Time, space and disciplinary dimensions considered

Heat stress occurs over hours to days, and involves biological and physical processes. Mammals need to maintain their internal body temperature constant (homoeothermy) under a fluctuating microenvironment. When external weather conditions venture outside a certain range (heat stress situation), maintaining homoeothermy strains animal traits (e.g. productivity, reproduction rate, etc.) and ultimately leads to death. However, many more aspects need to be considered to understand impacts of future climate induced heat stress on livestock at global scale. There are considerable challenges to extend the timescale to that of climate change. Throughout the year, weather regime varies and significantly alters local livestock traits only exceptionally (e.g. heat waves), whereas livestock traits also have their seasonality (e.g. lactation peak). Heat tolerance also varies seasonally at animal (e.g. acclimation) and herd (e.g. harvest effect) scales, and extending the timescale of modelling should thus not overlook these important high temporal resolution features. Over longer time scales, farmers can switch for breeds more tolerant to heat or adopt heat-mitigating management strategies, although selection historically leads to more productive but less heat tolerant breeds.

Similarly, high spatial resolution features have to be accounted for while seeking global coverage. There is a large differential in vulnerability to heat stress between species, breeds and type of management or climatic and sanitary conditions: the current spatial variability in livestock species, breeds and management systems partially mirrors that of climate. In addition, management measures can mitigate heat exposure and impacts but can have a significant cost, or rely on resources whose availability might drastically change (energy, water). On long time scales, the spatial distribution of livestock species, their management and the capacity to adapt will evolve, for climatic and non-climatic (e.g. diets, trade, policies, competitiveness, etc.) reasons. Factoring these factors in largely broadens the discipline scope to agricultural sciences, genetics, geography, climate, hydrology and socio-economics.

A brief review of existing types of modelling

When reviewing literature, we distinguished seven broad categories of relevant models (not exhaustive). The first type of models assesses the various routes of thermal exchange of an animal with their environments, and how various traits can be affected. A second category of models simulates the climatic conditions surrounding animals in the farms and human interventions (such as barn design, etc.). These first two types of models are the only ones able to render, respectively, the complexity of various routes of heat exchange and the climatic conditions surrounding animals. A third category of models, more parsimonious and widely developed, estimates statistical relationships between simplified metrics of heat (e.g. temperature humidity index) and animal traits based on observed data. A fourth type of model focusses on regional heat waves monitoring and provides real-time or forecasted indices of simple livestock heat stress risk. The fifth category of models integrates more complex processes at the farm scale, including farmer management decisions and farm economic or environmental impacts. The sixth type of model looks at regional to global-wide economics of the livestock sector over long time scales. Finally, the seventh type of model applies statistics to genetic material of livestock put under various heat conditions to identify genotypes responsible for tolerance to heat stress.

Potential solutions for bridging the gap and limits

There is thus a rich variety of models dealing with the relevant processes, but they are not yet connected over the required range of spatial and temporal scales. Linking small-scale heat stress models (e.g. Vitali et al., Reference Vitali, Segnalini, Bertocchi, Bernabucci, Nardone and Lacetera2009) on the one hand, and global economics models (e.g. Havlík et al., Reference Havlík, Leclère, Valin, Herrero, Schmid, Soussana and Obersteiner2015) on the other seems a promising avenue. However, although climatic variables are available at a daily resolution at 0.5° latitude–longitude grid, significant challenges remain for incorporating features related to high temporal and spatial resolution. Some additional processes related to acclimation, herd dynamics and farm management at the seasonal scale need to be incorporated, but could be inspired from existing farm and herd dynamics models. Similarly, while high resolution global gridded distribution of livestock species and traits are already available (Herrero et al., Reference Herrero, Havlík, Valin, Notenbaert, Rufino, Thornton, Blümmel, Weiss, Grace and Obersteiner2013), it needs to be enriched with additional parameters (e.g. breeds, seasonal cycle of traits) to meet the requirements for most accurate statistical models.

Building on the experience of crop model developments, a gridded version of statistical stress-strain models could be developed. It would incorporate daily weather variables and project spatially explicit heat stress impacts on the various traits used by global economic models, for various categories of livestock species and management conditions. Such a model should be modular to offer the possibility to develop and select alternative formulations either simple enough to harvest parameter values from the large and increasing amount of literature of heat stress case studies, or more complex allowing richer analysis in data rich environments. This would be helpful to fill knowledge gaps, such as processes ranging from seasonal to decadal timescale, or from animal to herd to farm scale. An open access (data and code) approach could potentially scale-up the sharing and collection of missing data, enforce the developments of data standards, facilitate exploiting the richness of various existing models and increase usability among various communities.