Modelling Impacts of Climate Change on Maize (Zea mays L.) Growth and Productivity: A Review of Models, Outputs and Limitations

The use of crop modelling in various cropping systems and environments to project and upscale agronomic decision-making under the facets of climate change has gained currency in recent years. This paper provides an evaluation of crop models that have been used by researchers to simulate maize growth and productivity. Through a systematic review approach, a comprehensive assessment of 186 published articles was carried out to establish the models and parameterization features, simulated impacts on maize yields and adaptation strategies in the last three decades. Of the 23 models identified, CERES-maize and APSIM models were the most dominant, representing 49.7% of the studies undertaken between 1990 and 2018. Current research shows projected decline in maize yields of between 8% 38% under RCP4.5 and RCP8.5 scenarios by the end of the 21 century, and that adaptation is essential in alleviating the impacts of climate change. Major agro-adaptation options considered in most papers are changes in planting dates, cultivars and crop water management practices. The use of multiple crop models and multi-model ensembles from general circulation models (GCMs) is recommended. As interest in crop modelling grows, future work should focus more on suitability of agricultural lands for maize production under climate scenarios.

IPCC (2018), the projected increase in global warming to 1.5˚C is likely to reduce productivity of key cereals including maize, rice and wheat in sub-Saharan Africa, Southeast Asia, and Central and South America.
In order to ensure today's food security and in the coming decades, efforts have been made on establishment of crop simulation models aimed at predicting growth, development and yield potential of a crop under certain environmental conditions (Basso et al., 2013;Wang et al., 2018;. Several dynamic crop simulations models (CSMs) have been developed and used widely to study physiological, physical and chemical processes of crop productivity under a changing climate (Kasampalis et al., 2018;Shi, Tao, & Zhang, 2013;White, Hoogenboom, Kimball, & Wall, 2011). The crop models and their outputs are then used to guide agronomic decision-making aimed at sustainable management and development of adaptive strategies for responding to impacts of climate change (Basso et al., 2013). Optimum management practices that are either strategic or tactful including planting dates, selection of crop variety, fertilizer usage and water application can be analysed through proven models for planning purposes (Boote, Jones, & Pickering, 1996). Furthermore, crop simulation models can play a major role in evaluating the potential impacts of climate change on agricultural systems in the world (IPCC, 2007;Stocker et al., 2013). Likewise, CSMs have played a key role in interpretation of agronomic results and are increasingly being used by farmers and policy-makers for decision-making in crop production. The differences in spatio-temporal scales and predicted changes in global climate and land use have led to development of several crop models by researchers for agronomic purposes. In addition, several general circulation models (GCMs) have also been developed for use in crop modelling and other related aspects (IPCC, 2007). Thus, the crop-modelling subject has attracted many researchers and policy makers and was one of the major issues discussed during the COP21 agreement in Paris in 2015.
Scientific research on climate and crop modelling in the last three decades has increased tremendously and it is of critical importance to have an overview of existing models and their use, including their outputs, to inform future investigations. Such an assessment can help answer weighty questions such as: 1) What study, we review papers published in journals for the past 28 years (from 1990 to 2018) with the main objective of exploring the models that are specifically used to simulate the impact of climate change on maize growth and productivity. This review forms a beneficial reference by addressing the limitations that exist on identification of the available models, the parameterization requirements, and the regions where such models have been applied. In the paper, we consider maize (Zea mays L.) since it is a major crop grown in most parts of the world to provide food and nutritional security for the vulnerable populations (Tesfaye et al., 2015). Like other crops, the main challenge that affects maize growth and development is the changing weather pattern, leading to intra-seasonal changes in yield . Other factors are variable soil properties, crop agronomic management practices including planting, fertilizer application, tillage among others (Ahmed et al., 2018;Lin et al., 2017;Ramirez-Cabral, Kumar, & Shabani, 2017;Tesfaye et al., 2015;Tesfaye et al., 2018). We believe that analysing the performance of various maize simulation models is vital for addressing the challenges posed by climate change.
The specific objectives of the review include: 1) identification of the various models and geographic locations used, including the input parameters and simulated processes; 2) examination of the projected simulated yields from various models and the adaptation measures applied; and 3) identification of limitations and future perspectives in crop modelling. This paper is intended to provide a basis for researchers as well as agronomic decision-makers who are interested in understanding the available maize simulation models and devising suitable perspectives in crop modelling. This paper is organized as follows: Section 2 describes the methodology used in the identification of peer-reviewed articles of interest. The results are presented in Section 3 focusing on the models available and their features, summary of articles published in the review period and overview on model objectives. In Section 4, a detailed discussion of the models and an overview of simulated impacts of climate change on maize are presented. Lastly, Section 5 is on conclusion and recommendations. Figure 1. We used ISI Web of Knowledge (IWK) as the main source for identification of relevant peer reviewed articles. IWK was selected because it is one of the oldest citation databases with wide coverage of high impact journals that are mostly written in English (Aghaei Chadegani et al., 2013;Levine-Clark & Gil, 2008). The first inclusion criteria involved searching of articles with the title maize and topics that included climate change, model, yield and simulation. The second inclusion criteria limited results to journal articles that were specific to maize and climate change on their titles. Lastly, the list of papers was filtered to exclude articles whose objectives were not focused on maize production, the non-peer reviewed journal articles, those written in languages other than English and those not available through internet sources. With an addition of five more papers that were identified through cross-referencing, the ultimate sources of literature for this study were 186 papers dating from 1990 to 2018. An excel spreadsheet was developed to enter specific information on location of the study areas, model used, objective of the study, input parameters and outputs. Schematic representation of the systematic review process is presented in Figure 1.
Following the screening process, the other steps were extraction of information from the articles according to the following thematic areas: • The simulation model used in the study • Focus of the model, input data and processes simulated • General circulation model (GCM) applied • Geographic location of the study • Objectives of the research, simulated outputs and adaptation strategies

