 The temperature is projected to increase significantly in the future, and this increase will be stronger with the high radiative forcing scenario at both Makoni and Hwedza (Fig. 2Ad–f, Bj–l). Compared with the baseline climate, the minimum temperature is projected to increase by between 3.2 °C and 5.4 °C and maximum temperature by between 3.2 °C and 4.9 °C for the time horizon 2070–2099 for all five GCMs, under the high radiative forcing of 8.5 W m 2 for Hwedza (Fig. 2Bl). The greatest increase in temperature is projected for the time horizon 2070–2099, under the high radiative forcing of 8.5 W m 2 (Fig. 2Af, Bl). The direction of possible change in total rainfall is less clear for all three future time periods and for each radiative forcing (Fig. 3Aa– f, Bg–l). Coinciding trends such as increased length of the largest period of consecutive dry days and increased rainfall intensity (see Supporting Information, Fig. S2A–C), observed in different time horizons, can nullify signals in the mean total rainfall trend. Although the mean annual rainfall change signal is unclear by 2100, rainfall variability is highly likely to increase (Fig. S2A–C).
Consequences of the changing climate and its interaction with adaptive management options on maize yield Impact of maize cultivar choice on yield under varied scenarios of future climate. The baseline maize yield simulated with the (1976–2005) bias-corrected GCM data was similar to that simulated with on-site observed historical climate data (Fig. 4a, b). Yield responses for all three maize cultivars to future changes in climate as well as to adaptive management of planting date and soil nutrient input were similar (data not shown). In other words, the differences in yield between cultivars were negligible, even under future change in climate in both study areas. Compared with the baseline climate when planting before mid-December with a high fertilizer rate in Hwedza, the simulated average grain yield for all three maize cultivars declined by an average of 13% for the time horizons 2010–2039 and 2040–2069, and by 18% for 2070–2099 with radiative forcing of 4.5 W m 2 (Fig. 5Bg). With the same management, but under radiative forcing of 8.5 W m 2, the simulated grain yield in Hwedza declined by an average of 13% for the time horizon 2010–2039, 18% for 2040– 2069 and 32% for 2070–2099 (Fig. 5Bj). Climate change effects on maize were small when little fertilizer was used at both study areas (Fig. 5Bh,k). Thus, the greatest maize yield loss is projected towards the end of the 21st century mostly under the high radiative forcing of 8.5 W m 2 when a high rate of fertilizer is applied. Although a greater yield loss was simulated for Hwedza, the trends in yield changes were similar with that in Makoni, a region that generally receives more rainfall (Fig. 5Aa–f). Delayed planting effects on maize yield under varied scenarios of future climate. The simulated average maize yield increased gradually with planting date from early November to mid-December at both study sites (Fig. 5Aa,b,d,e, Bg,h,j,k). After mid-December, simulated yields declined drastically. Overall, the maize yield response to planting date was similar from November to mid-December for all three future climate periods for each radiative forcing (Fig. 5). Under zero
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4593
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
4594 J . R U R I N D A et al.
Fig. 2 Projected maximum and minimum temperatures for three climate change periods: near-term (2010–2039), mid-term (2040–2069) and long-term (2070–2099), against simulated baseline temperatures (solid lines), for two radiative forcings: RCP 4.5 and RCP 8.5, for (A) Makoni and (B) Hwedza. The box and whisker plots show the temperature variation based on the ensembles of five global circulation models. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4595
Fig. 3 Projected daily rainfall in each month for three climate change periods: near-term (2010–2039), mid-term (2040–2069) and longterm (2070–2099), against simulated baseline rainfall (solid lines), for two radiative forcings: RCP 4.5 and RCP 8.5, for (A) Makoni and (B) Hwedza. The box and whisker plots show the rainfall variation based on the ensembles of five global circulation models. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
4596 J . R U R I N D A et al.
Fig. 4 Baseline average maize yield (1976–2005) obtained with varying planting date for three maize cultivars (SC401, SC501 and SC625) under three fertilization rates (CR – zero fertilization; LR – low rate: 35 kg N ha 1, 14 kg P ha 1, 3 t ha 1 manure; and HR – high rate: 90 kg N ha 1, 26 kg P ha 1, 7 t ha 1 manure), for (a) on-site measured historical climate data and (b) bias-corrected global circulation model data, for Hwedza.
