https://orcid.org/0000-0001-8068-5772
Figures
Abstract
Cattle farming, a vital component of the agricultural sector, is highly sensitive to environmental conditions, with heat stress effects accounting for a substantial share of its economic losses. Temperature and humidity are the two main climatic factors governing thermal balance in cattle, and the Temperature-Humidity Index (THI) provides a non-invasive estimate of their combined physiological impact on livestock.
We employed novel, global, high-temporal-resolution global projections of hourly THI extending to the end of the 21st century. Using validated thresholds for mild and severe thermal stress, we project the duration and intensity of thermal stress periods, as well as the frequency and duration of thermal stress waves, under different greenhouse gas emission scenarios.
Our results show that in the future, severe heat stress is likely to become a threat to cattle farming in several regions of the world, particularly under a business-as-usual scenario of greenhouse gas emissions. By cross-referencing our climatic projections with cattle distribution data, we identify ‘hotspots’ where future environmental conditions are likely to create challenges for cattle farming. The Americas, Africa, and Southeast Asia emerge as the regions with the largest projected increases in thermal stress conditions, underscoring the urgent need for targeted, cost-effective mitigation and adaptation strategies.
Citation: Neira M, Georgiades P, Proestos Y, Economou T, Araya J, Malas S, et al. (2026) Climate change and thermal stress in cattle: Global projections with high temporal resolution. PLOS Clim 5(1): e0000761. https://doi.org/10.1371/journal.pclm.0000761
Editor: Valerio Lembo, CNR: Consiglio Nazionale delle Ricerche, ITALY
Received: July 14, 2025; Accepted: December 4, 2025; Published: January 28, 2026
Copyright: © 2026 Neira et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data analyzed in this study have been deposited at the World Data Center for Climate at DKRZ, and are publicly available under a CC BY 4.0 license at https://doi.org/10.26050/WDCC/THI. The code used for performing the data analysis described in this study is available, under an MIT license, at https://github.com/pantelisgeor/cattle_THI.
Funding: MN, YP, TE, JA and JL are supported by the EMME-CARE project, which receives funding from the European Union’s (EU) Horizon 2020 Program, Grant Agreement No. 856612, as well as supplementary funding by the Government of Cyprus. PG is supported by the PREVENT project, which receives funding from the EU’s Horizon Europe Program, Grant Agreement No. 101081276. MO, DS and GH received funds from the Agricultural Research Institute’s Budget for Research. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
The effects of heat stress on cattle (Bos taurus) create important challenges for the farming sector, negatively affecting crucial factors such as animal health and welfare, productivity, and reproductive performance. The financial burden associated with heat stress accounts for a significant share of the losses incurred by the cattle industry. For example, data from the USA have shown that heat-related losses affecting lactating cows can reach between 1.0 and 1.8 billion USD per year, accounting for approximately 64% of the national yearly losses of the dairy sector [1–4]. Also in the USA, the effects of heat stress on ‘dry’ (i.e., non-lactating) dairy cows can impose additional revenue reductions of up to $810 million per year [5].
When subjected to extreme heat, cattle exhibit a variety of physiological and behavioral changes, including reduced feed intake to moderate metabolic heat production, increased water consumption, elevated respiration rates, endocrine alterations, and increased standing time [2,6–8]. These disturbances lead to marked declines in critical productivity factors such as weight gain, as well as milk yield and quality. In addition, cows affected by heat stress display increased susceptibility to diseases, such as mastitis, which can further deteriorate their general health and contribute to the reduction of milk quality [9].
Heat stress is also associated with altered estrus behavior, reduced pregnancy rates, and an increased risk of miscarriage, resulting in significantly decreased reproductive efficiency during hot periods [8,10]. When the magnitude and/or duration of heat stress overwhelm the animal’s homeostatic mechanisms, the subsequent pathological and physiological stress can be lethal [11]. This effect has been evidenced during particularly intense heatwaves, which have caused extensive mortality in cattle herds around the world [11–14].
The effects of climate change on cattle farming operations around the world are potentially severe. Rising temperatures and increasingly frequent heatwaves are likely to drive reductions in milk yield and quality, while alterations to precipitation patterns and prolonged drought periods can reduce water, pasture, and feed availability. [15,16]. Furthermore, climate change can indirectly affect cattle farming by triggering ecological changes that favor the transmission of infectious and/or parasitic diseases, and extreme weather events have the potential to disrupt the infrastructure and supply chains required to sustain this industry [16–18].
Cattle rely mainly on evaporative transpiration for thermal regulation. Therefore, their ability to cool down is strongly modulated by the relative humidity (RH) of their surroundings, becoming progressively less efficient with increasing humidity [19,20]. For this reason, any attempt to evaluate the effect of heat stress on cattle should consider the combined effects of temperature and humidity. The Temperature Humidity Index (THI), a metric that combines the effects of air temperature and relative humidity, has been widely used to assess the impacts of heat stress on cattle in a variety of environments [20–26].
THI is conventionally computed in daily time steps due to limitations in data availability and extensive computing requirements. Unfortunately, this temporal definition does not allow for accurate estimation of the varying thermal conditions occurring throughout a diel cycle, which in turn limits our ability to assess important factors such as the efficiency of nocturnal heat dissipation (associated with cooler temperatures during the night), and thermal load accumulation during extended periods of heat stress. Recently, a novel Machine-Learning (ML) method was developed to downscale the temporal frequency of THI projections from daily to hourly values, offering long-term global projections at high temporal resolution and under different greenhouse gas emission scenarios [27]. By combining this information with validated metrics of mild and severe thermal stress for cattle, we generated global projections of future thermal stress duration and intensity, and computed estimates for mid-century (2020 to 2050) and late-century (2070-2100) periods. Furthermore, we identified regions where expected changes in environmental parameters are likely to create particularly challenging conditions for cattle in the future, hence requiring the implementation of specific adaptation measures.