Maize Simulation Models
Based on the criteria we used in our assessment, we identified 23 models that have been developed to simulate the impacts of climate change on maize production ( Table 1). The dominant models identified were CERES Maize (59 articles) and APSIM (30 articles).

Focus of the Model, Input Data and Processes Simulated
In terms of the selected features, the 21 models mainly focused on simulation of maize growth and yields, with the exception of CLIMEX and MaxEnt models whose focus was on predicting geographic distribution of maize under climate change scenarios. The GCMs that dominated in most articles were ECHAM5, CCSM, HadCM3, CSIRO-MK3, CGCM3.1, UKLO and MIROC3.2. The input parameters that were established to be common in most models were: 1) weather data including temperature, daily rainfall and solar radiation; 2) soil data including soil type, soil depth, soil texture, soil organic carbon, bulk density, soil nitrogen; 3) crop information and management of crop species, planting date, row space, plant density; 4) field management practices such as water management, irrigation usage including scheduling, method and amount applied, fertilizer   usage and type, pesticide application and tillage practices. The physiological process simulated by most models are phenology, biomass and evapotranspiration. Specific input parameters for the models and processes simulated are presented in Table 2.

Objectives of the Papers
The objectives presented in different papers ranged from analysing the maize yields under the impacts of climate change, to adaptation strategies and model suitability (Table 3). From the findings, most (46.6%) of the papers investigated the impacts of climate change on maize production in terms of yields.

