The IPCC (Intergovernmental Panel on Climate Change) assessment reports confirm that climate change will hit developing countries the hardest. Adaption is on the agenda of many countries around the world. However, before devising adaption strategies, it is crucial to assess and understand the impacts of climate change at regional and local scales. In this study, the impact of climate change on rain-fed maize (Zea mays) production in the Wami-Ruvu basin of Tanzania was evaluated using the Decision Support System for Agro-technological Transfer. The model was fed with daily minimum and maximum temperatures, rainfall and solar radiation for current climate conditions (1971–2000) as well as future climate projections (2010–2099) for two Representative Concentration Pathways: RCP 4.5 and RCP 8.5. These data were derived from three high-resolution regional climate models, used in the Coordinated Regional Climate Downscaling Experiment program. Results showed that due to climate change future maize yields over the Wami-Ruvu basin will slightly increase relative to the baseline during the current century under RCP 4.5 and RCP 8.5. However, maize yields will decline in the mid and end centuries. The spatial distribution showed that high decline in maize yields are projected over lower altitude regions due to projected increase in temperatures in those areas.

INTRODUCTION

Climate change studies including the latest Intergovernmental Panel on Climate Change (IPCC) assessment reports consistently show that many of the world's regions are already experiencing increasing frequency and severity of droughts and floods, increasing inter-annual and inter-seasonal rainfall variability and higher temperatures (IPCC 2007, 2013). Developing countries are particularly vulnerable to climate change, especially extreme weather and climate events, due to their high dependence on rain-fed agriculture and natural resources for their livelihoods (IPCC 2007, 2012, 2013; Akram 2012; Sarker et al. 2012). The IPCC (2007) assessment report projected that between 75 and 250 million people in Africa will be exposed to increased water stress due to climate change by 2020. In some countries, yields from rain-fed agriculture are projected to decline by 50%. Agricultural production, including food, in many African countries will be severely compromised (IPCC 2007). This would further adversely affect food security and exacerbate malnutrition.

African countries need to prioritize the formulation of adaptation strategies and mitigation policy in order to prevent the destructive impacts of future climate change on agriculture (Rowhani et al. 2011). Tanzania will need to seriously consider its adaptation strategies, since agriculture is one the largest sectors of its economy (Ahmed et al. 2011). It is essential to ensure that there are adequate and credible adaptation strategies in this sector. However, in order to begin determining adaptation strategies for the agriculture sector, it is necessary to have credible assessment of climate change impacts on crop production to produce evidence to guide the formulation of adaptation policy.

Previous studies in Tanzania (Mwandosya et al. 1998; Agrawala et al. 2003; Ehrhart & Twena 2006; Enfors & Gordon 2008; Thornton et al. 2009, 2010; Ahmed et al. 2011; Arndt et al. 2011; Müller et al. 2011) have assessed the impact of climate change on crop production using climate simulation derived directly from general circulation models (GCMs). These models classically run at spatial resolution of 300 km or more such that their climate change projections cannot reproduce climate details at local or region scales where farming practices are predominant. The coarse resolution of GCMs severely limits the direct use of their output in regional and sub-regional decision-making or in impact studies (Masson & Knutti 2011; Daniels et al. 2012). This limitation is particularly strict for countries like Tanzania with high regional climate heterogeneity, influenced by heterogeneous topography.

The limitations of direct use of GCM outputs for decision-making at regional and sub-regional scales or in impact studies call into question the many prior assessments of climate change impact on crop productions based on GCMs. Furthermore, adaptation strategies and mitigation policy developed using GCM simulations are not realistic and might pose significant challenges for anticipatory adaptation in many countries.

In this study, we assess both the impact of climate change on maize (Zea mays) production over the Wami-Ruvu basin of Tanzania using high resolution climate data from regional climate models (RCMs) and the uncertainties associated with the assessment from individual RCMs and driving GCMs. The main aim is to update the results from previous studies that were based on coarse resolution climate data from the GCMs. Maize is the most important cereal crop for financial value and food security in Tanzania (Rowhani et al. 2011; Washington & Pearce 2012), and therefore deserves special attention in studies of how future climate will affect its production. The Wami-Ruvu basin of Tanzania was chosen because the basin contains many agriculture projects that aim to develop southern Tanzania, into a major regional food producer and engine of national economic development (Milder et al. 2012). Moreover, to the best knowledge of authors, there is no climate change impact study addressing the knowledge problem in rain-fed areas that have used high resolution climate information.

DATA AND METHODOLOGY

Study area

Tanzania, which is located in East Africa between latitudes 1 °S and 12 °S and longitudes 29 °E to 41 °E, has nine basins (URT 2013). The Wami-Ruvu basin, covering an area of about 66,820 km2, is located in six regions: Dar es Salaam, parts of Coast, Morogoro, Dodoma, Tanga and Manyara in eastern Tanzania along latitudes 5°–7°S and longitudes 36°–39°E. The Wami-Ruvu basin is covered by low lying and mountainous landscapes (Figure 1); the topographical features are described in detail in URT (2013). The dominant mountain landscape includes the Uluguru Mountains (altitude above mean sea level 400 m to 2,500 m), Nguru Mountains (400 to 2,000 m), Rubeho Mountains (500 to 1,000 m), Ukaguru Mountains (400 to 1,000 m) and Nguu Mountains located in the western part of Wami River (400 to 2,000 m). The low-lying areas include Mkata plains (400 to 800 m), Lower Wami (200 to 400 m), Kisaki located south east of the Uluguru Mountains (140 to 200 m) and Kimbiji and Mbezi located in the southern coastal area of Dar es Salaam (50 to 100 m).
Figure 1

Topographical map of the Wami-Ruvu basin (adapted from the Global Water for Sustainability Program), available at: http://dpanther.fiu.edu/dpService/glowsProjectServices/project/iWASH%20(Tanzania)# (accessed on 13.05.2016).

Figure 1

Topographical map of the Wami-Ruvu basin (adapted from the Global Water for Sustainability Program), available at: http://dpanther.fiu.edu/dpService/glowsProjectServices/project/iWASH%20(Tanzania)# (accessed on 13.05.2016).

The Wami-Ruvu basin is divided into three sub-basins: Wami sub-basin which covers an area of 43,946 km2, the Ruvu sub-basin with an area of 18,078 km2 and Coastal sub-basin with an area of 4,796 km2. These sub-basins are further divided into seven sub-catchments: Kinyasungwe, Mkondoa, Wami, Upper Ruvu, Ngerengere, Lower Ruvu and Coast sub-catchments (Figure 2).
Figure 2

Maps of Africa and Tanzania showing the location of the study area, the Wami-Ruvu basin and its seven sub-catchments.

Figure 2

Maps of Africa and Tanzania showing the location of the study area, the Wami-Ruvu basin and its seven sub-catchments.

The climate over the Wami-Ruvu basin is mainly controlled by movement of the Inter Tropical Convergence Zone (ITCZ): some areas within the basin receive unimodal rainfall (due to a single passage of the ITCZ), while others receive bimodal rainfall (due to two passages of the ITCZ). The former is experienced in the Wami sub-basin (Kinyasungwe sub-catchment), while the latter is experienced in the Wami sub-basin (Mkondoa and Wami sub-catchments) and Ruvu (upper and lower Ruvu), Ngerengere and Coast sub-catchments.

Although the ITCZ is the dominant driver of rainfall within the basin, rainfall distribution is possibly influenced by orographic features. For example, in the plains annual rainfall ranges from 800 mm to 1,000 mm near the Coast and 500 to 600 mm inland towards Dodoma and north of the Wami sub-basin (Kashaigili 2011). Over the high ground, in the Uluguru Mountains, the annual rainfall exceeds 1,500 mm (Mwandosya et al. 1998). Annually, the eastern slopes of the Uluguru Mountains receive the highest rainfall of 2,500 to 4,000 mm while the western slopes receive 1,200 to 3,100 mm (Mbwanga 2005). Annual rainfall is between 800 and 1,200 mm in the Nguru-Rubeho Mountains and 1,000 and 1,800 mm in the Ukaguru Mountains (Kashaigili 2011).

The Wami-Ruvu basin is characterized by 12 soil types: Cambisols, Ferralsols, Acrisols, Fluvisols, Luvisols, Lixisols, Arenosols, Leptosols, Nitisols, Vertisols, Planosols and Haplic Phaeozems (Figure 3) (URT 2013). The dominant soils are Cambisols which cover parts of Bagamoyo, Kisarawe, Mkuranga, Morogoro Rural, Dodoma Urban, Bahi and Chamwino. These types of soils make good agricultural land and are used intensively for agriculture production.
Figure 3

Major soil groups in the Wami-Ruvu basin (adapted from URT 2013).