fertilization, no effect of planting date was observed (Fig. 5Ac,f, Bi,l). Responses of maize yield to adjustments in the amount of nitrogen under a changing climate. Fertilizer increased yield significantly under both the baseline and future climates particularly when planting before mid-December compared to the control (zero fertilization; e.g. Fig. 5Aa, Bg). However, after mid-December, the impact of fertilizer on yield was drastically decreased (e.g. Fig. 5Aa, Bg). The response of maize to increased nitrogen applied decreased for all three climate periods for each radiative forcing, compared with the baseline climate. For example, at 80 kg N ha 1, a maize yield of about 4.5 t ha 1 was simulated for the baseline climate, against maize yields of about 3.5 t ha 1 for 2010–2069 and 3 t ha 1 for 2070–2099 under the RCP4.5 at both study sites (Fig. 6Aa–d, Bi–l). Maize yield responses to nitrogen were comparable between climate periods 2010–2039 and 2040–2069 for each radiative forcing (Fig. 6Ab,c, Bj,k). Maize yield responses to nitrogen further decreased with increasing climate change with time. The smallest yield response to nitrogen was simulated for the climate period 2070–2099 and for the high radiative forcing of 8.5 W m 2. For example, applying 60 kg N ha 1 resulted in a maximum yield benefit of only about 2.5 t ha 1 for the period 2070–2099 for RCP8.5 (Fig. 6Bp).
Discussion Our analyses suggest that the average yields for maize in Zimbabwe will decrease with the increasing temper-
atures predicted for the current century, although the yield decline will be relatively small ( 13%) until the middle of the century particularly for the low greenhouse gas emission scenario. Larger yield losses ( 32%) are projected for the more distant future, that is towards the end of the century, and under high fertilization rate for the high radiative forcing scenario of 8.5 W m 2. Although the yield trend for the two sites studied was similar, the impacts of climate change on maize yield were more pronounced for Hwedza, a semi-arid region characterized by higher temperatures and increased rainfall variability (Rurinda et al., 2014). The decline in future yields was driven mainly by increasing temperatures that increased the crop maturation rate and hence shortened the crop growing period (Springate & Kover, 2014). This means less crop biomass can be assimilated, and hence, total crop biomass is reduced, which directly results in yield reduction. Using a statistical model and for a shorter time horizon, that is 2030, Lobell et al. (2008) projected much larger losses of maize yield (< 30%) in southern Africa, mainly due to increased temperatures and decreased rainfall. Suboptimal crop management, studied here in relation to application of mineral fertilizer and planting date of maize, has much larger negative impacts on maize production in the near future. In particular, the nonlinear effect of delay in planting date on maize production is important, because it suggests that farmers have quite some flexibility in choosing their planting date up to mid-December. This finding is contrary to the existing literature in which delays in planting date are often directly associated with lower yields (e.g.
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4597
Fig. 5 Average seasonal yield distribution with planting date for maize cultivar SC401 under three fertilization rates (CR – zero fertilization; LR – low rate: 35 kg N ha 1, 14 kg P ha 1, 3 t ha 1 manure; HR – high rate: 90 kg N ha 1, 26 kg P ha 1, 7 t ha 1 manure), for the simulated baseline climate (1976–2005) and in response to climate change for three periods: near-term (2010–2039), mid-term (2040–2069) and long-term (2070–2099) and for two radiative forcings: RCP 4.5 and RCP 8.5, for (A) Makoni and (B) Hwedza. Data for the other two cultivars (SC501 and SC625) are not presented because the yields for all three cultivars were similar. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
4598 J . R U R I N D A et al.
Fig. 6 Simulated average maize yield for three maize cultivars planted early: SC401, SC501 and SC625, in response to nitrogen fertilization for three future climate periods: 2010–2039, 2040–2069 and 2070–2099, against the simulated baseline climate, 1976–2005, for two radiative forcings: RCP 4.5 and RCP 8.5, for (A) Makoni and (B) Hwedza. The standard deviation (SD) presented is for one cultivar (SC401) because the standard deviation for all three cultivars was similar.