Model Evaluation
From the assessment, a total of 23 different simulation models were identified from the articles (Table 1). Of these, CERES-maize and APSIM models were the   there is inadequate information that can be passed to stakeholders and farmers due to minimal research, lack of access and uncertainties associated with data used (Whitfield, 2013). Therefore, it is imperative that governments in developing countries or regions with limited data prioritise collection of weather data and support more research on modelling in order to plan strategies for adapting to the impacts of climate change on crop productivity.
Of the 186 articles, only 7.5% simulated maize growth with more than one crop model. Those that simulated results with more than one GCM were approximately 17.2%. This then points to a possibility of uncertainties in results obtained from several studies using single crop models and GCMs. In line with this, Zhang, Zhao, and Feng (2018) in their study, observed that GCMs contributed more uncertainties to maize-yield simulations compared to crop simulation models that use observed environmental data. Therefore, the use of more

Impacts of Climate Change on Maize Productivity
Most reviewed articles (81%) assessed the impacts of climate change on maize productivity. Others were on yield, biomass and leaf area index (14%), biomass (5%) and area suitability for maize production (1% and 2100 leading to reduction of total biomass by approximately 10% and LAI by 16%. In West Africa, Parkes, Sultan, and Ciais (2018), in their study using General Large Area Model (GLAM), projected a reduction of maize yields by 5.95% with an increase of temperature. A study by Araya et al. (2015) in Ethiopia using APSIM and CERES maize models under 20 GCMs and RCP 4.5 and 8.5 reported an increase in maize yields of between 1.7% and 4.2%. Araya et al. (2017) (Ruane et al., 2013).
In terms of area suitability for maize production, Ramirez-Cabral et al. (2017) in their study using CLIMEX distribution model with climate data CSIRO-Mk3.0 and MIROC-H GCMs predicted high loss of climate suitability for maize production between the tropics of Cancer and Capricorn (highest being in South America, followed by Africa and Oceania); whereas poleward regions (including Asia, Europe and North America) exhibit increase in suitability.
In analysing the outputs of the simulated results from the models, one of the challenges in this review was to undertake an inter-comparison of the estimated yields, which in most cases differed in various studies. This is supported by the works of Bassu et al. (2014) who pointed out that different models produce different projected impacts of climate change due to variability in parameterization processes. In addition, model simulations vary due to differences in structure, processes considered and their relative importance depending on the region where the model was developed (Challinor, Ewert, Arnold, Simelton, & Fraser, 2009 (Hansen & Jones, 2000;Reidsma, Ewert, Boogaard, & van Diepen, 2009).
In general, most studies and models reviewed reported mainly on maize production and yields compared to those that analysed variation in geographic suitability and distribution of the crop as influenced by climate change.

Adaptation Strategies and Their Implications
While estimates using different maize simulation models project decline in yields, most studies emphasized that future maize production can benefit from various adaptation strategies aimed at offsetting the negative impacts of climate change on maize production. Some of the dominant adaptation measures considered in 44 studies shown in Table 3 (25 on adaptation and 19 on impact and adaptation) were change of sowing date, cultivars and crop water management, respectively represented by 47%, 31% and 3% of the studies under this review.
The major adaptation strategies suggested for consideration on crop models include breeding new cultivars, proper irrigation and soil nutrient management (Bannayan, Paymard, & Ashraf, 2016;Lin et al., 2017;Moradi, Koocheki, Mahallati, & Mansoori, 2013;Rurinda et al., 2015;. The study by Lana et al. (2016) is a notable example that reported how change of planting date impacted on maize yields. In their works, by using CERES-Maize model without factoring in adaptation strategies, showed reduction of 11.5% -13.5% in total maize production across the cultivars used. They also reported that by combining cultivar and the best planting date, they simulated an increase in production by 15%. Similarly, a study in Northern China by Lin et al. (2017) showed that maize yields would decrease by 6.9% and increase by 15.9% if planting days were advanced or delayed by 15 days respectively. Another notable example is by Parent et al. (2018) who, in their study using APSIM model and six field experiments in South and Northern Europe projected an increase of 4% -7% in grain production through adaptation that involves genetic variability of flowering during the crop cycle. Rahimi-Moghaddam, Kambouzia, and Deihimfard (2018) found that, by combining early sowing and using cultivars which have high thermal time requirements in North Eastern, there is high likelihood of reducing the impact of climate change on maize productivity. Lashkari, Alizadeh, Rezaei, and Bannayan (2012), Bannayan et al. (2016) and (Araya et al., 2015;Araya et al., 2017;Reddy et al., 2016) in their studies also reported positive effects of early sowing dates as an adaptive strategy towards impact of climate change. D. P.  in their study in North China Plain using APSIM model concluded that cultivation of maize cultivars with longer growing periods and higher thermal requirements could be a potential adaptation measure towards mitigating the impacts of climate change on crop production.
Adequate polices on adapting agriculture to climate change have also been proposed. For example, Kang et al. (2009)