Figure 3

Major soil groups in the Wami-Ruvu basin (adapted from URT 2013).

Farming activities in the Wami-Ruvu basin are shaped by two agro-ecologies: semi-arid and sub-humid. The former covers parts of Dodoma region and the latter parts of Morogoro, Tanga and Coast regions. Crop production in the basin is mainly subsistence. However, in recent years, the basin has witnessed many agriculture projects, which are outlined in URT (2012), including the FEED the FUTURE (investment cost USD 300 million), Tanzania Bread-Basket Transformation Project (investment cost USD 173 million), Southern Agriculture Corridor of Tanzania (investment cost USD 3.4 billion) and Rural Livelihoods Development Programme (investment cost USD 21 million).

Data

Data from RCM simulations

This study makes use of high resolution climate simulation from the Coordinated Regional Climate Downscaling Experiment program regional climate models (CORDEX_RCMs). The CORDEX program is archiving outputs from a set of RCM simulations over different regions in the world. Figure 4 indicates the CORDEX domain for model integrations. Data sets from CORDEX Africa are accessed from http://cordexesg.dmi.dk/esgf-web-fe/. These data sets are quality controlled and may be used according to the terms of use at http://wcrp-cordex.ipsl.jussieu.fr/. The spatial grid resolutions of all CORDEX_RCMs are set to longitude 0.440 and latitude 0.440 using rotated pole system coordinates where the model operates over an equatorial domain with a quasi-uniform resolution of approximately 50 km by 50 km. For detailed description of CORDEX RCMs and their dynamics and physical parameterization the reader may consult Nikulin et al. (2012). The CORDEX_RCMs and their driving GCMs used here are listed in Table 1.
Figure 4

CORDEX regional domain (adapted from Pechlivanidis 2013).

Figure 4

CORDEX regional domain (adapted from Pechlivanidis 2013).

Table 1

Details of CORDEX-RCMs and the driving GCMs

No RCM Model Centre Short name of RCM Short name GCM 
DMI HIRHAM5 Danish Meteorological Institute HIRHAM5 ICHEC-EC-EARTH 
SMHI Rossby Swedish Meteorological and Hydrological Institute (SMHI) RCA4 MPI-M-MPI-ESM-LR 
Center Regional ICHEC-EC-EARTH 
Atmospheric CNRM-CERFACS- 
Model (RCA4) CNRM-CM5 
KNMI Regional Royal Netherlands Meteorological Institute (KNMI) RACMO22T ICHEC-EC-EARTH 
Atmospheric Climate 
Model, version 2.2 
(RACMO2.2T) 
No RCM Model Centre Short name of RCM Short name GCM 
DMI HIRHAM5 Danish Meteorological Institute HIRHAM5 ICHEC-EC-EARTH 
SMHI Rossby Swedish Meteorological and Hydrological Institute (SMHI) RCA4 MPI-M-MPI-ESM-LR 
Center Regional ICHEC-EC-EARTH 
Atmospheric CNRM-CERFACS- 
Model (RCA4) CNRM-CM5 
KNMI Regional Royal Netherlands Meteorological Institute (KNMI) RACMO22T ICHEC-EC-EARTH 
Atmospheric Climate 
Model, version 2.2 
(RACMO2.2T) 

Daily values of minimum and maximum temperatures, rainfall and solar radiation from CORDEX_RCMs for two Representative Concentration Pathway (RCP) scenarios (RCP 4.5) and (RCP 8.5) for the periods 1971–2000, 2010–2039, 2040–2069 and 2070–2099 were used. Since climate models simulate climate variables at grids, the interpolation technique of inverse square distance weighting average is used to transfer the model grid climate simulation to the location where farming is carried out. This enables use of site-specific climate information to simulate maize yields. For detailed descriptions of the interpolation technique used, the reader may consult Hartkamp et al. (1999) and Ly et al. (2013). Interpolated daily minimum and maximum temperatures, rainfall and solar radiations from individual CORDEX_RCMs are used by crop model to simulate maize growth, development and yields.

Data for soil profiles and management practices

Data for 20 soil profiles were used in this study: eight were excavated within the study region and 12 were obtained from the Africa soil profiles database. The hydrological characteristics of layers for each soil profile were estimated using a soil water properties calculator (Saxton & Rawls 2009), where input parameters were soil type (sand, silt or clay) and organic matter, and the output parameters are drained lower limit (SLLL; cm3/cm3), drained upper limit (SLDUL; cm3/cm3), saturation (SLSAT) and water content for each soil layer (Table 2).

Table 2

Soil physical and chemical characteristics at Mvomero research station, Morogoro, Tanzania

Soil depth (cm) Lower limit (cm3/cm3Upper limit (cm3/cm3SAT (cm3/cm3BD (g/cm3pH NO3 (ugN/g) NH4 (ugN/g) ORG (%) 
0–5 0.116 0.253 0.432 1.1 5.4 0.45 0.15 0.24 
5–15 0.116 0.253 0.432 1.1 5.4 0.45 0.15 0.24 
15–30 0.116 0.31 0.452 1.1 5.2 0.45 0.15 0.24 
30–45 0.116 0.31 0.452 1.1 5.2 0.45 0.15 0.24 
45–67 0.131 0.31 0.468 1.2 0.45 0.15 0.11 
67–90 0.131 0.31 0.468 1.2 0.45 0.15 0.11 
Soil depth (cm) Lower limit (cm3/cm3Upper limit (cm3/cm3SAT (cm3/cm3BD (g/cm3pH NO3 (ugN/g) NH4 (ugN/g) ORG (%) 
0–5 0.116 0.253 0.432 1.1 5.4 0.45 0.15 0.24 
5–15 0.116 0.253 0.432 1.1 5.4 0.45 0.15 0.24 
15–30 0.116 0.31 0.452 1.1 5.2 0.45 0.15 0.24 
30–45 0.116 0.31 0.452 1.1 5.2 0.45 0.15 0.24 
45–67 0.131 0.31 0.468 1.2 0.45 0.15 0.11 
67–90 0.131 0.31 0.468 1.2 0.45 0.15 0.11 

SAT, saturation.

BD, bulk density.

ORG, organic matter.

Management practices and actual and previous yield information, types of fertilizers were obtained from a comprehensive household panel survey database (National Bureau of Statistics 2012) (Table 3). Information about planting and harvesting dates, planting density and the type of maize cultivars used on each farm were obtained from interviews conducted across the study region. This information was used to create crop model input data files.

Table 3

Detailed information about the farms used to create crop model data input files

Location No. farms Planting date Planting month Planting density (plants/m2Planting spacing (cm) Nitrogen (kg/ha) Maize yields (kg/ha) By-product (kg/ha) Planting window (mm/dd) 
Dodoma 15 1–30 12–1 75 50 683 12/01–12/15 
Kongwa 13 1–30 12–1 100 40 603 12/01–12/15 
Mlali 16 2–28 2–3 75 571 02/02–02/16 
Morogoro 2–31 3–3 75 40 736 03/02–03/16 
Mlali–village 11 3–29 12–1 100 40 1,229 12/03–12/17 
Ukaguru 15 2–28 2–3 75 50 871 02/02–02/16 
Nondoto 11 2–30 12–1 75 979 12/02–12/16 
Chibwangula 10 3–26 12–1 75 609 12/03–12/17 
Idifu 2–30 12–1 100 222 12/02–12/16 
Kiberashi 11 1–23 2–3 75 800 02/01–02/15 
Wami–tuliani 16 1–23 3–3 77 30 996 03/01–03/15 
Kilosa 11 5–31 3–3 75 10 1,224 03/05–03/19 
Ulaya 11 7–25 12–1 80 10 960 12/07–12/21 
Mbwewe 1–28 2–3 75 40 1,287 02/01–02/15 
Kibakwe 15–9 12–1 100 40 2,167 12/15–12/29 
Grand total 168       861  
Location No. farms Planting date Planting month Planting density (plants/m2Planting spacing (cm) Nitrogen (kg/ha) Maize yields (kg/ha) By-product (kg/ha) Planting window (mm/dd) 
Dodoma 15 1–30 12–1 75 50 683 12/01–12/15 
Kongwa 13 1–30 12–1 100 40 603 12/01–12/15 
Mlali 16 2–28 2–3 75 571 02/02–02/16 
Morogoro 2–31 3–3 75 40 736 03/02–03/16 
Mlali–village 11 3–29 12–1 100 40 1,229 12/03–12/17 
Ukaguru 15 2–28 2–3 75 50 871 02/02–02/16 
Nondoto 11 2–30 12–1 75 979 12/02–12/16 
Chibwangula 10 3–26 12–1 75 609 12/03–12/17 
Idifu 2–30 12–1 100 222 12/02–12/16 
Kiberashi 11 1–23 2–3 75 800 02/01–02/15 
Wami–tuliani 16 1–23 3–3 77 30 996 03/01–03/15 
Kilosa 11 5–31 3–3 75 10 1,224 03/05–03/19 
Ulaya 11 7–25 12–1 80 10 960 12/07–12/21 
Mbwewe 1–28 2–3 75 40 1,287 02/01–02/15 
Kibakwe 15–9 12–1 100 40 2,167 12/15–12/29 
Grand total 168       861  