Methods
THI data
All THI data analyzed in this report were produced by the machine learning (ML)-based temporal downscaling approach method described in [27]. This methodology employed an Extreme Gradient Boosting (XGBoost) regression model trained on data available in the 5th-generation global historical climate dataset issued by the European Centre for Medium-Range Weather Forecasts Reanalysis (ERA5) [28]. This historical reanalysis data was used to estimate hourly THI values from daily climate and temporal variables using the formula proposed by [29]:
This approach was designed to address the limitation of daily-level THI data, which don’t capture adequately the diurnal thermal load variations and cumulative heat stress effects, especially during heatwave periods when cooling during night-time is insufficient. The trained model was subsequently utilized to generate hourly THI projections through the end of the 21st century using bias-adjusted daily climate projections from the Coupled Model Intercomparison Project Phase 6 (GDPP-NASA-NEX-CMIP6) [30].
The input data for the model comprised climate variables (daily minimum, maximum and average temperature, daily average relative humidity), daily average THI (calculated from daily average temperature and relative humidity), temporal variables (day length, day of year, and hour of the day) and land-sea mask. The XGBoost model was trained incrementally on approximately 130 billion examples of ERA5 historical data (1980-2017), with ocean-only areas excluded from training and inference. The incremental training approach was necessitated by the computational requirements of processing the extensive dataset, with the model trained on CPUs using 128 parallel processes. Input variables were scaled to the 0-1 range using MinMaxScaler, and early stopping was implemented during each training epoch to prevent overfitting. Model performance was evaluated on a validation dataset (February 2018-December 2020) not seen during training, achieving a mean absolute error (MAE) of 3.4 THI units, mean squared error (MSE) of 19, and coefficient of determination (R2) of 0.94. The model demonstrated best performance in equatorial regions with MAE around 1 THI unit, while higher errors (4-6 THI units) were observed in mountainous regions due to larger diurnal temperature variations at high altitudes. Spatial analysis revealed no systematic bias across the dataset, confirming the model’s robustness for hourly THI predictions in most regions relevant to heat stress applications. The trained model was also successfully validated on ERA5-Land data at 9 km spatial resolution, demonstrating transferability to higher-resolution datasets. Further details on the design and training of the ML model, as well as validation of the resulting dataset, are available in [27].
The dataset used for our study (accessible online at https://doi.org/10.26050/WDCC/THI) contains hourly THI values for 12 different GDPP-NASA-NEX-CMIP6 climate models (listed in S1 Table), spanning until the end of the present century, with global geographic coverage, and 0.25° spatial resolution. The THI projections presented in our analysis were built using an ensemble approach that combines the outputs of these 12 models, thereby generating projection intervals. For each grid point, hourly THI values correspond to the mean value of the 12 aforementioned GDPP-NASA-NEX-CMIP6 climate model outputs, and 95% confidence intervals (representing ensemble model uncertainty). Furthermore, our projections consider GHG emission scenarios for two different Shared Socioeconomic Pathways (SSP): SSP2-4.5 and SSP5-8.5. These two scenarios were chosen because they represent one ‘middle of the road’ scenario (SSP2-4.5, characterized by a balance between challenges to mitigation and adaptation strategies), and one ‘business as usual’ scenario (SSP5-8.5, which is largely dominated by progressively increasing emissions and significant challenges to mitigation) [31].
Uncertainty quantification
Ensemble uncertainty
The uncertainty in our multi-model ensemble approach was quantified through the inter-model variability across the 12-model ensemble. For each grid cell s, and time step t, we computed the following metrics:
- Ensemble mean:
(2)
wheredenotes the THI value relating to climate model m.
- Inter-model dispersion (standard deviation):
(3)
- 95% confidence interval for ensemble mean:
(4)
with critical t- value t0.025,11 = 2.201. - Uncertainty metric (confidence interval width):
(5)
The t-distribution was used as opposed to the conventional approximations of normal distribution due to the small ensemble size ), since the t-distribution accounts for estimation uncertainty in the variance [32].
Machine-learning prediction uncertainty
In addition to the multi-model ensemble uncertainty, we also quantified the prediction uncertainty in the temporal downscaling from the ML model output (stemming from estimating the ML model). For this purpose, we used the held-out portion of the data, for the period spanning 2018-2020, which was used for evaluating the performance of the model, as described in [27,33]. We quantified the prediction uncertainty using a non-parametric bootstrapping approach applied to model residuals (THI minus prediction) at each 0.25° grid cell [34]. For each spatial grid point, residuals were resampled with replacement (1,000 iterations) to estimate Mean Error (ME), defined as:
where N is the number of samples (time steps), yt is the t-th ground truth value and xt is the t-th prediction. We also calculated the standard error (SE) of the prediction error, computed as:
where is the mean of the b-th bootstrap sample and
is the overall mean residual. Spatial heterogeneity in uncertainty was preserved by performing the analysis independently per grid cell.
