The impact of climate change on croplands
Phenology and yield
The growth duration of a crop determine the time length of light interception and photosynthesis, therefore the crop yield and its water and energy balance. As sensitivity of crop to climate variations are not equally sensitive across the growing season (Porter & Semenov, 2005), which can be particularly sensitive to climate variations at certain reproductive phase (e.g. Hatfield et al., 2011; Espe et al., 2017), the timing of key phenological events are also of great importance. Thus, understanding how phenology respond to climate change is a prerequisite to understand how climate change affects crop ecosystems. It has long been recognized that climate change has significant impacts on crop growth duration. Each 1°C of warmer temperature shorten growth duration by ~7 days by average (Muchow et al., 1990; IPCC, 2007). However, this average sensitivity cannot really represent the theories and observations. A widely-adopted theory of cardinal temperatures suggest that, crop growth accelerates with warmer temperature when it is below the optimum temperature for crop development (e.g. Hatfield et al., 2011), which shorten the growth duration. However, when temperature is above its optimum, the acceleration with higher temperature may disappear. This critical temperature threshold (the optimum temperature) may differ largely across crops and varieties ranging from 20°C to 35°C (Sanchez et al., 2014). It should be noted that how crop-climate relationship may change above the optimum temperature is largely uncertain and differ across crops (Craufurd & Wheeler, 2009). For example, some studies found rapid senescence of wheat after exposure to 32-34oC during flowering period (Asseng et al., 2011; Lobell et al., 2013). However, for rice, the limited number of researches indicate that growth duration is not responsive to temperature when it goes beyond the optimum (Yoshida, 1983). These observational evidences, however, have not been well accounted in many widely used crop models (Sanchez et al., 2014). For example, the CERES model used for assessing food security under climate change in China (e.g. Xiong et al., 2009; Xiong et al., 2010) only considers the acceleration effects of warming but not the high temperature stress (e.g. Lobell et al., 2013)。
For natural ecosystems, many studies have consistently shown that global warming over the past few decades has advanced the spring onset date (e.g. Menzel et al., 2006; Wang et al., 2015), lengthening the growing season (e.g. Garonna et al., 2014), though it may reverse over a few regions due to climate flucturations (e.g. Piao et al., 2011). However, unlike the consistency found for natural ecosystem, the trend in crop growth duration was quite diversified in different researched. For example, Siebert et al. (2012) found growth duration of oat over Germany is shortening by 0.1-0.4 day/10a over past five decades; Tao et al. (2006) found growth duration for rice over China have also shortened over past two decades. However, more recent researched over past two to three decades found growth duration for major cereal crops (rice, wheat and maize) over China has become longer (e.g. Liu et al., 2012; Liu et al., 2013; Tao et al., 2013; Xiao et al., 2013; Li et al., 2014). These results appear contradictory, but can be reconciled with adaptation measures by selecting long-duration varieties.
Crop yield can be affected by temperature change through different pathways. First, rising temperature directly drives change in photosynthetic rate (Figure 1.1). When temperature is below the optimum temperature, rising temperature will enhance photosynthetic rate, while it suppress photosynthetic rate when temperature goes beyond the optimum. Respiration processes also subject to temperature regulations. However, the optimum temperature for respiration is usually higher than that of photosynthesis and outside measurement range (Figure 1.1). It is therefore commonly believed that higher temperature will lead to higher respiration rate. The net effect of temperature on photosynthesis and respiration is the temperature effects on crop productivity. Night-time warming was believed to negatively affect crop yield as respiration increase while photosynthesis is still zero (Peng et al., 2004; Lobell et al., 2012a). However, due to potential compensation effects that enhance photosynthesis on the day (Wan et al., 2009), warmer nighttime temperature may also improve crop productivity. Second, certain phase of crop reproductive growth (e.g. silking and grain filling) is sensitivity to high/low temperature stress (e.g. Schar et al., 2004; Espe et al., 2017). For example, high temperature stress can lead to failure of flowering, grain formation and grain filling, leading to reduced crop yield (Schar et al., 2004; Porter & Semenov, 2005; Asseng et al., 2011; Teixeira et al., 2013). Third, as mentioned in previous paragraph, temperature change will affect the length of growing duration, which affect the accumulation of photosynthesis and thus yield. Usually, higher temperature lead to shorter growing duration and lower yield (e.g. Iqbal et al., 2009; Giannakopoulos et al., 2009; Lobell et al., 2012b). Finally, increase in temperature lead to exponential increment of vapor pressure deficit, which may also stress the productivity of croplands (e.g. Lobell et al., 2013).