Waddington & Hlatshwayo, 1991) and widely applied ‘rules-of-thumb’ which suggest a 2.3% decline in grain yield per day delay in planting (Shumba, 1989) over the
period of October to mid-December. Our results even suggest a slight yield increase with small delays in planting, which would also hold under a changing © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4599 climate (see e.g. Fig. 5Aa). Delays in planting are caused mainly by lack of draught power, farm labour and fertilizer or farmers’ perceptions of rainfall patterns (Shumba et al., 1992). The similarity in simulated yields until planting dates after mid-December reinforced the findings from our earlier experimental studies that there is no yield difference between crops planted early (25 October–20 November) or during what farmers considered to be a normal planting window (21 November–15 December; see Fig. S3; Rurinda et al., 2013). Pushing for very early planting will not bring much yield benefit under a changing climate, while the risk of losing the crop might increase due to increased rainfall variability during the start of the growing season (Raes et al., 2004). Similarly, in other regions of Africa, changing sowing dates during the crop growing period is projected to be ineffective in counteracting adverse climatic effects because of the narrow window of rainfall that affects the timing of farm operations (Tingem et al., 2009). Towards the end of the 21st century, none of the farm management options considered in this study can avoid large yield losses of more than 30%. In particular, the maize response to mineral nitrogen is projected to decline significantly as the climate changes. Although fertilizer use will remain a key strategy to enhance crop production also under climate change, if crops are not planted too late (see Fig. 5; Rurinda et al., 2013), maize yields might plateau at smaller mineral nitrogen rates than under current climate (see e.g. Fig. 6Bp). Our simulation results indicate that farmers who can afford to buy more fertilizers could reduce their application rate to about 80 kg N ha 1 by mid-century and 60 kg N ha 1 towards the end of the century, against the current recommendation rate of 120 kg N ha 1 in Zimbabwe to increase economic returns. Rurinda et al. (2013) demonstrated that the use of smaller amounts of fertilizers would not compromise household food selfsufficiency in smallholder farming systems. However, if fertilizer use is reduced, farmers will not be able to maximize production suggesting that investment outside crop production will be critical to enhance household income and food security. The yields of all three cultivars, which represent the maturity range of the cultivars currently available in much of southern Africa (http://seeds.seedco.co/zimbabwe, accessed 20 July 2015), are likely to respond similarly to the predicted future climates. The range in time to maturity of these cultivars is small, with a difference between the relatively short (SC403) and long (SC635) duration cultivars of only 11 days. The yield potential of these cultivars is claimed to vary between 4 and 9 t ha 1 (see, © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
Table S2), but we found no yield difference tested in the field under farmers management (Rurinda et al., 2013). Given that farmers are now experiencing delayed onset of rainfall with no change in the date of the end of rainy season (Rurinda et al., 2013), breeding for shorter maturing cultivars with greater tolerance to more frequent droughts could help to stabilize future crop production. By contrast, Traore (2014) found that short-duration varieties seemed more vulnerable to predicted increases in temperature than long-duration varieties in West Africa. The model analysis of Traore (2014) suggests that the growing period of the short-duration maize varieties is shortened due to increased temperatures, so that overall biomass and grain production is reduced. This could mean that although shortduration varieties are perceived to be more climate robust because they are not dependent on a long growing season with ample rainfall, the temperature response of such varieties might counteract this positive characteristic. More detailed crop experimentation is needed to investigate how these interactions work out in reality compared with model analyses. However, new cultivars will only help smallholder farmers to adapt to climate change if other binding constraints to crop production such as poor soil fertility are first removed (Tittonell & Giller, 2013). Overall, our results indicate that in the short-term improved farm management practices can buffer the negative impacts of climate change. Yet in the longer term, climate change will further increase food insecurity of smallholder farmers in southern Africa given that none of the current farm management practices, including fertilization, planting date adjustments and cultivar choice, can avoid large yield loss. The limitation of the current farm management options to compensate for the huge yield loss due to long-term climate change suggests that there is need to transform the cropping systems in southern Africa to offset the negative impacts of climate change and build a climate resilient livelihood system. Opportunities are also needed for farmers to explore livelihood options outside agriculture.
Acknowledgements We thank the International Development Research Centre (IDRC) and Department for International Development (DFID) for funding through the Climate Change Adaptation in Africa (CCAA) Grant #104140 to the University of Zimbabwe. Climate data were provided through the Combine project (European Commission’s 7th Framework Programme Grant Agreement No. 226520) and the IPOP Sustainable and Smart Food Supply programme (Wageningen University, the Netherlands). We would also like to thank the two anonymous reviewers whose comments greatly improved the manuscript.