Thermal stress definitions
To estimate the duration and load of future thermal stress, we used the reference THI value of 68.8 to indicate the threshold for mild thermal stress, and THI=84 as the threshold for severe thermal stress. The threshold for mild stress was selected following a recent report by [3] in which the authors performed a systematic review of the scientific literature, covering a wide range of ecological conditions in 22 countries, and spanning 20 years (1999-2020), identifying THI 68.8 as the median threshold for the onset of heat stress on dairy cattle, above which negative effects start to be observed in productivity (i.e. milk yield and quality), fertility, and survival. The threshold for severe thermal stress was selected following the guidelines of the Livestock Weather Safety Index, where THI 84 is identified as the threshold for emergency conditions, above which heat-related mortality occurs in healthy feedlot cattle without management interventions [35–38].
Estimation of thermal stress duration and thermal stress load
Following the approach proposed by [4], we define ‘thermal stress duration’ (TSD) as the time interval (expressed as number of hours) during which THI values remain above a specific THI threshold, and ‘thermal stress load’ (TSL) as the summation of THI units above threshold throughout the TSD period. Formulas 8 and 9 below were used to estimate the daily TSD and TSL, respectively. A graphic representation of these concepts is shown in Fig 1.
The curved line represents an idealized version of the fluctuation of THI over a 24h period. The continuous red line marks the selected THI threshold (in this example, THI 84, corresponding to severe thermal stress). Thermal stress duration is defined as the time interval during which the THI values remain above the selected threshold (represented by the space between the two red dashed lines). Thermal stress load is defined as the area under the curve delimited by the selected THI threshold (area in pink).
The yearly TSD values represent the cumulative number of hours above a specific threshold for all days of the corresponding year:
Similarly, yearly TSL values represent the cumulative TSL for all days within a specific year:
Reference periods
We provide projections for two future reference periods: mid-century (2020 to 2050) and late-century (2070 to 2100). For the estimation of the magnitude of change from historical values, our historical reference period spans the years 1990 to 2019. For any of the aforementioned periods, the projected values of any parameter correspond to the mean of yearly values for the period in question (e.g. the TSD above THI 68.8 for the mid-century period, corresponds to the mean of the yearly TSD values above THI 68.8 for the period between 2020-2050).
Cattle distribution
Global distribution of cattle was obtained from the 4th release of the Gridded Livestock of the World (GLW4) dataset, maintained by the Food and Agriculture Organization of the United Nations (FAO) [39,40]. For this study, the GLW4 cattle distribution data were interpolated to match the 0.25° spatial resolution of the Temperature-Humidity Index (THI) dataset. The cattle distribution data were then combined with datasets representing the transitions in TSD and TSL between the historic and end-of-century reference periods, under the SSP5-8.5 scenario. This integration enabled the creation of bi-variate choropleth maps, which quantify the combined effects of cattle distribution and thermal stress indicators. For the parameters analyzed in our study, we defined ‘hotspot’ as any grid cell that (a) contain at least 3,000 cattle heads, and (b) is expected to undergo increments in TSD or TSL above the global median when comparing historical averages to end-of-century forecasts under SSP5-8.5.
Furthermore, to estimate the global proportion of cattle at risk of experiencing extended periods of heat stress, we estimated the percentage of hours per year above threshold for each grid cell, sorted them into quartiles based on this criterion (i.e. 0-25%, 25-50%, 50-75% and 75-100% of yearly hours above threshold), and aggregated the number of cattle heads contained within all cells belonging to each quartile.
Thermal stress waves
Periods of consecutive days with unusually hot thermal conditions, commonly known as ‘heatwaves’, can be particularly dangerous for the health of both humans and animals. In spite of being the focus of intensive research for their biological and socioeconomic impacts, no consensus exists regarding the definition of such periods [41–43]. A recent systematic review found a wide variation of definitions in the literature, with up to 21 different sets of criteria used for this purpose [42].
In 2021, the Intergovernmental Panel on Climate Change (IPCC) defined a heatwave as “a period of abnormally hot weather, often defined with reference to a relative temperature threshold, lasting from two days to months” [44]. Therefore, for this component of our work, we combined this IPCC definition of heatwave with the data presented by [14] and [45], showing that the thermal recovery threshold for dairy cows (i.e., the threshold under which animals start dissipating accumulated heat) corresponds to THI 74. Because our work is based on THI, a well-established metric that incorporates the effect of both heat and humidity, we use the term ‘thermal stress wave’ (TSW) instead of ‘heatwave’. Accordingly, we define a TSW as a period of at least 48 consecutive hours during which all hourly THI values are 74.
For our analysis, we estimated the maximum and mean TSW duration for each year, as well as the projected frequency of TSWs (i.e., number of TSW events per year), and the number of TSW hours/year (i.e., the cumulative duration of all TSWs projected within a year).
Results
Uncertainty quantification
We quantified uncertainty for both the ML model used to temporally downscale daily projections to hourly THI values, and for our 12-model ensemble approach. These uncertainties are quantified in terms of the standard error, as described in the Methods section. The mean and maximum standard errors for each of the datasets we produced are presented in detail in Table 1.
The uncertainty introduced by the ML model was at least an order of magnitude smaller than that of the ensemble variability and, as such, it was considered negligible for the scope of this study. For the remaining analyses, the standard errors were relatively small percentage-wise when compared to the magnitude of the respective values, indicating robust estimation. Notably, for the Thermal stress load, we present the standard error for both SSP2-45 and SSP5-85 scenarios, as their values differed significantly, a divergence not observed in the other analyses. Comprehensive maps illustrating the mean value for each dataset and the spread of the ensemble are provided in the Supplementary Information (S1 Fig to S5 Fig).