Limitations of Crop Modelling
The limitations inherent in crop models include the input parameters, calibration, evaluation and validation procedures and methods of simulating crop responses to various environmental and management factors, leading to uncertainties in prediction of crop yields and identification of appropriate measures towards adaptation (Bassu et al., 2014;Moradi et al., 2013). Likewise, as the intended objectives of various models differ, the structure of the model including the input parameters may result in differences in projected climatic impacts, which in most cases are based on estimations. The performance of models is also constrained by the accuracy and precision of the input data, which can be affected by poor calibration of the sensors used in the research study before model applications (Boote et al., 1996). For example, most researchers use "above-ground" crop data as compared to data related to root growth and development which are not extensive and with enormous sampling errors. In addition, very few studies accounted for important factors which play an important role in crop development, such as weeds, diseases, insects, cultivation and phosphorous. According to Basso et al. (2013), one of the challenges in crop modelling is the use of observed and simulated results to determine the cause of spatial and temporal crop variability and how to manage the crop from agronomic, environmental and economic perspectives. Other limitations associated with crop models are related to sensitivity to CO 2 , which has major influence on projected changes and remains an obstacle to the assessment of the impacts of climate change (Ruane et al., 2013).
The use of GCMs provides reasonable accuracy on wide-scale assessments.
However, their use for the prediction of climate scenarios is not without limitations (Kang et al., 2009;Whitfield, 2013). These include systematic errors such as the tendency of northward displacement during winter in the northern hemisphere, too wet simulations in the middle latitudes of both hemispheres and underestimation of clouds in the tropics (D'andrea & Vautard, 2000). Therefore, the use of GCMs with higher spatial resolutions is recommended in order to acquire realistic projections on the impacts of climate change on crop production at a regional scale (Reddy & Hodges, 2000).
Though Basso et al. (2013) argued that remote sensing may not be suitable for use in mixed agricultural lands and small farm sizes, especially in developing countries where available satellite data is not of good quality, the future of crop modelling is more expected to integrate remote sensing.

Conclusion and Future Work
This paper presents the crops models that are used to simulate maize growth and development. Through a systematic review of literature, we identified 23 simulation models from ISI Web of Knowledge that have been used to research and simulate maize productivity under the impact of climate variability. For the last two decades, these models have played a significant role in research, farm-level management and agronomic decision-making. To emphasize our findings, the following points briefly summarize our review and the future perspectives drawn: • Most articles have focused on projecting impact of climate change on maize production compared to adaptation and suitability of the geographic area for maize production. It is clear that with the anticipated effects of climate change, adaptation of maize production systems is essential. • The process-based models that are used in maize simulation vary in their complexity but share some inter-comparable input parameters and plant processes that include phenology, canopy and biomass establishment. Comparability of the input parameters in the models is relevant for climate change studies where results can inform decision-making and policy direction.
• With the broadening of crop simulation models, concerns include the reliability of outcomes from a single simulation model that may not adequately factor in all the pertinent processes. To address this challenge, the use of multiple crop models and GCMs to simulate crop growth remains a major consideration in future crop modelling in order to minimize uncertainties in simulated results that can be linked with individual predictions.
• Integration of remote sensing and crop models presents feasible agronomic consideration for monitoring of crop growth and yield forecasting. Though there could be some research gap regarding the criteria used in this systematic review and in analysing the large amount of diverse peer-reviewed articles, the results of this study provide not only important insights on the diversified maize simulation models and projected outputs, but also provide better understanding of the projected impacts, adaptations and future works towards sustainable maize production under the adverse impacts of climate change.