Crop model

Crop models simulate crop growth by numerical integration of constituent processes with the aid of computers (Graves et al. 2002). They are used to study crop growth and calculate growth responses to environmental changes. There are different types of crop models that can be classified as descriptive and explanatory. The descriptive model simulates the behaviour of the system (plant organs and processes) in a simple way (Miglietta & Bindi 1993), where experimental data are used to develop one or more mathematical equations to describe the system. This type of model is suitable only when a quick look is required to describe the behaviour of a crop under field conditions when conditions remain relatively stable. Explanatory crop models describe quantitatively the mechanisms and processes that cause the behaviour of the system. To develop an explanatory crop model, a system is analysed and its processes and mechanisms are quantified separately. An example of explanatory crop model is the Decision Support System for Agro-technological Transfer (DSSAT). This model has been commonly used worldwide over the past two decades because it is reasonably accurate, process-oriented, simple to use and requires minimum data sets. The model requires daily solar radiation, minimum and maximum temperatures and rainfall. The other input parameters of the model include soil properties, crop characteristics and management practices such as planting and harvesting dates, row spacing, plant population, irrigation amount, fertilizer rate and date of application. We used DSSAT version 4.5 to simulate maize yields over the Wami-Ruvu basin of Tanzania. This version has 28 different crop models and new tools to facilitate creation and management of experimental, soil and weather data files. DSSAT version 4.5 has new application programs for analysing crop rotation to assess economic risks and environmental impacts. For instance, impacts from irrigation, fertilizer and nutrient management, climate variability, climate change, soil carbon sequestration and precision management.

MODEL INPUT FILES

A new protocol of the Agricultural Model inter-comparison and Improvement Project (AgMIP) was used. This protocol has new computing tools for crop modelling and has capacity to handle data from multiple sites through different software, such as Data overlay for Multi-model Export that contains data for field overlay, and the AgMIP Crop Experiment (ACE) database that contains site-based crop experimental or farm survey results. The data are translated into format ready for crop modelling using QuadUI desktop utility. For detailed information about AgMIP protocol the reader may visit www.agmip.org. The data preparation was done by creating three Excel files: (1) FIELD ALL containing all information from each farm/field, i.e. planting dates, soil profiles, type and amount of fertilizer used per farm, plant population, plant row spacing and weather stations used; (2) FIELD OVERLAY containing information about soil organic carbon, soil water and nitrogen contents, types of cultivars (in this study SITUKA maize was used due to its drought tolerance and use by many farmers in the study area); (3) SEASONAL STRATEGIES containing information about the span of the year for which simulation is carried out, concentration of atmospheric carbon dioxide, automatic planting date and windows. In this study, the planting date was set to start automatically within the planting window only if cumulative rainfall reached 25 mm on three consecutive days. These Excel files are translated by QuadUI desktop utility to DSSAT folder. From this folder simulation was initiated by running DSSAT45.BAT file.

Model calibration and validation

The Crop–Environment–Resource–Synthesis (CERES)-Maize model is incorporated within DSSAT 4.5 to simulate maize yields as influenced by several factors. The calibration and validation of CERES-Maize model to obtain reasonable estimates of model genetic coefficient was performed by comparing simulated and observed maize yields data at 168 farms (Table 3). The calibration was robust since there was good agreement between simulated and observed yields with coefficient of determination (R2) of 0.91. For detailed description on how the model was calibrated and validated, a reader may consult Mourice et al. (2014).

Simulation of maize yield

Using the calibrated CERES maize model, maize yields simulation was carried out with historical climate data from individual RCMs for the baseline period 1971–2000 and future projections: 2010–2039, 2040–2069 and 2070–2099 for RCP 4.5 and RCP 8.5 scenarios. In order to address properly the uncertainties introduced by the climate models in simulating maize yields, an ensemble average of three CORDEX_RCMs driven by three different GCMs was constructed for RCP 4.5 and RCP 8.5 and used to force the CERES maize model to simulate maize yields for baseline period (1971–2000), near future (2010–2039) and mid (2040–2069) and end centuries (2070–2099).

Spatial interpolation of climate variables, length of growing seasons and maize yields

The spatial distribution of length of growing season, maize yields, seasonal minimum (TN) and maximum (TX) temperatures, rainfall and solar radiation was estimated using the inverse distance weighting (IDW) interpolation technique. This is a deterministic interpolation method and produces interpolated values within the range of input values. The influence of input point data on interpolated values is isotropic since the influence on the input data point is distance related (Philip & Watson 1986). The results from this method may not represent the desired surface when the sampling input points are sparse or uneven (Watson & Philip 1985). This method works in such a way that the characteristics of the interpolated surface can be controlled by limiting the input data points used in the calculation of each output. This can be done using a variable or fixed search method. Under the variable search method, an interpolated value is calculated even if there are not enough points within the maximum distance search radius to meet the number of points criterion; the points within the maximum distance are used. If the minimum point required for estimation of unsampled data point is unavailable, the search radius is increased until the required number of points is satisfied. The interpolated points used were from 15 farms distributed in the basin almost evenly.

RESULTS

The impact of climate change on maize production in the Wami-Ruvu basin of Tanzania was assessed using a dynamic crop growth model CERES-maize embedded within DSSAT version 4.5. The model was run under fixed atmospheric carbon dioxide concentration (360 ppm). This was to limit simulation of other crop processes such as photosynthesis that can affect yields. The change in maize yields simulated here are due to change in climate variables which are derived from the climate models during the present, mid and end centuries under RCP 4.5 and RCP 8.5 scenarios. The climate projections from the RCMs driven by GCMs are used to drive the CERES maize model. It is important to note that all presented results are for the main growing season in the study area, i.e. from December to early June (Table 3).

Temporal averaged maize yields for the baseline period (1971–2000)

The temporal averaged maize yields over Wami-Ruvu basin simulated by CERES forced by climate data from different RCM-GCM combinations for the baseline period (1971–2000) are presented in Tables 48. These tables show that average maize yields over Wami-Ruvu basin for the baseline period (1971–2000) differ when CERES is forced with different RCM-GCM sets. The highest maize yield of 1,303 kg/ha is simulated by CERES forced with RACMO22T-ICHEC and the lowest maize yield of 951 kg/ha is simulated by CERES forced with RCA4-CNRM. CERES simulates maize yields over Wami-Ruvu basin differently even when forced with RACMO22T and HIRHAM5, both driven by the same GCM. Mean maize yields of 1,303 kg/ha and 1,158 kg/ha are simulated by CERES forced with RACMO22T and HIRHAM5 respectively. This variation is mainly due to differences in RACMO22T and HIRHAM5 formulation.