Thermal Stress Duration (TSD)
Our model calculations project a global and gradual increase in the average number of hours above the thresholds for both mild and severe thermal stress, for both GHG emissions scenarios (Fig 2). Table 2 shows the average TSD values for the different scenarios and periods considered in this work. Under a ‘business as usual’ scenario (SSP5-8.5), by late century, the average year will be subjected to 1.7 fold increases in the average annual TSD for mild stress, and up to a 12-fold increase in severe TSD compared to historical averages. Under the moderate SSP2-4.5 (stabilization) scenario, the end of the century is expected to see increases of 1.5- and 5.8-fold, relative to the historical averages, for mild and severe TSD, respectively.
Solid lines show yearly global mean values. Light-colored bands show 95% confidence intervals, representing model uncertainty.
Values shown correspond to the global average number of hours/year above threshold for each period and GHG emissions scenario. Numbers in parentheses show the corresponding fold increase from historical period averages. All values are rounded to one decimal place.
As shown in panel A of Fig 3, several regions of the world -particularly around the tropics- already experience THI conditions above the threshold for mild thermal stress for more than 8,000 hours per year (equivalent to >90% of the total number of hours in a year). However, as the century progresses, temperate latitudes will see progressive increases in the number of hours per year above the threshold for mild thermal stress. In contrast, panel B of Fig 3 shows that conditions for severe thermal stress have not been nearly as prevalent during our historical reference period, with most tropical regions experiencing less than 1,500 hours per year (i.e. <20% of the time) above THI 84, and temperate regions either not reaching this threshold at all, or doing so very rarely. However, our projections suggest that by mid-century, even under the relatively optimistic SSP2-4.5 emissions scenario, most of the world will experience at least some hours per year above THI 84. By the late century, under SSP 5-8.5, some tropical areas could spend 4,000 hours (
50% of the time) above this threshold in an average year.
Panel A shows mean TSD above the threshold for mild stress. Panel B shows mean TSD above the threshold for severe stress.
To identify the regions of the world where future shifts in TSD are expected to be relatively severe, we plotted the expected differences in average yearly TSD between our historical and late-century periods (Fig 4) under SSP5-8.5. By cross-referencing this information with the latest available data on worldwide cattle distribution (GLW4 dataset; [39,40]), we identify hotspots with high cattle density that are also likely to experience particularly large increases in TSD in the future. As shown in Fig 5, several such hotspots for mild TSD can be identified in North and South America, Central and Eastern Africa, South-East Asia, and Oceania, while hotspots for severe TSD are mostly located in South America and South-east Asia.
Maps plot the difference in average hours/year above threshold between the late-century and historical reference periods. The two upper maps correspond to expected changes in TSD above the threshold for mild stress, and the two lower maps show expected changes in TSD above the threshold for severe stress. Red hues mark expected increments, while blue hues mark expected reductions.
Panel A shows data for TSD above the threshold for mild stress. Panel B shows data for TSD above the threshold for severe stress. Both panels correspond to increases in TSD expected under SSP5-8.5. Blue hues show expected increases in yearly average TSD towards the end of century, relative to the historical average. Pink hues indicate the current cattle density per grid cell. Yellow areas denote ‘hotspots’ (i.e. regions that contain at least 3,000 cattle heads per grid cell and are expected to suffer increases in TSD above the global median between the historical and end-of-century reference periods).
As can be seen in Fig 6, currently 41% of the world’s cattle experience conditions associated with mild thermal stress (THI
68.8) very frequently (75 to 100% of the hours in a year), while
26% of the world’s cattle either do not experience these conditions, or do so rarely (0 to 25% hours per year). However, in a business as usual scenario -and assuming no major shifts in global cattle distribution- the proportion of cattle experiencing mild stress conditions very frequently is expected to increase to
59% by the end of the century, while the percentage of cattle that rarely experiences these conditions is projected to drop below 13%.
Each color represents a different frequency of hours above threshold, as indicated on the inset (e.g., blue lines show the global proportion of cattle at risk of experiencing THI above threshold between 0 and 25% of the total number of hours per year, etc.) Light-colored bands show 95% confidence intervals, representing model uncertainty.
When it comes to severe thermal stress, the projected outcomes are striking. Currently, 99% of the cattle in the world either do not experience THI
84, or do so rarely (≤25% hours per year). However, under a business-as-usual scenario, by the end of the century the proportion of cattle that rarely experiences this level of stress would drop below 80%,
18% of the world’s cattle would experience severe thermal stress between 25% and 50% of the time, and
3% of the cattle would experience these conditions between 50 and 75% of the time (Fig 6). Interestingly, changes in the proportion of cattle exposed to THI
84 would be much less pronounced under SSP2-4.5, with over 96% of cattle remaining in the category ‘rarely or never’ (0-25% of hours in a year) at the end of the century (Fig 6).
Thermal Stress Load (TSL)
The projected trends in global TSL fluctuations follow patterns similar to those observed for TSD, showing progressive increases under both emissions scenarios, which become more pronounced during the second half of the century (Fig 2). As shown in Table 3, the estimated increase between the historical and mid-century periods remains relatively constant (about 1.7-fold), independent of the emissions scenario. However, by late-century, a somewhat larger change is expected under SSP5-8.5 (2.3-fold) than under SSP2-4.5 (2-fold).
Values shown correspond to the global average thermal load/year above threshold for each period and GHG emissions scenario. Numbers in parentheses show the corresponding fold increase from historical period values. All values are rounded to one decimal place.