The impact of precipitation change on crop yield remains more controversial. Some studies show that 20% decrease in precipitation will still have limited impacts on maize yield over USA (Lobell et al., 2013), while other studies found precipitation change as more dominant factor than change in temperature and atmospheric CO2 on crop yield (Ko et al., 2010). Probably due to expansion of irrigation, which may alleviate the water stress to crop production, the studies on impact of precipitation on crop yield is much less than that of temperature. However, climate change will lead to change in irrigation demands (Elliot et al., 2014) and spatio-temporal distribution of available water resources. Whether sufficient irrigation water can be provided is a urgent research question to answer. In addition, projected increase in exteme events, such as droughts and flood (IPCC, 2012), may also leads to fluctuations of global crop productions (Lesk et al., 2016).
Solar radiation reaching the land surface is the energy source of photosynthesis and thus crop productivity. Interannual variations of solar radiation has significant impacts on rice yield over China (Zhang et al., 2010). However, it is so commonly assumed that crop growth was more stressed by temperature and water availability (Hatfield et al., 2011), the impact of variations of solar radition on crop yield remains largely uncertain.
Despite growing knowledge on the mechanism how climate change could influence crop yield, our knowledge on the key parameters (e.g. cardinal temperature) and dominant climatic factors driving yield change remains unclear. Regional and inter-crop differences may further complex situation. Large uncertainties, therefore, still exist in quantifying climate change impacts on crop production (IPCC, 2013a). A synthesis of 66 studies on climate change impacts on crop yield (IPCC, 2013a) found that warming of 1-2 oC may lead to decline of wheat and maize yield. However, rice in tropical region and maize in temperate regions show different response to warming in different studies. As a result, even qualitative conclusions are difficult to make. Different global studies drew different conclusions on how rice yield respond to climate change. For example, Lobell et al. (2011) found climate change over past three decades may slightly enhance the yield, while recent multi-model intercomparison study (Rosenzweig et al., 2014) found climate change will reduce rice yield, without considering the CO2 fertilization effect. Therefore, detailed regional studies are warranted in order to reduce the uncertainties. However, regional studies based on statistics, long-term agro-meteorological site observations and crop models drew contrast conclusions on how climate change affects rice yield over China (Lin et al., 2005; Yao et al., 2007; Tao et al., 2008; Xiong et al., 2007; Xiong et al., 2009; Zhang et al., 2010; Welch et al., 2010; Tao et al., 2012), highlighting large uncertainties in the estimates. Single model studies are prevalent among previous ones (e.g. Lin et al., 2005; Xiong et al., 2007; Yao et al., 2007; Xiong et al., 2009; Tao and Zhang, 2012) , but the uncertainties related to model structures and parameters remains largely unexplored. Recent studies seems indicating the multi-model ensemble may improve confidence in projecting how crop yield may respond to the changing climate (Asseng et al., 2015; Martre et al., 2015; Li et al., 2015).
Land surface energy and water exchange
Irrigation accounts for ~70% of global water widraw (Shiklomanov & Rodda, 2003), which is also a key variable for projecting crop production and food security (IPCC, 2013a). The irrigation water requirements of croplands are determined by balance of precipitation and evapotranspiration, both of which are affected by climate change. Anthropogenic climate change is projected to alter the spatial distribution of annual precipitation (IPCC, 2013b), which will change the water availability over contemporary cropping area. The seasonal distribution of precipitation may also altered (IPCC, 2013b), which may induced seasonal shortage of water supply during growing season.