4600 J . R U R I N D A et al.
References
Mtambanengwe F, Mapfumo P (2005) Organic matter management as an underlying cause for soil fertility gradients on smallholder farms in Zimbabwe. Nutrient
Challinor AJ, Ewert F, Arnold S, Simelton E, Fraser E (2009) Crops and climate
Cycling in Agroecosystems, 73, 227–243. Mtambanengwe F, Mapfumo P (2006) Effects of organic resource quality on soil profile N dynamics and maize yields on sandy soils in Zimbabwe. Plant and Soil, 281, 173–191. Neukom R, Nash DJ, Endfield GH et al. (2014) Multi-proxy summer and winter precipitation reconstruction for southern Africa over the last 200 years. Climate Dynamics, 42, 2713–2726. Phillips JG, Cane MA, Rosenzweig C (1998) ENSO, seasonal rainfall patterns and sim-
change: progress, trends, and challenges in simulating impacts and informing adaptation. Journal of Experimental Botany, 60, 2775–2789. Chikowo R, Corbeels M, Tittonell P, Vanlauwe B, Whitbread A, Giller KE (2008) Aggregating field-scale knowledge into farm-scale models of African smallholder systems: summary functions to simulate crop production using APSIM. Agricultural Systems, 97, 151–166. Dessai S, van der Sluijs J (2007) Uncertainty and Climate Change Adaptation – A Scoping Study. Netherlands Environmental Assessment Agency, Utrecht. Fischer G, Shah M, Tubiello FN, Van Velhuizen H (2005) Socio-economic and climate change impacts on agriculture: an integrated assessment, 1990–2080. Philosophical Transactions of the Royal Society B: Biological Sciences, 360, 2067–2083. Folberth C, Yang H, Gaiser T, Abbaspour KC, Schulin R (2013) Modeling maize yield responses to improvement in nutrient, water and cultivar inputs in sub-Saharan Africa. Agricultural Systems, 119, 22–34. Fraser EDG, Simelton E, Termansen M, Gosling SN, South A (2013) “Vulnerability hotspots”: integrating socio-economic and hydrological models to identify where cereal production may decline in the future due to climate change induced drought. Agricultural and Forest Meteorology, 170, 195–205. Haddeland I, Heinke J, Voß F, Eisner S, Chen C, Hagemann S, Ludwig F (2012) Effects of climate model radiation, humidity and wind estimates on hydrological simulations. Hydrology and Earth System Sciences, 16, 305–318. Haerter JO, Hagemann S, Moseley C, Piani C (2011) Climate model bias correction and the role of timescales. Hydrology and Earth System Sciences, 15, 1065–1079. Hagemann S, Chen S, Haerter JO, Heinke J, Gerten D, Piani C (2011) Impact of a statistical bias correction on the projected hydrological changes obtained from three GCMs and two hydrology models. Journal of Hydrometeorology, 12, 556–578. Hartemink AE, Huting J (2008) Land cover, extent, and properties of Arenosols in Southern Africa. Arid Land Research and Management, 22, 134–147. Hawkins E, Sutton R (2009) The Potential to Narrow Uncertainty in Regional Climate Predictions. Bulletin of the American Meteorological Society, 90, 1095–1107. Holzworth DP, Huth NI, deVoil PG et al. (2014) APSIM – evolution towards a new generation of agricultural systems simulation. Environmental Modelling & Software, 62, 327–350. Howden SM, Soussana JF, Tubiello FN, Chhetri N, Dunlop M, Meinke H (2007) Adapting agriculture to climate change. Proceedings of the National Academy of Sciences, USA, 104, 19691–19696. Hussein J (1983) A review of methods for determining available water capacities of soils and description of an improved method for estimating field capacity. Zimbabwe Journal of Agricultural Research, 21, 73–87. Hussein J (1987) Agro-climatological analysis of growing season in natural regions lll, lV and V of Zimbabwe. Cropping in the semi-arid areas of Zimbabwe. Proceedings of a workshop held in Harare 24th to 28th August 1987. (ed. Agritex), pp. 25–189. Agritex/DR&SS/GTZ-GART, Harare, Zimbabwe. IPCC (2013) Summary for policymakers. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (eds Stocker TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM), pp. 1–27. Cambridge University Press, Cambridge. Keating BA, Carberry PS, Hammer GL et al. (2003) An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy, 18, 267–288. Liu Z, Hubbard KG, Lin X, Yang X (2013) Negative effects of climate warming on maize yield are reversed by the changing of sowing date and cultivar selection in Northeast China. Global Change Biology, 19, 3481–3492. Lobell DB, Burke MB, Tebaldi C, Mastrandrea MD, Falcon WP, Naylor RL (2008) Prioritizing climate change adaptation needs for food security in 2030. Science, 319, 607–610. Lobell DB, B€anziger M, Magorokosho C, Vivek B (2011) Nonlinear heat effects on African maize as evidenced by historical yield trials. Nature Climate Change, 1, 42–45. Ludwig F, Asseng S (2006) Climate change impacts on wheat production in a Mediterranean environment in Western Australia. Agricultural Systems, 90, 159–179. Mapfumo P, Adjei-Nsiah S, Mtambanengwe F, Chikowo R, Giller KE (2013) Participatory action research (PAR) as an entry point for supporting climate change adaptation by smallholder farmers in Africa. Environmental Development, 5, 6–22. Moss RH, Edmonds JA, Hibbard KA et al. (2010) The next generation of scenarios for climate change research and assessment. Nature, 463, 747–756.