Interestingly, our data shows that during our historical reference period, TSL above the threshold for severe stress was extremely low worldwide, with the global average approaching 0 (Table 3 and Fig 7). However, as more regions of the world start experiencing conditions above this threshold over time, the global average is expected to increase by mid-century to 0.5 - 0.6 (for SSP2-4.5 and SSP5-8.5, respectively), and could eventually reach up to 2.3 by the end of the century under a business-as-usual scenario (Table 3).
Panel A shows mean TSL above the threshold for mild stress. Panel B shows mean TSL above the threshold for severe stress.
The maps presented in Fig 7 depict the expected progression of the mean annual TSL through the time periods considered in this work. Most regions of the world currently experience at least some TSL above the threshold for mild stress, and significant increases are expected under both emissions scenarios, particularly in and near tropical regions (Fig 8). For TSL above the threshold for severe stress, the projection follows a somewhat different pattern: Although most parts of the world do not currently experience meaningful thermal stress loads in this range, by mid-century, even under the moderate SSP2-4.5 scenario, most regions are likely to experience at least some severe TSL, where the steepest increases are expected between latitudes 40oN and 40oS (Fig 8).
Maps plot the difference in average hours/year above threshold between the late-century and historical reference periods. The two upper maps correspond to expected changes in TSL above the threshold for mild stress, and the two lower maps show expected changes in TSL above the threshold for severe stress. Red hues mark expected increments, while blue hues mark expected reductions.
When we cross-reference our projections for late-century mean yearly TSL under SSP5-8.5 with the current global distribution of cattle (Fig 9), we identify relevant hotspots for mild TSL located mainly in the Americas, the Caribbean, Africa, and South-east Asia. Hotspots for severe TSL are found mainly in South America and Asia.
Panel A shows data for TSL above the threshold for mild stress. Panel B shows data for TSL above the threshold for severe stress. Both panels correspond to increases in TSL expected under SSP5-8.5. Blue hues show expected increases in yearly average TSL towards the end of century, in reference to the historical average. Pink hues indicate current cattle density per grid cell. Yellow areas denote ‘hotspots’ (i.e. regions that contain at least 3,000 cattle heads per grid cell, and are expected to suffer increases in TSL above the global median between the historical and end-of-century periods).
Thermal Stress Waves (TSW)
Fig 10 and Table 4 show the data corresponding to our projections for TSW at a global level. Interestingly, our results suggest that the global average frequency of TSWs will not increase extensively (1.6 fold) relative to our historic reference period (Table 4), and is likely to change little from 2020 until the end of the century, at 2.7 events per year, independent of the emission scenario (Fig 10). However, during the same period, the number of TSW hours per year is expected to grow progressively, increasing by factors of
3.7 and
5.3 under SSP2-4.5 and SSP5-8.5, respectively (Table 4). Together, these results suggest that TSW events are likely to become progressively longer over time. Our results suggest that both the mean and maximum duration of TSWs are expected to increase towards the end of the century, under both emission scenarios. Even under the relatively moderate SSP2-4.5 scenario, the average TSW duration is likely to increase by factors of 2.7 and 3.5 (for mid- and late- century, respectively) in comparison to our historic reference period (Table 4). As expected, these changes would be more pronounced under SSP5-8.5, with increases of up to 4.7 fold for the average event duration, and 3.7 fold for the maximum event duration.
Upper left panel shows the average number of TSWs per year. Upper right panel shows the total number of hours per year that are expected to be part of TSWs. The lower left panel shows the maximum expected duration of TSWs, and lower right panel shows the mean expected duration of TSWs. Solid lines represent projected values. Light-colored bands show 95% confidence intervals, representing model uncertainty.
Values shown represent global averages for each period and GHG emissions scenario. Numbers in parentheses show the corresponding fold increase from historical period averages. All values are rounded to one decimal place.
Fig 11 presents the transition between our historic reference period and the end of the century for several TSW parameters (the transition between historical and mid-century reference periods for the same parameters is shown in S6 Fig). As shown in panel A, several regions, particularly in and near the tropics, will experience a reduction in the average number of TSWs per year. However, panel B of the same figure shows that in most of these regions the average number of TSW hours per year will increase considerably, suggesting that because of the magnitude of the expected increase in the duration of TSW, these regions will shift from experiencing several short TSWs per year, to fewer but significantly longer events. In fact, in tropical regions of the Americas, Africa, and South-east Asia, the average number of TSW hours per year is projected to more than double as early as the mid-century, even under SSP2-4.5 (S7 Fig).
A) TSW frequency. B) Number of TSW hours per year. C) Average TSW duration. D) Maximum TSW duration. For all panels, maps on the left side correspond to predictions under SSP2-4.5, and maps on the right side correspond to projections under SSP5-8.5.
Discussion
In this work, we employed a recently developed ML model capable of downscaling daily-level climate predictions [27] to generate hourly THI projections on a global scale, under different emission pathways. The uniquely high temporal resolution of this dataset allowed us to project the duration and intensity of thermal stress periods, as well as the frequency and duration of thermal stress waves. We estimated the uncertainty for both the ensemble and the ML model, and found uncertainty from the ML model to be negligible compared to that of the ensemble. Therefore, all uncertainty estimates presented correspond to the ensemble.