Table 4

Maize yields and seasonal climatic variable as simulated by crop model fed with climate data from RACMO-RCM for RCP 4.5 and RCP 8.5

  1971–2000
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 1,303 1,274a −2.2a 1,266a −2.8a 1,275a −2.2a 
1,273b −2.3b 1,259b −3.4b 1,240b −4.8b 
Length of growing season (days) 130 121a −6.9a 114a −12.3a 111a −14.6a 
121b −6.9b 110b −15.4b 101b −22.3b 
Seasonal maximum temp (°C) 24 25a 4.2a 25a 4.2a 26a 8.3a 
25b 4.2b 26b 8.3b 27b 12.5b 
Seasonal minimum temp (°C) 15 16a 6.7a 17a 13.3a 17a 13.3a 
16b 6.7b 17b 13.3b 19b 26.7b 
Seasonal solar radiation (MJ/m2/day) 19 19a 0a 19a 0a 19a 0a 
19b 0b 18b −5.3b 18b 5.3b 
Total rainfall (mm) 425 406a −4.5a 403a −5.2a 404a −4.9a 
418b −1.7b 407b −4.2b 392b −7.8b 
Total evapotranspiration (mm) 366 354a −3.3a 355a −3a 355a −3a 
358b −2.2b 357b −2.5b 343b −6.3b 
  1971–2000
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 1,303 1,274a −2.2a 1,266a −2.8a 1,275a −2.2a 
1,273b −2.3b 1,259b −3.4b 1,240b −4.8b 
Length of growing season (days) 130 121a −6.9a 114a −12.3a 111a −14.6a 
121b −6.9b 110b −15.4b 101b −22.3b 
Seasonal maximum temp (°C) 24 25a 4.2a 25a 4.2a 26a 8.3a 
25b 4.2b 26b 8.3b 27b 12.5b 
Seasonal minimum temp (°C) 15 16a 6.7a 17a 13.3a 17a 13.3a 
16b 6.7b 17b 13.3b 19b 26.7b 
Seasonal solar radiation (MJ/m2/day) 19 19a 0a 19a 0a 19a 0a 
19b 0b 18b −5.3b 18b 5.3b 
Total rainfall (mm) 425 406a −4.5a 403a −5.2a 404a −4.9a 
418b −1.7b 407b −4.2b 392b −7.8b 
Total evapotranspiration (mm) 366 354a −3.3a 355a −3a 355a −3a 
358b −2.2b 357b −2.5b 343b −6.3b 

aRCP 4.5.

bRCP 8.5.

Table 5

Maize yields and seasonal climatic variable as simulated by crop model fed with climate data from HIRHAM5-RCM for RCP 4.5 and RCP 8.5

  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % Change Mean % change Mean % change 
Maize-yield (kg/h) 1,158 1,132a −2.2a 1,119a −3.4a 1,123a −3.0a 
1,152b −0.5b 1,111b −4.1b 1,080b −6.7b 
Length of growing season (days) 110 104a −5.5a 100a −9.1a 98a −10.9a 
104b −5.5b 97b −11.8b 90b −18.2 
Seasonal maximum temp (°C) 25 26a 4.0a 27a 8.0a 27a 8.0a 
26b 4.0b 28b 12.0b 29b 16.0b 
Seasonal minimum temp (°C) 18 19a 5.6a 20a 11.1a 20a 11.1a 
19b 5.6b 20b 11.1b 22b 22.2b 
Seasonal solar radiation (MJ/m2/day) 16 17a 6.3a 17a 6.3a 17a 6.3a 
16b 0.0b 17b 6.3b 17b 6.3b 
Total rainfall (mm) 356 345a −3.1a 373a 4.8a 373a 4.8a 
340b −4.5b 358b 0.6b 396b 11.2b 
Total evapotranspiration (mm) 259 255a −1.5a 257a −0.8a 260a 0.4a 
255b −1.5b 257b −0.8b 258b −0.4b 
  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % Change Mean % change Mean % change 
Maize-yield (kg/h) 1,158 1,132a −2.2a 1,119a −3.4a 1,123a −3.0a 
1,152b −0.5b 1,111b −4.1b 1,080b −6.7b 
Length of growing season (days) 110 104a −5.5a 100a −9.1a 98a −10.9a 
104b −5.5b 97b −11.8b 90b −18.2 
Seasonal maximum temp (°C) 25 26a 4.0a 27a 8.0a 27a 8.0a 
26b 4.0b 28b 12.0b 29b 16.0b 
Seasonal minimum temp (°C) 18 19a 5.6a 20a 11.1a 20a 11.1a 
19b 5.6b 20b 11.1b 22b 22.2b 
Seasonal solar radiation (MJ/m2/day) 16 17a 6.3a 17a 6.3a 17a 6.3a 
16b 0.0b 17b 6.3b 17b 6.3b 
Total rainfall (mm) 356 345a −3.1a 373a 4.8a 373a 4.8a 
340b −4.5b 358b 0.6b 396b 11.2b 
Total evapotranspiration (mm) 259 255a −1.5a 257a −0.8a 260a 0.4a 
255b −1.5b 257b −0.8b 258b −0.4b 

aRCP 4.5.

bRCP 8.5.

Table 6

Maize yields and seasonal climatic variable as simulated by crop model fed with climate data from RCA4-ICHEC-RCM for RCP 4.5 and RCP 8.5

  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 1,126 1,120a −0.5a 1,109a −1.5a 1,118a −0.7a 
1,118b −0.7b 1,087b −3.5b 1,061b −5.8b 
Length of growing season (days) 112 106a −5.4a 101a −9.8a 99a −11.6a 
105b −6.3b 98b −12.5b 91b −18.8b 
Seasonal maximum temp (°C) 27 27a 0.0a 28a 3.7a 28a 3.7a 
27b 0.0b 28b 3.7b 30b 11.1b 
Seasonal minimum temp (°C) 16 17a 6.3a 18a 12.5a 19a 18.8a 
18b 12.5b 19b 18.8b 21b 31.3b 
Seasonal solar radiation (MJ/m2/day) 21 21a 0.0a 21a 0.0a 20a −4.8a 
21b 0.0b 21b 0.0b 20b −4.8b 
Total rainfall (mm) 349 372a 6.6a 372a 6.6a 382a 9.5a 
377b 8.0b 386b 10.6b 418b 19.8b 
Total evapotranspiration (mm) 280 288a 2.9a 284a 1.4a 290a 3.6a 
284b 1.4b 286b 2.1b 292b 4.3b 
  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 1,126 1,120a −0.5a 1,109a −1.5a 1,118a −0.7a 
1,118b −0.7b 1,087b −3.5b 1,061b −5.8b 
Length of growing season (days) 112 106a −5.4a 101a −9.8a 99a −11.6a 
105b −6.3b 98b −12.5b 91b −18.8b 
Seasonal maximum temp (°C) 27 27a 0.0a 28a 3.7a 28a 3.7a 
27b 0.0b 28b 3.7b 30b 11.1b 
Seasonal minimum temp (°C) 16 17a 6.3a 18a 12.5a 19a 18.8a 
18b 12.5b 19b 18.8b 21b 31.3b 
Seasonal solar radiation (MJ/m2/day) 21 21a 0.0a 21a 0.0a 20a −4.8a 
21b 0.0b 21b 0.0b 20b −4.8b 
Total rainfall (mm) 349 372a 6.6a 372a 6.6a 382a 9.5a 
377b 8.0b 386b 10.6b 418b 19.8b 
Total evapotranspiration (mm) 280 288a 2.9a 284a 1.4a 290a 3.6a 
284b 1.4b 286b 2.1b 292b 4.3b 

aRCP 4.5.

bRCP 8.5.

Table 7

Maize yields and seasonal climatic variable as simulated by crop model fed with climate data from RCA4-MPI-RCM for RCP 4.5 and RCP 8.5

  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 1,212 1,150a −5.1a 1,117a −7.8a 1,083a −10.6a 
1,122b −7.4b 1,087b −10.3b 1,032b −14.9b 
Length of growing season (days) 108 102a −5.6a 97a −10.2a 95a −12.0a 
100b −7.4b 93b −13.9b 86b −20.4b 
Seasonal maximum temp (°C) 27 28a 3.7a 29a 7.4a 29a 7.4a 
28b 3.7b 29b 7.4b 31b 14.8b 
Seasonal minimum temp (°C) 17 18a 5.9a 19a 11.8a 20a 17.6a 
19b 11.8b 20b 17.6b 22b 29.4b 
Seasonal solar radiation (MJ/m2/day) 21 20a −4.8a 21a 0.0a 21a 0.0a 
21b 0.0b 20b −4.8b 20b −4.8b 
Total rainfall (mm) 415 473a  14.0a 434a 4.6a 430a 3.6a 
425b 2.4b 439b 5.8b 443b 6.7b 
Total evapotranspiration (mm) 314 318a 1.3a 304a −3.2a 307a −2.2a 
307b −2.2b 307b −2.2b 302b −3.8b 
  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 1,212 1,150a −5.1a 1,117a −7.8a 1,083a −10.6a 
1,122b −7.4b 1,087b −10.3b 1,032b −14.9b 
Length of growing season (days) 108 102a −5.6a 97a −10.2a 95a −12.0a 
100b −7.4b 93b −13.9b 86b −20.4b 
Seasonal maximum temp (°C) 27 28a 3.7a 29a 7.4a 29a 7.4a 
28b 3.7b 29b 7.4b 31b 14.8b 
Seasonal minimum temp (°C) 17 18a 5.9a 19a 11.8a 20a 17.6a 
19b 11.8b 20b 17.6b 22b 29.4b 
Seasonal solar radiation (MJ/m2/day) 21 20a −4.8a 21a 0.0a 21a 0.0a 
21b 0.0b 20b −4.8b 20b −4.8b 
Total rainfall (mm) 415 473a  14.0a 434a 4.6a 430a 3.6a 
425b 2.4b 439b 5.8b 443b 6.7b 
Total evapotranspiration (mm) 314 318a 1.3a 304a −3.2a 307a −2.2a 
307b −2.2b 307b −2.2b 302b −3.8b 

aRCP 4.5.

bRCP 8.5.