ulated maize yield variability in Zimbabwe. Agricultural and Forest Meteorology, 90, 39–50. Piani C, Weedon GP, Best M, Gomes SM, Viterbo P, Hagemann S, Haerter JO (2010) Statistical bias correction of global simulated daily precipitation and temperature for the application of hydrological models. Journal of Hydrology, 395, 199–215. Raes D, Sithole A, Makarau A, Milford J (2004) Evaluation of first planting dates recommended by criteria currently used in Zimbabwe. Agricultural and Forest Meteorology, 125, 177–185. Reidsma P, Ewert F, Boogaard H, Diepen KV (2009) Regional crop modelling in Europe: the impact of climatic conditions and farm characteristics on maize yields. Agricultural Systems, 100, 51–60. Rockstro¨m J (2004) Making the best of climatic variability: options for upgrading rainfed farming in water scarce regions. Water Science and Technology, 49, 151–156. Rosenzweig C, Parry ML (1994) Potential impact of climate change on world food supply. Nature, 367, 133–138. Ruane AC, Cecil LD, Horton RM et al. (2013) Climate change impact uncertainties for maize in Panama: farm information, climate projections, and yield sensitivities. Agricultural and Forest Meteorology, 170, 132–145. Rurinda J, Mapfumo P, van Wijk MT, Mtambanengwe F, Rufino MC, Chikowo R, Giller KE (2013) Managing soil fertility to adapt to rainfall variability in smallholder cropping systems in Zimbabwe. Field Crops Research, 154, 211–225. Rurinda J, Mapfumo P, van Wijk MT, Mtambanengwe F, Rufino MC, Chikowo R, Giller KE (2014) Sources of vulnerability to a variable and changing climate among smallholder households in Zimbabwe: a participatory analysis. Climate Risk Management, 3, 65–78. Shongwe ME, Van Oldenborgh GJ, Van Den Hurk BJJM, De Boer B, Coelho CAS, Van Aalst MK (2009) Projected changes in mean and extreme precipitation in Africa under global warming. Part I: Southern Africa. Journal of Climate, 22, 3819–3837. Shumba EM (1989) An agronomic study of appropriate maize (Zea mays) tillage and weed control technologies in a communal area of Zimbabwe. DPhil. Thesis. University of Zimbabwe, Zimbabwe. Shumba EM, Waddington SR, Rukuni M (1992) Use of tine-tillage, with atrazine weed control, to permit earlier planting of maize by smallholder farmers in Zimbabwe. Experimental Agriculture, 28, 443–452. Springate DA, Kover PX (2014) Plant responses to elevated temperatures: a field study on phenological sensitivity and fitness responses to simulated climate warming. Global Change Biology, 20, 456–465. Tadross M, Suarez P, Lotsch A et al. (2009) Growing-season rainfall and scenarios of future change in southeast Africa: implications for cultivating maize. Climate Research, 40, 147–161. Taylor SH, Ripley BS, Martin T, De-Wet L-A, Woodward FI, Osborne CP (2014) Physiological advantages of C4 grasses in the field: a comparative experiment demonstrating the importance of drought. Global Change Biology, 20, 1992–2003. Tingem M, Rivington M, Bellocchi G (2009) Adaptation assessments for crop production in response to climate change in Cameroon. Agronomy for Sustainable Development, 29, 247–256. Tittonell P, Giller KE (2013) When yield gaps are poverty traps: the paradigm of ecological intensification in African smallholder agriculture. Field Crops Research, 143, 76–90. Traore B (2014) Climate change, climate variability and adaptation options in smallholder cropping systems of the Sudano-Sahel region in West Africa. PhD Thesis. Wageningen University, Wageningen, The Netherlands. Tubiello FN, Ewert F (2002) Simulating the effects of elevated CO2 on crops: approaches and applications for climate change. European Journal of Agronomy, 18, 57–74. Turner NC, Rao KPC (2013) Simulation analysis of factors affecting sorghum yield at selected areas in eastern and southern Africa, with emphasis on increasing temperatures. Agricultural Systems, 121, 53–62. Unganai LS (1996) Historic and future climatic change in Zimbabwe. Climate Research, 6, 137–145. Waddington SR, Hlatshwayo M (1991) Agronomic Monitoring to Determine Causes of Low Yield in Later Planted Maize: Mangwende Communal Area, Zimbabwe: 1987/88 and 1988/89 Seasons. CIMMYT, Harare, Zimbabwe.