Our results suggest that conditions associated with both mild and severe thermal stress in cattle are likely to increase in duration and intensity toward the end of the century, independent of GHG emissions scenario. Because several parts of the world (particularly in the tropics) already endure THI conditions 68.8 during a significant portion of the year, the magnitude of the expected changes in global average TSD and TSL for mild stress is expected to be comparatively modest. However, our projections show that conditions associated with severe thermal stress (THI
84), which were rare (or nonexistent) throughout the world during the historical reference period, will become more prevalent in the following decades. By mid-century, most regions are projected to experience at least some hours above this threshold during an average year. Importantly, some of the most significant changes projected by our analysis correspond to increases in global TSD and TSL associated with severe thermal stress, indicating that not only will these conditions become much more prevalent, but also that in some areas THI is likely to increase above 84. These results are consistent with previous research that has projected alarming increases in the frequency, duration, and intensity of severe thermal stress for livestock in the coming decades, at both the global and regional levels [46–48]. Considering that such environmental conditions are associated with potentially lethal effects on healthy animals, this trend indicates important future challenges for the cattle industry in those regions.
It should be noted that the areas in which the highest increases in TSD and TSL are projected do not necessarily coincide with the areas generally identified as hotspots of extreme heat caused by global warming [49]. This is because our analysis is not based on temperature alone, but is instead based on THI units, which reflect the combined effects of both temperature and humidity. Therefore, high humidity conditions can amplify the impact of even moderate increases in temperature. Conversely, low humidity conditions can buffer the effect that high temperatures have on THI values.
By cross-referencing our THI predictions with the geographic distribution of cattle, we identify hotspots of increased risk for cattle farming. Mild thermal stress hotspots are particularly prevalent in South and Central America, Central and Eastern Africa, South-east Asia and Oceania, while severe thermal stress, hotspots are mostly found in South and Central America (particularly in Venezuela, Colombia, Brazil, and Mexico), as well as South- and South-east Asia (Pakistan, Vietnam, Cambodia, Thailand, and Indonesia).
To assess the risks to the cattle industry across the two SSPs considered in our work, our projections assume a spatial distribution of cattle fixed to the baseline year 2020 (GLW4). This approach aims to isolate the direct effects of climate variability from confounding factors such as changes in agricultural practices or cattle relocation. We acknowledge that cattle distributions may adjust in response to climate pressures or socioeconomic factors, potentially altering the projected risks. However, modeling these dynamic changes is beyond the scope of this study because of their significant complexity, especially at the global scale. By assuming a constant cattle distribution, climate-driven risks become manifested in a transparent context. Furthermore, this approach is consistent with previous research, notably [48], [50], and [47], who similarly used fixed livestock distributions to project future climate-related impacts.
Our results also indicate that the proportion of cattle experiencing severe thermal stress strongly depends on the emissions scenario. These results corroborate previous reports which indicate that the proportion of cattle exposed to thermal stress is strongly dependent on future GHG emission scenarios and development pathways [51,52], and that shared socio-economic pathways leading to marked reductions in GHG emissions are likely to result in significant reductions in the proportion of the cattle in the world exposed to thermal stress [3].
TSW events are particularly harmful to cattle, with previous studies showing that the mortality of adult animals increases significantly both during and immediately after such events [12,14,53]. Although our data suggest that on a global level the average number of TSW events per year is likely to remain relatively constant during the rest of the century, this apparent stability results from a complex pattern of variations across different geographic locations, with some regions shifting toward more frequent TSWs, while other regions shift towards fewer, but significantly longer events. Considering that the rates of heat-related cattle mortality increase with the duration of TSWs [12], the projected increases in both the average and maximum duration of TSWs represent significant future challenges for the cattle industry, especially in the tropics.
It is important to mention that the thresholds used in our definitions of severe thermal stress (THI and TSW (THI
74) either coincide with or are above the thresholds for thermal stress onset in several livestock groups of major economic importance. Bos taurus indicus, a subspecies of cattle that is considered to have increased physiological resistance to heat and is therefore widely farmed in tropical regions, is reported to begin experiencing thermal stress when THI conditions reach values
74 [54,55]. Similarly, the onset of heat stress has been reported to occur at THI
70 in goats, and THI ≥72 in sheep [56,57]. Furthermore, ([46] identified THI 83 as the threshold for “moderate to severe heat stress for highly relevant livestock species (cattle, swine and poultry)”. Taken together, these indicators suggest that our projections regarding severe thermal stress and TSW have implications not only for different cattle breeds, but also for the livestock industry at large. More research is required to assess the precise effects of these environmental phenomena on different livestock species.
Our results suggest that regions located between latitudes 20°N and S are projected to experience the highest increases in thermal stress parameters related to THI. These results are in line with previous research that identified similar geographic patterns [46]. In contrast, cattle farming areas in latitudes above 40°N are likely to see the lowest future increases in this type of thermal stress. Our projections for TSD, TSL, and TSW, point to parts of the Americas, Africa, and South-east Asia as regions where cattle are likely to experience the strongest growth in risks associated with thermal stress. These results are in general agreement with a recent report that identifies these regions among those with the highest projected increases in heat-related risks for dairy cattle [3].