Table 8

Maize yields and seasonal climatic variable as simulated by crop model fed with climate data from RCA4-CNRM for RCP 4.5 and RCP 8.5

  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 951 947a −0.4a 915a −3.8a 909a   − 4.4a 
962b 1.2b 893b −6.1b 849b −10.7b 
Length of growing season (days) 105 99a −5.7a 95a −9.5a 91a −13.3a 
99b −5.7b 92b −12.4b 85b −19.0b 
Seasonal maximum temp (°C) 28 29a 3.6a 30a 7.1a 31a 10.7a 
29b 3.6b 31b 10.7b 32b 14.3b 
Seasonal minimum temp (°C) 17 18a 5.9a 18a 5.9a 19a 11.8a 
18b 5.9b 19b 11.8b 20b 17.6b 
Seasonal solar radiation (MJ/m2/day) 23 23a 0.0a 23a 0.0a 23a 0.0a 
23b 0.0b 23b 0.0b 23b 0.0b 
Total rainfall (mm) 256 274a 7.0a 263a 2.7a 269a 5.1a 
292b 14.1b 264b 3.1b 267b 4.3b 
Total evapotranspiration (mm) 230 233a 1.3a 233a 1.3a 235a 2.2a 
246b 7.0b 232b 0.9b 234b 1.7b 
  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 951 947a −0.4a 915a −3.8a 909a   − 4.4a 
962b 1.2b 893b −6.1b 849b −10.7b 
Length of growing season (days) 105 99a −5.7a 95a −9.5a 91a −13.3a 
99b −5.7b 92b −12.4b 85b −19.0b 
Seasonal maximum temp (°C) 28 29a 3.6a 30a 7.1a 31a 10.7a 
29b 3.6b 31b 10.7b 32b 14.3b 
Seasonal minimum temp (°C) 17 18a 5.9a 18a 5.9a 19a 11.8a 
18b 5.9b 19b 11.8b 20b 17.6b 
Seasonal solar radiation (MJ/m2/day) 23 23a 0.0a 23a 0.0a 23a 0.0a 
23b 0.0b 23b 0.0b 23b 0.0b 
Total rainfall (mm) 256 274a 7.0a 263a 2.7a 269a 5.1a 
292b 14.1b 264b 3.1b 267b 4.3b 
Total evapotranspiration (mm) 230 233a 1.3a 233a 1.3a 235a 2.2a 
246b 7.0b 232b 0.9b 234b 1.7b 

aRCP 4.5.

bRCP 8.5.

CERES simulates maize yields differently when forced by the same RCM driven by different GCMs. It simulates maize yields of 1,126 kg/ha, 1,212 kg/ha and 951 kg/ha when forced with RCA4 driven by ICHEC, MPI and CNRM, respectively. These variations mainly come from differences in formulation of the driving GCMs. Therefore simulation of maize yields using climate data from individual RCM-GCM sets have large uncertainties that come from both RCM and driving GCM.

To account for uncertainties that arise from RCM and driving GCM, the ensemble averages of five climate model members were constructed and used to force CERES. Table 9 shows maize yields simulated by CERES forced with the ensemble averages. These maize yields differ greatly from that of individual models. This is an indication of large uncertainties involved in climate change impact studies particularly for the agriculture sector. However, results from the ensemble averages that take into account the uncertainties from individual RCMs and driving GCMs are the best estimate of future climate change in the basin. Therefore CERES driven by ensemble averages provide reliable estimate of future maize yields in the Wami-Ruvu basin.

Table 9

Maize yields and seasonal climatic variable as simulated by crop model fed with climate data from ENSEMBLE-RCM for RCP 4.5 and RCP 8.5

  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 1,238 1,241a 0.2a 1,229a −0.7a 1,220a −1.5a 
1,257b 1.5b 1,188b −4.0b 1,198b −3.2b 
Length of growing season (days) 112 106a −5.4a 101a −9.8a 99a −11.6a 
105b −6.3b 98b −12.5b 90b −19.6b 
Seasonal maximum temp (°C) 26 27a 3.8a 28a 7.7a 28a 7.7a 
27b 3.8b 28b 7.7b 30b 15.4b 
Seasonal minimum temp (°C) 17 18a 5.9a 18a 5.9a 19a 11.8a 
18b 5.9b 19b 11.8b 21b 23.5b 
Seasonal solar radiation (MJ/m2/day) 20 20a 0.0a 20a 0.0a 20a 0.0a 
20b 0.0b 20b 0.0b 20b 0.0b 
Total rainfall (mm) 349 357a 2.3a 360a 3.2a 357a 2.3a 
357b 2.3b 355b 1.7b 369b 5.7b 
Total evapotranspiration (mm) 375 376a 0.3a 371a −1.1a 375a 0.0a 
375b 0.0b 372b −0.8b 371b −1.1b 
  1971–2005
 
2010–2039
 
2040–2069
 
2070–2099
 
 Mean Mean % change Mean % change Mean % change 
Maize-yield (kg/h) 1,238 1,241a 0.2a 1,229a −0.7a 1,220a −1.5a 
1,257b 1.5b 1,188b −4.0b 1,198b −3.2b 
Length of growing season (days) 112 106a −5.4a 101a −9.8a 99a −11.6a 
105b −6.3b 98b −12.5b 90b −19.6b 
Seasonal maximum temp (°C) 26 27a 3.8a 28a 7.7a 28a 7.7a 
27b 3.8b 28b 7.7b 30b 15.4b 
Seasonal minimum temp (°C) 17 18a 5.9a 18a 5.9a 19a 11.8a 
18b 5.9b 19b 11.8b 21b 23.5b 
Seasonal solar radiation (MJ/m2/day) 20 20a 0.0a 20a 0.0a 20a 0.0a 
20b 0.0b 20b 0.0b 20b 0.0b 
Total rainfall (mm) 349 357a 2.3a 360a 3.2a 357a 2.3a 
357b 2.3b 355b 1.7b 369b 5.7b 
Total evapotranspiration (mm) 375 376a 0.3a 371a −1.1a 375a 0.0a 
375b 0.0b 372b −0.8b 371b −1.1b 

aRCP 4.5.

bRCP 8.5.

The impact of climate change on maize yields over Wami-Ruvu basin

The percentage change in maize yields relative to the baseline over Wami-Ruvu basin for present (2010–2039), mid (2040–2069) and end (2070–2099) centuries under RCP 4.5 and RCP 8.5 is presented in Tables 49. It is clear that the impacts of climate change on future maize yields are not very large compared with the simulated baseline yields. Table 4 presents maize yields when CERES is driven by RACMO22T-ICHEC. From this table, decreases in maize yields by 2.2%, 2.8% and 2.2% are projected in 2010–2039, 2040–2069, and 2070–2099, respectively, under RCP 4.5. Slightly more decrease in maize yields is projected under RCP 8.5 where maximum decrease in maize yields of 4.8% is projected during the end century. These projected changes in maize yields are mainly attributed to increased maximum and minimum temperatures. Hampton et al. (2012) indicates the increase in temperatures reduces pollen viability, reduces seed yields and reduces seed germination. Moore et al. (2012) suggests that increase in temperatures reduce the length of growing season that can either decrease yields (if currently warm) or increase yields (if currently cool).

Tables 4 and 5 allow the analysis of the performance of RCMs in simulating future maize yield over Wami-Ruvu basin whereas CERES forced with RACMO22T and HIRHAM5, both driven by ICHEC GCM, simulates future maize yields differently under RCP 4.5 and 8.5. For instance, decreases in maize yields by 2.2%, 3.4% and 3.0% are projected in 2010–2039, 2040–2069 and 2070–2099, respectively, when CERES was forced with HIRHAM5 under RCP 4.5. This change in maize yields slightly differs from what was observed when CERES was forced with RACMO22T under the same scenario during mid and end centuries. Table 5 shows that maximum decrease in maize yields is projected during the end century under RCP 8.5 with a decrease of 6.7%, whereas the minimum decrease in maize yield is projected during the present century (2010–2039) with a decrease of 0.5% under the same scenario RCP 8.5.