Climate change affect evapotranspiration through three pathways. First, it affects crop productivity, which consume water affect the rate of evapotranspiration; Second, it regulates length of growing season, which affects the annual sum of evapotranspiration; Finally, warmer temperature will directly change saturated water vapor pressure and stomatal conductance, the net effect of which may accelerate the crop evapotranspiration (e.g. Ben-Asher et al., 2008). One factor often dismissed in studies on crop evapotranspiration is the impact of solar radiation (Hatfield et al., 2011), which directly alter the energy balance of the land surface (Wild et al,. 2005). The commonly used empirical equation (Penmman-monteith) in crop models does not include effects of solar radiation, which may underestimate variations of evapotranspiration. Rising atmospheric CO2 will lead to decrease of stomatal conductance and thus reducing transpiration (Leaky et al., 2006). Across different FACE experiments, stomatal conductance by average reduce by 20% in response to enhanced CO2 at 550 ppm (Ainsworth et al., 2005). The reduction of stomatal conductance may further enhanced to 30%-40% under doubling CO2 concentration (Hatfield et al., 2011). However, at canopy level, the observed change of evapotranspiration under double CO2 is only 8%-13% (Hatfield et al., 2011), which can result from negative feedbacks result from higher CO2 induced higher leaf temperature and photosynthetic rate (Leaky et al., 2009; Burkart et al., 2011). Rising atmospheric CO2 and temperature drive change evapotranspiration in different direction, which is a hotspot for impact studies and remains largely uncertain (Liu & Tao, 2013). Complex interactions among climate change factors in affecting evapotranspiration may have not been fully understood and incorporated in the models. For example, rising CO2 may enhance vegetation growth, and thus surface roughness, resulting in reduced wind speed (Vautard et al., 2010). The lower wind speed resulted from rising CO2 may thus reduce evapotranspiration.
Overall, experimental and model studies show that warmer temperature lead to increasing cropland evapotranspiration (e.g. Guo et al., 2010; Hoff et al., 2010; Gerten et al., 2011). Field observational studies in general agree that rising atmospheric CO2 will lead to decrease of cropland evapotranspiration (e.g. Reddy et al., 1995; Leaky et al., 2006; Bernacchi et al., 2007). Assuming no change of crop varieties, the global modelling study show the overall effect of climate change following RCP8.5 will be reducing global crop irrigation demand by 8%-15% (Elliot et al., 2014), but the sign and magnitude change across crops and regions. Uncertainties are still large, as hydrological models and crop models differ, by average two times, in the estimate of crop irrigation demand (Elliot et al., 2014).
Crop models, from sites to the globe
Crop models are the essential tool integrating our knowledge of climate change impacts on croplands. The field-scale crop model started from 1960s with two genres: The waegningen group led by de Wit (1965) developed crop growth model based on light use efficiency module. Crop models such as WOFOST and ORYZA(Bouman & Van Laar, 2006) are evolution of this type of models. The other genre is the CERES type of model (Ritchie et al., 1985) based on earlier work by Duncan (1967), including CROPGRO. DSSAT is the platform integrating both CERES and CROPGRO. APSIM is an Australian model also belongs to this genre. Despite the differences among these models, there are some resemblance on them, such as the use of radiation use efficiency (RUE) module or water use efficiency (WUE) module, the thermal accumulation module to drive crop phenology, the use of variants of Penman-Monteith equation for calculation of evapotranspiration. These traditional crop model have strong suits in detailed simulation of organ developments, given a large number of parameters. However, the equations used are often highly empirical. For example, the water and nutrient stress to crop phenology development, the ratio of actual to potential evapotranspiration are often empirical parameter between 0 and 1. Such formulation of equations will easily lead to over-parameterization and uniformality issues in representing physiological process. The photosynthesis in these models are semi-empirical WUE or RUE model (Soussana et al., 2010), instead of the physiology based Farquhar(Farquhar & Sharkey, 1982). Under contemporary climate, these crop models may be parameterized to reflect the characteristics of the croplands, but its robustness to be extrapolated into future and project impacts of climate change could be dubious (Nowak et al., 2004; Soussana et al., 2010). For example, Wang et al. (2012) show WUE and RUE model may predict contrast response of productivity to climate change over China. There are a long list of this type of crop models developed by researches from different countries (e.g. STICS (Brisson et al., 2008), SIMRIW(Horie, 1987; Zhang et al., 2014), Agro-C (Huang et al., 2009), RiceGrow (Tang et al., 2009), McWLA(Tao and Zhang, 2012)), which have been developed and tuned for a certain crop-region. As a result , in recent model intercomparison of crop models for wheat, maize and rice, no models can out-perform others in four test sites at different regions of the globe (Li et al., 2015; Martre et al., 2015).
Researchers have realized the difficult in applying the site-scale model at regional and global scales (Challinor et al., 2009), at which climate change impacts and economy models have to operate. The other generation of crop models was thus developed to explore large scale crop-climate relationships, such as IMAGE (Leeman & Solomon, 1993). These model typically divide the globe into several agro-ecological zones. Empirical relationship between climate and yield was then built usually with agro-statistics. Some selective process may also be incorporated into these models for model improvements, such as GLAM (Challinor et al., 2004). Compared with traditional crop models mentioned in previous paragraph, these models have far less input requirements and parameters and low requirement of computing resources, which facilitates large-scale applications. However, its empirical nature may hurdle further exploration on how management practices may affect the croplands’ response to climate change (Challinor et al., 2004). When climate change beyond its contemporary range of variations (Mora et al., 2013), it is hard to prove whether the contemporary empirical relationship may still apply. Similar issues also apply for different types of statistical models (e.g. Lobell et al., 2011).