In high-income countries, the negative effects of high THI on cattle productivity can potentially be mitigated through various technologies and strategies, including structural adaptations (e.g., the expansion of shaded areas to shield animals from direct sun exposure, improved shed design to maximize air flow, increased number of hydration stations, installation of cow showers and fogging devices, etc.) and management strategies (e.g., changes in feed composition and feeding schedules to reduce metabolic heat production during peak thermal stress times, seasonal management of calving, etc.) [4,58–60]. However, most of the hotspots identified in the current study are located in low- and middle-income regions, where the implementation of such technologies and practices can be challenging due to economic constraints. As a result, cattle in these areas will remain highly vulnerable to the negative effects of high THI values, emphasizing the need for low-cost, scalable solutions and proactive international support. The adoption of targeted local genetic improvement programs for breeding heat-tolerant animals could offer a cost-effective and sustainable strategy, although accurate identification of productive phenotypes associated with resilience to environmental disturbances remains challenging [59,61]. In addition, innovative and interdisciplinary tools and concerted practices must be developed to mitigate the adverse impacts of thermal stress on livestock.
Finally, we should also point out some relevant limitations of our study: by focusing on temperature and humidity (the two parameters used to estimate THI), we have excluded other variables that influence thermal stress, such as solar radiation, cloud cover, and airflow. The reasons for excluding these factors are multi-layered: for one, Global Climate Models (GCMs), such as those in CMIP6, are fundamentally based on the physics of thermodynamics and fluid dynamics [62]. As such, parameters like near-surface air temperature and humidity are core prognostic outputs that are simulated with a relatively high degree of confidence and consistency across models. On the other hand, parameters such as cloud fraction, or surface radiation, are not fundamental thermodynamic outputs, but are instead products of complex, parameterized sub-models that incorporate factors such as convection schemes, cloud microphysics, and radiative transfer, among other. These parameterization schemes can differ substantially from one GCM to another, leading to a much larger inter-model spread, and therefore increased ensemble uncertainty [63,64]. Introducing variables with high intrinsic uncertainties at the very beginning of our analytical pipeline would inject a significant amount of “noise” on the model, which would be subsequently propagated (and potentially amplified) through the ML downscaling, and all downstream calculations. The result of this uncertainty cascade would be final projections with substantially wider confidence intervals.
The scale with which solar radiation is represented in GCMs constitutes another limitation for the use of this parameter in our model: GCMs provide a single value for the average downwelling radiation within an entire grid cell, representing a spatial average over an area covering hundreds of square kilometers that usually encompass diverse landscapes such as open fields, forests, and built environments. Because of this complex physical context, the actual solar radiation load experienced by any animal is determined by micro-environmental and individual factors that are completely unresolved by the GCM, including cloud coverage and availability of shade from trees, barns, sheds, etc. Therefore, while the difference in radiation experienced by animals under direct sunlight vs. animals protected by shade is substantial [65], this difference is entirely invisible to the GCM. Additionally, factors such as an animal’s shade-seeking behavior, and even phenotypic traits (e.g. coat coloration) can further impact the actual radiative heat load experienced by cattle on the ground [65–68].
Although there are several alternative indices that incorporate other parameters in addition to temperature and humidity [36,69], their use for the development of predictive models is limited by the nature and availability of the data they incorporate, which often requires detailed field measurements, over extensive areas and periods of time, and involves the use of complex and expensive equipment [69,70]. These restrictions are particularly relevant for global-scale, long-term modeling, where the primary goal is to construct a model that is not only accurate but also robust, transparent, and minimally susceptible to the propagation of error. To this end, it is often necessary to rely on the principle of methodological parsimony (i.e. a model should be as simple as possible for its given purpose), aiming not to incorporate every conceivable variable that might have an influence, but rather to select the most causally significant and reliably modeled drivers of the phenomenon under investigation. In this context, a large body of literature has substantiated the relationship between THI and physiological indicators of heat stress, changes in productivity, and altered fertility in cattle [20,23,25,71], making it the de-facto standard for evaluating the interaction between livestock and its thermal environment [36].
As previously mentioned, another limitation of our work is the use of a static model for global cattle distribution for our projections. To the best of our knowledge, global long-term projections of cattle distribution are unfortunately not available at present, largely due to the complexities involved in identifying, quantifying, and modeling the many factors that can influence said distribution (e.g., the socioeconomic development paths of different countries, shifts in human population growth and distribution, consumer behavior, future decisions on land use, tariffs and trade agreements, changes in environmental protection laws, etc.) [3,72,73]. In light of this complexity, the generation of such a model is beyond the scope of our work. Nevertheless, considering the nature of the infrastructural, economic, and cultural aspects involved in livestock husbandry, present areas of historically continuous cattle farming are likely to remain relevant to the industry in the coming decades, and are therefore reliable general indicators of future cattle distribution. Therefore, this approach remains critical for identifying the most vulnerable regions and informing targeted mitigation and adaptation strategies in livestock management, while also highlighting the need for future research to explore adaptive responses under evolving cattle distributions.
Supporting information
S1 Table. List of NEX-GDDP-CMIP6 models used to generate hourly THI projections (from [27]).
https://doi.org/10.1371/journal.pclm.0000761.s001
(DOCX)
S1 Fig. Spatial distribution of the standard error for days with at least one hour above the THI threshold.
The two THI thresholds (68.8 and 84) employed in this study are displayed, with 68.8 on the left and 84 on the right columns, respectively. The top row illustrates the SSP2-45 scenario, while the bottom row presents the SSP5-85 scenario.
https://doi.org/10.1371/journal.pclm.0000761.s002
(PDF)
S2 Fig. Spatial distribution of the standard error for hours spend above the THI threshold.
The two THI thresholds (68.8 and 84) used in this study are show, in the left and right columns of panels, respectively. The top row illustrates the SSP2-45 scenario, while the bottom row presents the SSP5-85 scenario.
https://doi.org/10.1371/journal.pclm.0000761.s003
(PDF)
S3 Fig. Spatial distribution of the standard error for the Thermal Stress Load metric.
