Context: Climate change will alter the global distribution of climatically suitable space for many species, including agricultural crops. In some locations, warmer temperatures may offer opportunities to grow novel, high value crops, but non-climatic factors also inform agricultural decision-making. These non-climatic factors can be difficult to quantify and incorporate into suitability assessments, particularly for uncertain futures. Objective: to demonstrate how qualitative and quantitative techniques can be combined to assess crop suitability with consideration for climatic and non-climatic factors. Methods: We carried out a horizon scanning exercise that used Delphi methodology to identify possible novel crops for a region in south-west England. We show how the results of the expert panel assessment could be combined with a crop suitability model that only considered climate to identify the best crops to grow in the region. Results and conclusions: Whilst improving climate and crop models will enhance the ability to identify environmental constraints to growing novel crops, we propose horizon scanning as a useful tool to understand constraints on crop suitability that are beyond the parameterisation of these models and that may affect agricultural decisions. Significance: a similar combination of qualitative and quantitative approaches to assessing crop suitability could be used to identify potential novel crops in other regions and to support more holistic assessments of crop suitability in a changing world.

Context: Against a background of unprecedented climate change, humanity faces the challenge of how to increase global food production without compromising the natural environment. Crop suitability models can indicate the best locations to grow different crops and, in doing so, support efficient use of land to leave space for, or share space with, nature. However, challenges in downscaling the climate data needed to drive these models to make predictions for the future has meant that they are often run using national or regional climate projections. At finer spatial scales, variation in climate conditions can have a substantial influence on yield and so the continued use of coarse resolution climate data risks maladaptive agricultural decisions. Opportunities to grow novel crops, for which knowledge of local variation in microclimate may be critical, may be missed. Objective: We demonstrate how microclimate information can be acquired for a region and used to run a mechanistic crop suitability model under present day and possible future climate scenarios. Methods: We use microclimate modelling techniques to generate 100 m spatial resolution climate datasets for the south-west of the UK for present day (2012–2017) and predicted future (2042–2047) time periods. We use these data to run the mechanistic crop model WOrld FOod STudies (WOFOST) for 56 crop varieties, which returns information on maximum crop yields for each planting month. Results and conclusions: over short distances, we find that the highest attainable yields vary substantially and discuss how these differences mean that field-level assessments of climate suitability could support land-use decisions, enabling food production whilst protecting biodiversity. Significance: We provide code for running WOFOST in the WofostR R package, thus enabling integration with microclimate models and meaning that our methodology could be applied anywhere in the world. As such, we make available to anyone the tools to predict climate suitability for crops at high spatial resolution for both present day and possible future climate scenarios.