The impact of driving GCMs on simulated maize yield is characterised in Tables 68, where CERES forced with RCA4 driven by three different GCMs simulates maize yields differently. The variations in simulated maize yields are mainly due to differences in GCM formulations.

In general, CERES forced with different RCM-GCM sets simulates maximum decrease in maize yields during the end century under RCP 8.5. Furthermore, projected maize yields over the Wami-Ruvu basin show a consistent decline in present, mid and end centuries, except a slight increase of maize yields is simulated by CERES forced with RCA4-CNRM during the present century under RCP 8.5.

Table 9 shows simulated maize yields when CERES is forced by ensemble averages. From the table, it is clear that increased maize yields are projected during the current century under RCP 4.5 and RCP 8.5 and decreased maize yields are projected during the mid and end centuries under RCP 4.5 and RCP 8.5. Minimum decrease in maize yields of 0.7% is projected during mid century under RCP 4.5 and maximum decrease of 4% is projected during mid century under RCP 8.5. The increase in maize yields during the current century and decrease in maize yields in mid and end centuries is due to slightly increased temperatures and rainfall in the current century and greater increases temperatures during the mid and end centuries that shortened the length of growing season by triggering maturity stages faster.

Spatial distribution of climate variables, length of growing seasons and maize yields

The spatial distribution of minimum temperature (TN), maximum temperature (TX), rainfall, length of growing season and maize yields during the base line period (1971–2000) are presented in Figure 5. This figure shows high temperatures greater than 26°C for TX and 17°C for TN are observed in the lower altitude areas, particularly in the eastern part of the basin. Temperatures are particularly higher over the eastern part of Wami sub-catchment. Lower temperatures are found in high altitude areas in the western side of the basin, particularly over the western part of Wami sub-catchment, and the northern and eastern parts of Kinyasungwe sub-catchment.
Figure 5

The spatial distribution minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during the baseline period (1971–2000) under RCP 4.5.

Figure 5

The spatial distribution minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during the baseline period (1971–2000) under RCP 4.5.

Seasonal rainfall is greater in high than low altitude areas, particularly over the Kinyasungwe sub-catchment, eastern parts of Mkondoa, western parts of Wami and over the upper Ruvu sub-catchments. Meanwhile lower rainfall is found over low than high altitude areas, particularly over Coast, the eastern part of Wami, lower Ruvu and Ngerengere sub-catchments. The length of the growing season is higher in high than in low altitude areas, which have lower temperatures and higher rainfall. Meanwhile shorter growing seasons are found in low altitude areas, which have higher temperatures and lower rainfall. Simulated maize yields are highest in areas with highest rainfall, particularly over southeastern parts of the Kinyasungwe and northwestern parts of Mkondoa sub-catchments.

Figure 6 shows the distribution of changes in temperature, rainfall, maize yields and length of growing season during the present century under RCP 4.5. Highest change in temperatures of 1.8°C for TX and 1.2°C for TN are found over southeastern parts of Mkondoa and over upper Ruvu sub-catchments. Meanwhile lower changes in temperatures of 0.5°C for TX and 0.7°C for TN are found over Kinyasungwe sub-catchment. Decrease in rainfall of 5 mm is found over a large part of Mkondoa, and the eastern part of Ngerengere and southern part of Wami sub-catchments. The greatest decrease in length of growing season is found over the northern and southeastern parts of Mkondoa, western parts of Wami and over upper Ruvu sub-catchments. This is due to increased TX and TN over these areas. Maize yields are projected to decrease by 1 to 5% over northwestern and southeastern parts of Mkondoa, western parts of upper and lower Ruvu, and western and northern parts of Wami sub-catchments. Meanwhile, maize yields over large parts of Mkondoa, Kinyasungwe and Wami sub-catchments are projected to increase by 3 to 9%. Maize yields over Coast and western parts of upper and lower Ruvu are projected to increase by 1 to 3%.
Figure 6

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during the present century (2010–2039) under RCP 4.5.

Figure 6

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during the present century (2010–2039) under RCP 4.5.

Figure 7 shows the distribution of change in temperatures, rainfall, maize yields and the length of growing season during mid century under RCP 4.5. High change in temperatures of 1.8°C to 2.7°C for TX and 1.9°C to 2.2°C for TN are found over upper and lower Ruvu, Mkondoa, Ngerengere, Coast and western parts of Wami sub-catchments. Lower change in temperatures of 0.9°C for TX and 1.5°C for TN are found over Kinyasungwe, southwestern parts of Mkondoa and some eastern parts of Wami sub-catchments. Rainfall is projected to increase by 14 to 49 mm over Coast, southern parts of upper Ruvu, southeastern parts of Mkondoa, lower Ruvu, large parts of Kinyasungwe and Wami sub-catchments. Meanwhile decrease in rainfall of 3 mm is projected over Ngerengere, southern parts of Wami, large parts of Mkondoa and southeastern parts of Kinyasungwe sub-catchments. The greatest decrease in length of growing season is found over upper Ruvu, western parts of Wami and large parts of Mkondoa sub-catchments. Maize yields are projected to decrease from zero to 8% over upper and lower Ruvu, eastern and northern parts of Wami, Ngerengere, southeastern parts of Mkondoa and western parts of coast sub-catchments. Meanwhile maize yields over large parts of Mkondoa, Kinyasungwe, southern parts of Wami and eastern parts of coast sub-catchments are projected to increase by 2 to 10%. This increase is mainly due to increased rainfall and a small decrease in length of growing season or a small increase in TX and TN over those areas. Generally the eastern part of the basin is projected to have decreased maize yield. This is due to the high increase in TX and TN that reduced the length of the growing season. It is important to note that some areas, such as north of Mkondoa, will benefit from projected temperature increase, as maize yields continue to increase. Meanwhile other areas such as upper Ruvu, Ngerengere, lower Ruvu, northern and western parts of Wami sub-catchments are projected to have decreased maize yields in present, mid and end centuries, due to increased TX and TN.
Figure 7

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during mid century (2040–2069) under RCP 4.5.

Figure 7

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during mid century (2040–2069) under RCP 4.5.

Figure 8 shows the distribution of temperatures, maize yields, rainfall and length of the growing season during the end century. High change in temperatures in the range of 2.3°C to 3.2°C for TX and 2.3°C to 2.7°C for TN are found over upper and lower Ruvu, large part of Mkondoa, Ngerengere, Coast and western parts of Wami sub-catchments. Lower change in temperatures of 1.5°C for TX and 1.9°C for TN are found over Kinyasungwe, southwestern parts of Mkondoa and eastern parts of Wami sub-catchments. Rainfall is projected to increase by 4 to 30 mm over almost the entire basin except eastern parts of Ngerengere, southwestern parts Mkondoa and southeastern parts of Kinyasungwe sub-catchments where it is projected to decrease by 9%. Highest decrease in length of growing season is found over upper Ruvu, western parts of Wami and large parts of Mkondoa sub-catchments. Maize yields are projected to decrease by 1 to 10% over upper and lower Ruvu, eastern and northern parts of Wami, Ngerengere, southeastern parts of Mkondoa and western parts of coast sub-catchments. Meanwhile maize yields over large parts of Mkondoa, Kinyasungwe, southern parts of Wami and eastern parts of coast sub-catchments are projected to increase by 2 to 11%.
Figure 8

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during mid century (2070–2099) under RCP 4.5.

Figure 8

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during mid century (2070–2099) under RCP 4.5.

The distribution of TX, TN, rainfall and length of the growing season during present, mid and end centuries under RCP 8.5 are presented in Figures 911. Temperatures are projected to increase more over central and south eastern parts of the basin. This will reduce maize yields particularly during mid and end centuries. Generally rainfall is projected to increase in many areas of the basin but maize is projected to decrease. This is due to the fact that the entire basin is projected to warm in present, mid and end centuries. This warming will reduce the length of the growing season which in turn will reduce maize yields. This is observed under both RCP 4.5 and RCP 8.5. Thus, regardless of projected increased rainfall that seems to benefit agriculture production, the need arises to undertake more studies that address how to adapt to future increase in TX and TN.
Figure 9

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during present century (2010–2039) under RCP 8.5.

Figure 9

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during present century (2010–2039) under RCP 8.5.

Figure 10

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during mid century (2040–2069) under RCP 8.5.

Figure 10

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during mid century (2040–2069) under RCP 8.5.

Figure 11

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during end century (2070–2099) under RCP 8.5.

Figure 11

Change in spatial distribution of minimum (TN) and maximum (TX) temperatures, rainfall, length of growing season and maize yields over Wami-Ruvu basin during end century (2070–2099) under RCP 8.5.