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
C L I M A T E C H A N G E A D A P T A T I O N O F M A I Z E - B A S E D S Y S T E M S 4601 Waha K, M€ uller C, Bondeau A, Dietrich JP, Kurukulasuriya P, Heinke J, Lotze-Campen H (2013) Adaptation to climate change through the choice of cropping
World Soil Resource Base (1998) World Soil Resources Report No. 84. Food and Agriculture Organization, Rome.
system and sowing date in sub-Saharan Africa. Global Environmental Change, 23, 130–143.
Zinyengere N, Crespo O, Hachigonta S (2013) Crop response to climate change in southern Africa: a comprehensive review. Global and Planetary Change, 111, 118–126.
Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. A map showing the location of Makoni and Hwedza in Zimbabwe under the respective dry sub-humid and semi-arid climatic zones of Africa. Map adapted from World Meteorological Organization (WMO), United Nations Environment Programme (UNEP), Climate Change 2001: Impacts, Adaptation, and Vulnerability, Contribution of working group II to the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC). Figure S2. (A) Average change in number of periods of at least five consecutive dry days (cdd) for three climate periods, against the baseline simulated climate data, for two radiative forcings: representaive concentration pathways (rcp) of 4.5 and 8.5 W m 2, in Southern Africa. (B) Average change in length of the largest period of consecutive dry days (cdd) for three climate periods, against the baseline simulated climate data, for two radiative forcings: representative concentration pathways (rcp) of 4.5 and 8.5 W m 2, in Southern Africa. (C) Average change in rainfall intensity (mm) for three climate periods, against the baseline simulated climate data, for two radiative forcings: representative concentration pathways (rcp) of 4.5 and 8.5 W m 2, in Southern Africa. Figure S3. Simulated average maize yields for on-site observed climate data, and five GCMs (CNRM-CM5: atmospheric resolution – 1.9 9 1.9; EC-EARTH: atmospheric resolution-, HADGEM2-ES: atmospheric resolution – 2.5 9 3.75), IPSL-CM5A-LR: atmospheric resolution – 2.5 9 3.75 and MPI-ESM-LR: atmospheric resolution – 1.9 9 1.9) for the baseline climate (1976–2005) and for two radiative forcings: RCP 4.5 and RCP 8.5, for climate period, 2010–2039, for Hwedza. Error bars represent standard deviation. Figure S4. Maize grain yield in response to cultivar, planting date, and fertilization rate for (a) 2009/10 and (b) 2010/11 seasons in Makoni; and for (c) 2009/10 and (d) 2010/11 seasons in Hwedza. Error bars represent SED for a = time of planting, b = fertilization rate, c = crop cultivar. Source: Rurinda et al. (2013). Table S1. Soil physical and chemical properties used for the simulations in APSIM for Hwedza. Soil data for Makoni is not shown because it was similar to that of Hwedza. Table S2. Crop parameters for three maize cultivars used for the simulations in APSIM. Table S3. Description of the two Representative Concentration Pathways (RCPs): RCP8.5 and RCP4.5, and resolutions of the five Global Circulation Models (GCMs), used in the study.
© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 4588–4601