Compared with previously mentioned models, terrestrial ecosystem models have more physiology-based formulations. However, previous studies often neglect or simplified representation of crop ecosystems (e.g.Piao et al., 2009). The simplified module cannot represent the generally short growth duration of crops (Smith et al., 2010) and different allocation strategy of croplands than natural ecosystems (Bondeau et al., 2007). All earth system models in CMIP5 did not include a specific crop module. As croplands role in global biogeochemical cycle being gradually brought more attentions, there are some efforts introducing crop modules into the ecosystem models (Drewniak et al., 2013). For example, Kucharik (2003) bring crop phenology, irrigation and fertilization module into IBIS model, resulting in better representation of spatio-temporal variations of maize yield over US due to climate and management differences (Kucharik, 2003; Kucharik, 2008). Levis et al. (2012) bring Agro-IBIS into community land model, finding improved representation of dynamics in leaf area index (LAI), net ecosystem exchange and thus seasonal variations of atmospheric CO2 concentration. Bondeau et al. (2007) introduce crop functional type to LPJ. The improved LPJmL model, though only introduce improvements of phenology at that time, simulated 24% less global vegetation carbon pool than original LPJ model and produce significant difference in spatio-temporal variations of net primary productivity. Similarly, ORCHIDEE has also tried to introduce STICS model for simulating crop phenology, finding the model become better representing interannual variations of LAI and net primary production. Overall, the introduction of crop module can improve ecosystem models in representing spatio-temporal variations of cropland ecosystems, making it an alternative choice to study regional and global croplands, how they may respond to climate change.
Despite the large differences in the complexity of introduced crop module, the agro-ecosystem models still have limitations in representing the crop growth dynamics, such as the morphology of crop organs, the grain quality, and the lack of nutrient cycling, particularly for micro-nutrients such as potassium. In addition, the process-based ecosystem models usually requires larger amount of computational resources following the same protocol of simulations. The consumption of computing resource by ORCHIDEE-crop is one magnitude larger than that required by pDSSAT and pAPSIM (Elliot & Wang, personal communication). It becomes a bottleneck for the application of agro-ecosystem models, though increasing computing power globally may gradually alleviate the pressure.
The Global Gridded Crop Model Inter-comparison (GGCMI) project brought different types of crop models together to perform simulations forced with consistent climate and management forcing (Elliot et al., 2015). This ongoing global effort will help us further understand the advantage and disadvantage of different crop models and reduce large uncertainties in estimating crop yield response to climate change at global and regional scale.
Objectives and structure of this thesis
The general goal of this PhD thesis is to describe the efforts using both statistical tools and processed based crop models to 1) detect climate change impacts on crop phenology and yield, identifying key climatic factors regulating crop yield variations and estimating the temperature sensitivity of crop yield, and 2) attribute the crop yield change to climate and management factors, at regional and global scale.
Table of contents :
CHAPTER 1 INTRODUCTION
1.1 THE IMPACT OF CLIMATE CHANGE ON CROPLANDS Phenology and yield Land surface energy and water exchange
1.2 CROP MODELS, FROM SITES TO THE GLOBE
1.3 OBJECTIVES AND STRUCTURE OF THIS THESIS
CHAPTER 2 DETECTING CLIMATE CHANGE IMPACTS ON MAIZE YIELD IN NORTHEAST CHINA SUMMARY
CHAPTER 3 ATTRIBUTING HISTORICAL TRENDS IN CHINA’S RICE GROWING SEASON BASED ON CALIBRATED ORCHIDEE-CROP MODEL SUMMARY
CHAPTER 4 REANALYZING GLOBAL CROP YIELD RESPONSE TO WARMER TEMPERATURE USING MANIPULATION EXPERIMENTS AND GLOBAL CROP MODELS
CHAPTER 5 GLOBAL IRRIGATION CONTRIBUTION TO WHEAT AND MAIZE YIELD
CHAPTER 6 CONCLUSIONS AND PERSPECTIVES