DISCUSSION

In this paper, high resolution climate change information derived from three RCMs driven by three GCMs and process based crop model CERES maize embedded in DSSAT version 4.5 was used to simulate maize yields over the Wami-Ruvu basin during current, mid and end centuries under RCP 4.5 and RCP 8.5 scenarios. The primary aim was to examine how climate change will affect maize yields in the basin. The analysis is limited to the growing season (December to June). We found that temperatures are projected to increase through the basin. Different RCM-GCM combinations project increase in temperatures differently. However, all the RCM-GCM combinations agree that the basin will experience high change in temperatures during the end century under RCP 8.5. The highest increase in TN of 31.3% (5°C) is projected by RCA4 driven by ICHEC under RCP 8.5 and no change in temperature is projected by the same model during the current century under both RCP 4.5 and RCP 8.5. The ensemble averages also suggest highest change in temperatures within the basin during the end century under RCP 4.5 and RCP 8.5. TN will increase by 23.5% (4°C) under RCP 8.5 and TX will increase by 15.4% (4.4°C) under the same scenario RCP 8.5. These increases in temperature will reduce the length of the growing season and reduce maize yields, particularly in warmer low altitude areas (Moore et al. 2012). In general, results from the present study agree with a prior study by GLOWS–FIU (2014), which indicates that the temperatures within the basin will rise by 4°C in the last quarter of the century.

Our results differ greatly from other previous studies (e.g. Mwandosya et al. 1998; Paavola 2003; Matari et al. 2008; Moore et al. 2012; GLOWS – FIU 2014), which examined climate change within the basin or assessed its impacts on maize yields based on GCM simulations. For instance, a much cited paper by Mwandosya et al. (1998), which was used for development of climate change policies in Tanzania, indicated that maize yields over central Tanzania (Dodoma) region will decline by 80 to 90% towards the end of the century. In the present study, it is found that maize yields over Dodoma will increase by 5 to 8% and 7 to 23% during the end century under RCP 4.5 and RCP 8.5, respectively. Generally Dodoma region, which is located in southwestern part of Kinyasungwe sub-catchment, is predicted to have small change in temperatures, length of growing season and increasing rainfall during present, mid and end centuries under RCP 4.5 and RCP 8.5. Moore et al. (2012) using data from the GCMs indicated that maize yields over Ngerengere, lower and upper Ruvu sub-catchments will increase by 20 to 30% towards the mid century. However, we found that maize yields will decrease in those areas during mid century under both RCP 4.5 and RCP 8.5. Maize yields will decrease by 3 to 8% and 3 to 12% under RCP 4.5 and RCP 8.5 respectively during mid century. Studies that used climate simulations from GCMs to characterize the climate over different regions in Tanzania may have considerable uncertainties since the country has heterogeneous climate over shorter distances which are unlikely to be resolved by coarse resolution GCMs. More recently, GLOWS–FIU (2014) used 16 GCMs to assess the vulnerability of water resources to climate/forest cover change over the Wami-Ruvu basin; the results indicated that the western part of the basin is expected to see higher temperatures as compared with the coast, because of the proximity to the Indian Ocean which regulates temperatures over the coast. The report indicated further that temperatures within the basin will increase from inland to coast. In the present study, we found the opposite, temperatures are expected to increase over the lower altitudes compared to high altitudes, the eastern part of the basin over coast, upper and lower Ruvu, Ngerengere and eastern side of the Wami sub-catchments are projected to have increased temperatures in current, mid and end centuries. Recently Tumbo et al. (2015) analysed the impact of climate change on maize yields over the Wami-Ruvu basin where 20 GCMs were statistically downscaled using the delta method. The downscaled GCM data were used to force CERES to simulate maize yields in the present and future centuries under RCP 4.5 and RCP 8.5. They found that maize yields over the basin will decrease by 5 to 40% due to projected increase in temperature. Statistical downscaling may improve the coarse scale of the GCMs but depends on the choices of the predictor (Wilby & Dawson 2004) and the choice of mathematical transfer function (Wilby & Hayley 2011). The delta method is not strictly a downscaling method (Hewitson et al. 2014). The uncertainties of this method are described in detail by Wilby & Hayley (2011). In the present study, it was found that maize yields in the Wami-Ruvu basin will increase during the current century under RCP 4.5 and RCP 8.5 and decrease during mid and end centuries. The minimum decrease in maize yields of 0.7% and maximum decrease of maize yield of 4% is projected in the mid century under RCP 8.5 and RCP 4.5, respectively. The decrease in maize yields during the mid and end centuries are attributed to projected increase in temperatures that will shorten the growing season. The spatial distribution indicates that climate changes will contribute to increased maize yields over high altitude areas such as over Kinyasungwe, western side of Wami and Mkondoa sub-catchments. However, the low altitude areas such as lower and upper Ruvu, coast, Ngerengere and western parts of Wami sub-catchments will have decreased maize yields due to projected increase in temperatures. Furthermore, western parts of the basin are projected to have increased rainfall, small change in temperatures and length of growing season. Meanwhile the eastern side of the basin is projected to have high increase in temperatures, decrease in rainfall and length of growing season.

Presented results may be used by farmers and decision-makers to plan how to adapt to the projected increase in temperatures particularly over low altitude areas where the negative impact of climate on maize yields are projected. Since the Wami-Ruvu basin has many national and international projects and contains many sources of water for big cities regions like Dar es Salaam and Morogoro, it is crucial to developed adaptation strategies to the projected increase in temperatures and decrease in rainfall particularly in the eastern side of the basin. It is also recommended that more studies should be carried out that address the impact of climate change on crop production in many agro-ecological zones of Tanzania using high resolution climate change projections. It is important to note that actual crop yield data from one season (2009/2010) was the only available data used to validate the crop model in the present study. This may be a limitation, therefore more research is needed uses long period actual yield data to validate crop models and to update the findings of the present study.

CONCLUSIONS

In the present study, the assessment of the impacts of climate change on maize production over the Wami-Ruvu region was carried out using high resolution RCMs. The RCMs used are those driven by boundary condition from GCMs. Daily minimum, maximum temperatures, rainfall and solar radiations for the period of 1971–2000, 2010–2039, 2040–2069 and 2070–2099 were fed into DSSAT to simulate maize growth and yields. In addition to climate data, detailed field and household survey information (crop yield, soil and management data inputs) were used to calibrate the crop model. Maize simulations were carried out under RCP 4.5 and RCP 8.5. We found that the crop model (DSSAT) simulates maize yield over the Wami-Ruvu basin differently when fed with climate data from various RCM-GCM combinations. In general, DSSAT simulates decrease in maize yields over the Wami-Ruvu basin during mid and end centuries. This decrease is small relative to the baseline. Since climate data fed to crop models result in different maize yields production, climate data from ensemble averages of five model members was constructed and used as input into DSSAT to simulate maize growth and yields. Results also showed that maize yield slightly decreased relative to the base line. The maximum decrease of maize yield is projected during the mid century under RCP 8.5 and the minimum decrease is projected over the same time period (mid century) under RCP 4.5. On other hand, there is increase in maize yields in the current century (2010–2039) under RCP 4.5 and RCP 8.5 respectively.

The spatial distribution of seasonal TX, TN, rainfall, length of growing season and maize yields during present, mid and end centuries have shown similar patterns. The central and the eastern parts of the basin are projected to have high increased TX and TN. Importantly, in some places TX and TN are projected to increase by 4.7°C and 4.9°C during the end century under RCP 8.5. The western and northwestern parts are projected to have low increased temperature. Rainfall is projected to increase in most areas during present, mid and end centuries under RCP 4.5 and RCP 8.5. However maize yields are projected to be impacted more by increased temperature than increased rainfall. The increase in temperature shortens the length of the growing season and so contributes to the projected decrease in maize yields. However, in some cases a slight increase in maize yield is projected due to slight increase in rainfall. The assumption made during maize simulation was that all agronomical and management practices were constant. We recommend more research geared towards minimizing the uncertainties and necessary for designing effective and adequate adaptation strategies that will enhance and sustain crop production in the context of the projected climate change over the Wami-Ruvu basin.

ACKNOWLEDGEMENTS

Authors are grateful to Tanzania Meteorological Agency research section for providing computing facilities and CORDEX Africa for providing model data used in this study.