Species distribution models (SDMs) have played a pivotal role in predicting how species might respond to climate change. These models most often rely on correlative methods, whereby a statistical relationship between species’ known occurrences and the climate of those locations is determined and then extrapolated to predict future distributions under climate change scenarios. Such modelling approaches, which seek to find and recreate patterns in species distributions, are not directly associated with the physiological processes that ultimately underlie species’ responses to climate. By incorporating these processes more explicitly into SDMs, the proximal limitations on species distributions can be better understood and subsequent predictions will be more robust over space and time.
This thesis presents research to promote and support the integration of physiological processes into SDMs. It explores major themes in model construction and use and demonstrates how process-based (mechanistic) models might be applied to predict future crop suitability.
In Chapter 2, I review the plant SDM literature and find that physiologically important variables are frequently neglected in models. Ten physiologically relevant variables for plants are identified and in Chapter 3, I present a new global climate classification (CCS) that accounts for variation in these aspects of climate. I show how the popular Köppen CCSs, for which boundaries of zones were chosen to reflect major vegetation patterns, do not entirely reflect the physiological processes that determine plant distributions. I discuss how predictions of climate-driven changes in plant distributions may be unreliable in areas where zone assignment using physiologically relevant variables is different to that of the Köppen systems.
In Chapter 4, I demonstrate the use of microclimate modelling techniques to generate 100m spatial resolution climate data for present and future time periods. I use these data to run the mechanistic crop model WOrld FOod STudies (WOFOST) and show how, by capturing spatial variation in climate suitability, microclimate data could provide better approximations of predicted yields and inform agricultural decision-making. Then, in Chapter 5, I show that incorporating interannual variability into climate suitability assessments or understanding the extent to which average climate data might obscure this variance is also important to consider, even when using mechanistic models.
Finally, I consider how mechanistic crop models might be applied to inform agricultural decisions. In Chapter 6, I demonstrate how the results from a mechanistic crop model may be combined with an expert-informed qualitative assessment of crop suitability to give a holistic understanding of the best crops to grow based on climatic and non-climatic factors. In Chapter 7, I examine how climate-driven changes to crop suitability may lead to conflict between agricultural land use and conservation. I model global crop suitability for current and future time periods and show that agricultural expansion is a major threat to remaining wilderness. I conclude that to protect wilderness and its many values, agricultural systems will need to be transformed.
Overall, this thesis shows why physiological process should become central in endeavours to understand the effects of climate change on species distributions and presents methods to achieve this in ecological research. It shows how reliable predictions of the impacts of climate change on crop suitability could help reconcile food security and conservation goals.

Species distribution models (SDMs) have played a pivotal role in predicting how species might respond to climate change. These models most often rely on correlative methods, whereby a statistical relationship between species’ known occurrences and the climate of those locations is determined and then extrapolated to predict future distributions under climate change scenarios. Such modelling approaches, which seek to find and recreate patterns in species distributions, are not directly associated with the physiological processes that ultimately underlie species’ responses to climate. By incorporating these processes more explicitly into SDMs, the proximal limitations on species distributions can be better understood and subsequent predictions will be more robust over space and time.
This thesis presents research to promote and support the integration of physiological processes into SDMs. It explores major themes in model construction and use and demonstrates how process-based (mechanistic) models might be applied to predict future crop suitability.
In Chapter 2, I review the plant SDM literature and find that physiologically important variables are frequently neglected in models. Ten physiologically relevant variables for plants are identified and in Chapter 3, I present a new global climate classification (CCS) that accounts for variation in these aspects of climate. I show how the popular Köppen CCSs, for which boundaries of zones were chosen to reflect major vegetation patterns, do not entirely reflect the physiological processes that determine plant distributions. I discuss how predictions of climate-driven changes in plant distributions may be unreliable in areas where zone assignment using physiologically relevant variables is different to that of the Köppen systems.
In Chapter 4, I demonstrate the use of microclimate modelling techniques to generate 100m spatial resolution climate data for present and future time periods. I use these data to run the mechanistic crop model WOrld FOod STudies (WOFOST) and show how, by capturing spatial variation in climate suitability, microclimate data could provide better approximations of predicted yields and inform agricultural decision-making. Then, in Chapter 5, I show that incorporating interannual variability into climate suitability assessments or understanding the extent to which average climate data might obscure this variance is also important to consider, even when using mechanistic models.
Finally, I consider how mechanistic crop models might be applied to inform agricultural decisions. In Chapter 6, I demonstrate how the results from a mechanistic crop model may be combined with an expert-informed qualitative assessment of crop suitability to give a holistic understanding of the best crops to grow based on climatic and non-climatic factors. In Chapter 7, I examine how climate-driven changes to crop suitability may lead to conflict between agricultural land use and conservation. I model global crop suitability for current and future time periods and show that agricultural expansion is a major threat to remaining wilderness. I conclude that to protect wilderness and its many values, agricultural systems will need to be transformed.
Overall, this thesis shows why physiological process should become central in endeavours to understand the effects of climate change on species distributions and presents methods to achieve this in ecological research. It shows how reliable predictions of the impacts of climate change on crop suitability could help reconcile food security and conservation goals.