REFERENCES

REFERENCES
Agrawala
S.
Moehder
A.
Hemp
A.
Van Aalst
M.
Hitz
S.
Smith
J.
Meena
H.
Mwakifamba
S.
Hyera
T.
Mwaipopo
O.
2003
Development and Climate Change in Tanzania: Focus on Mount Kilimanjaro
.
OECD
,
Paris
.
Ahmed
S. A.
Diffenbaugh
N. S.
Hertel
T. W.
Lobell
D. B.
Ramankutty
N.
Rios
A. R.
Rowhani
P.
2011
Climate volatility and poverty vulnerability in Tanzania
.
Global Environ. Change
21
,
46
55
.
Arndt
C.
Farmer
W.
Strzepek
K.
Thurlow
J.
2011
Climate Change, Agriculture, and Food Security in Tanzania
.
United Nations University, UNU-WIDER
,
Katajanokanlaituri, Helsinki, Finland
.
Daniels
A. E.
Morrison
J. F.
Joyce
L. A.
Crookston
N. L.
Chen
S. C.
McNulty
S. G.
2012
Climate Projections FAQ. Gen. Tech. Rep. RMRS-GTR-277WWW
.
Fort Collins, CO, USA. Department of Agriculture, Forest Service, Rocky Mountain Research Station
.
Ehrhart
C.
Twena
M.
2006
Climate Change and Poverty in Tanzania: Realities and Response Options for CARE. Background Report
.
CARE International Poverty-Climate Change Initiative
.
GLOWS–FIU
2014
Climate, Forest Cover and Water Resources Vulnerability
.
Wami/Ruvu Basin
,
Tanzania
.
Graves
A. R.
Hess
T.
Matthews
R. B.
Stephens
W.
Middleton
T.
2002
Crop simulation models as tools in computer laboratory and classroom-based education
.
J. Natural Res. Life Sci. Educ.
31
,
48
58
.
Hampton
J. G.
Boelt
B.
Rolston
M. P.
Chastain
T. G.
2012
Effects of elevated CO2 and temperature on seed quality
.
J. Agric. Sci.
150
,
1
9
.
Hartkamp
A. D.
De Beurs
K.
Stein
A.
White
J. W.
1999
Interpolation Techniques for Climate Variables. NRG-GIS Series 99-01
.
CIMMYT
,
Mexico, D.F
.
Hewitson
B.
Daron
C. J.
Crane
R. G.
Zermoglio
M. F.
Jack
C.
2014
Interrogating empirical-statistical downscaling
.
Clim. Change
122
,
539
554
.
IPCC
2007
Working Group II Fourth Assessment Report
.
Climate Change: Climate Change Impacts, Adaptation and Vulnerability
.
IPCC
2012
Managing the risks of extreme events and disasters to advance climate change adaptation – summary for policy makers
. In:
A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change
(
Field
C. B.
Barros
V.
Stocker
T. F.
Qin
D.
Dokken
D. J.
Ebi
K. L.
Mastrandrea
M. D.
Mach
K. J.
Plattner
G.-K.
Allen
S. K.
Tignor
M.
Midgley
P. M.
, eds).
World Meteorological Organization
,
Geneva
,
Switzerland
.
IPCC
2013
Summary for policy makers
. 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
(
Stocker
T. F.
Qin
D.
Plattner
G.-K.
Tignor
M.
Allen
S. K.
Boschung
J.
Nauels
A.
Xia
Y.
Bex
V.
Midgley
P. M.
, eds).
Cambridge University Press
,
Cambridge
,
United Kingdom
and New York, NY, USA
.
Kashaigili
J.
2011
Rapid Environmental Flow Assessment for the Ruvu River, A Consultancy Report submitted to iWASH
.
Matari
E. E.
Chang'a
L. B.
Chikojo
G. E.
Hyera
T.
2008
Climate change scenarios development for second national communication, Tanzania
.
TMA Res. J.
1
,
8
18
.
Mbwanga
M.
2005
On Farm Verification and Promotion of Green Manure for Enhancing Upland Rice Productivity on Striga Infested Fields in Tanzania, R No. 8194 (ZA No0511)
.
Miglietta
F.
Bindi
M.
1993
Crop growth simulation models for research, farm management and agro-meteorology
.
EARSel Adv. Remote Sens.
2
,
148
157
.
Milder
J. C.
Buck
E. L.
Hart
K. A.
Scherr
J. S.
2012
A Green Growth Investment Framework for SAGCOT, the SAGCOT Greenprint
.
Moore
N.
Alagarswamy
G.
Pijanowski
B.
Thornton
P.
Lofgren
B.
Olson
J.
Andresen
J.
Yanda
P.
Qi
J.
2012
East African food security as influenced by future climate change and land use change at local to regional scales
.
Clim. Change
110
,
823
844
.
Müller
C.
Cramer
W.
Hare
W. L.
Lotze-Campen
H.
2011
Climate change risks for African agriculture
.
Proc. Natl. Acad. Sci. USA
108
,
4313
4315
.
Mwandosya
M. J.
Nyenzi
B. S.
Luhanga
M. L.
1998
The Assessment of Vulnerability and Adaptation to Climate Change Impacts in Tanzania
.
Centre for Energy, Environment, Science and Technology (CEEST)
,
Dar es Salaam
,
Tanzania
.
CEEST Book Series No. 11
.
Nikulin
G.
Jones
C.
Giorgi
F.
Asrar
G.
Büchner
M.
Cerezo-Mota
R.
Christensen
O. B.
Déqué
M.
Fernandez
J.
Hänsler
A.
van Meijgaard
E.
Samuelsson
P.
Sylla
M. B.
Sushama
L.
2012
Precipitation climatology in an ensemble of CORDEX-Africa regional climate simulations
.
J. Clim.
3
,
6057
6078
.
Paavola
J.
2003
Vulnerability to climate change in Tanzania: Sources, substance and solutions, a paper presented at the Inaugural Workshop of Southern Africa Vulnerability Initiative (SAVI) in Maputo
,
Mozambique
,
June 19–21, 2003
.
Pechlivanidis
I.
2013
India-HYPE has been set-up for the Indian region to evaluate CORDEX-SA projections of future changes in water storage and fluxes. Available online at: http://www.smhi.se/en/research/research-departments/hydrology/in-hype-1.28022.
Philip
G. M.
Watson
D. F.
1986
Matheronian geostatistics – Quo vadis?
Math. Geol.
18
,
93
117
.
Rowhani
P.
Lobell
D. B.
Linderman
M.
Ramankutty
N.
2011
Climate variability and crop production in Tanzania
.
Agr. Forest Meteorol.
151
,
449
460
.
Saxton
K. E.
Rawls
W. J.
2009
Soil water properties calculator (version 6.02). Available online at: http://hydrolab.arsusda.gov/soilwater/Index.htm.
The National Bureau of Statistics
2012
National Panel Survey 2009-2010. Available from http://www.nbs.go.tz
.
Thornton
P. K.
Jones
P. G.
Owiyo
T.
Kruska
R. L.
Herrero
M.
Orindi
V.
Bhadwal
S.
Kristjanson
P.
Notenbaert
A.
Bekele
N.
Omolo
A.
2009
Climate change and poverty in Africa: mapping hotspots of vulnerability
.
Afr. J. Agr. Resour. Econ.
2
,
24
44
.
Thornton
P. K.
Jones
P. G.
Alagarswamy
G.
Andresen
J.
2010
Spatial variation of crop yield response to climate change in East Africa
.
Glob. Environ. Change
19
,
54
65
.
Tumbo
D. S.
Mzirai
O.
Mourice
S.
Msongaleli
B.
Wambura
F.
Kadigi
I.
Sanga
C.
Kahimba
F.
Ngongolo
H.
Sangalugembe
C.
Mutabazi
K.
Sumari
N.
2015
Impacts of Climate Change on Agriculture: What, When, Where and How? Technical Report
.
URT
2012
Global agriculture and food security programme request for funding – public sector window. Tanzania Agriculture and Food Security Investment Plan, GAFSP Gap Financing Proposal, available online at https://www.gafspfund.org/sites/gafspfund.org/files/Documents/TANZANIA_GAFSP_PROPOSAL_FINAL.pdf.
URT
2013
Status of Gauging Station Wami/Ruvu Basin. DPG Tan, Annual Basin Hydrological Report
.
Washington
R.
Pearce
H.
2012
Climate change in East African agriculture: recent trends, current projections, crop-climate suitability, and prospects for improved climate model information
. In:
CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS)
,
Copenhagen, Denmark
.
Available online at: www.ccafs.cgiar.org.
Watson
D. F.
Philip
G. M.
1985
A refinement of inverse distance weighted interpolation
.
Geo-Processing
2
,
315
327
.
Wilby
L. R.
Dawson
C. W.
2004
Using SDSM Version 3.1 — A decision support tool for the assessment of regional climate change impacts
.
Wilby
L. R.
Hayley
F. J.
2011