The two THI thresholds (68.8 and 84) are presented in the left and right columns of panels, respectively. The top row presents the SSP2-45 scenario, whereas the bottom row presents the SSP5-85 scenario.
https://doi.org/10.1371/journal.pclm.0000761.s004
(PDF)
S4 Fig. Standard error spatial distribution for the metric related to heat waves for the SSP2-45 scenario.
The top row of panels presents the SE for the number of occurrences (left) and number of event hours per year (right). In the bottom row of panels the SE for the maximum duration (left) and mean duration (right) are presented.
https://doi.org/10.1371/journal.pclm.0000761.s005
(PDF)
S5 Fig. Standard error spatial distribution for the metric related to heat waves for the SSP5-85 scenario.
The top row of panels presents the SE for the number of occurrences (left) and number of event hours per year (right). In the bottom row of panels the SE for the maximum duration (left) and mean duration (right) are presented.
https://doi.org/10.1371/journal.pclm.0000761.s006
(PDF)
S6 Fig. TSW parameters - changes from historic reference period to mid-century reference period.
A) TSW frequency. B) Number of TSW hours per year. C) Average TSW duration. D) Maximum TSW duration. For all panels, maps on the left side correspond to predictions under SSP2-4.5, and maps on the right side correspond to predictions under SSP5-8.5.
https://doi.org/10.1371/journal.pclm.0000761.s007
(PDF)
S7 Fig. Average number of TSW hours per year for each reference period and emissions scenario considered in our work.
https://doi.org/10.1371/journal.pclm.0000761.s008
(PDF)
Acknowledgments
The authors would like to thank the High Performance Computing Facility of the Cyprus Institute for their support in the computational and storage needs for this study.
References
- 1.
Key N, Sneeringer S, Marquardt D. Climate change, heat stress, and US dairy production. 175. 2014. https://doi.org/10.22004/ag.econ.186731
- 2. Cartwright SL, Schmied J, Karrow N, Mallard BA. Impact of heat stress on dairy cattle and selection strategies for thermotolerance: a review. Front Vet Sci. 2023;10:1198697. pmid:37408833
- 3. North MA, Franke JA, Ouweneel B, Trisos CH. Global risk of heat stress to cattle from climate change. Environ Res Lett. 2023;18(9):094027.
- 4. St-Pierre NR, Cobanov B, Schnitkey G. Economic losses from heat stress by US Livestock Industries. Journal of Dairy Science. 2003;86:E52–77.
- 5. Ferreira FC, Gennari RS, Dahl GE, De Vries A. Economic feasibility of cooling dry cows across the United States. J Dairy Sci. 2016;99(12):9931–41. pmid:27743663
- 6. Kadzere CT, Murphy MR, Silanikove N, Maltz E. Heat stress in lactating dairy cows: a review. Livestock Production Science. 2002;77(1):59–91.
- 7. Polsky L, von Keyserlingk MAG. Invited review: Effects of heat stress on dairy cattle welfare. J Dairy Sci. 2017;100(11):8645–57. pmid:28918147
- 8. Mishra SR. Behavioural, physiological, neuro-endocrine and molecular responses of cattle against heat stress: an updated review. Trop Anim Health Prod. 2021;53(3):400. pmid:34255188
- 9. Neculai-Valeanu A-S, Ariton A-M. Udder health monitoring for prevention of bovine mastitis and improvement of milk quality. Bioengineering (Basel). 2022;9(11):608. pmid:36354519
- 10. Whitfield L. Heat stress and its impact on fertility in dairy cattle. Livestock. 2016;21(4):218–21.
- 11. Burhans WS, Rossiter Burhans CA, Baumgard LH. Invited review: lethal heat stress: the putative pathophysiology of a deadly disorder in dairy cattle. J Dairy Sci. 2022;105(5):3716–35. pmid:35248387
- 12. Vitali A, Felici A, Esposito S, Bernabucci U, Bertocchi L, Maresca C, et al. The effect of heat waves on dairy cow mortality. J Dairy Sci. 2015;98(7):4572–9. pmid:25958287
- 13. Morignat E, Perrin J-B, Gay E, Vinard J-L, Calavas D, Hénaux V. Assessment of the impact of the 2003 and 2006 heat waves on cattle mortality in France. PLoS One. 2014;9(3):e93176. pmid:24667835
- 14. Hahn GL. Dynamic responses of cattle to thermal heat loads. J Anim Sci. 1999;77 Suppl 2:10–20. pmid:15526777
- 15.
Sejian V, Bhatta R, Soren NM, Malik PK, Ravindra JP, Prasad CS, et al. Introduction to concepts of climate change impact on livestock and its adaptation and mitigation. Climate change impact on livestock: adaptation and mitigation. Springer India; 2015. p. 1–23. https://doi.org/10.1007/978-81-322-2265-1_1
- 16. Cheng M, McCarl B, Fei C. Climate change and livestock production: a literature review. Atmosphere. 2022;13(1):140.
- 17. Bett B, Kiunga P, Gachohi J, Sindato C, Mbotha D, Robinson T, et al. Effects of climate change on the occurrence and distribution of livestock diseases. Prev Vet Med. 2017;137(Pt B):119–29. pmid:28040271
- 18. Godde CM, Mason-D’Croz D, Mayberry DE, Thornton PK, Herrero M. Impacts of climate change on the livestock food supply chain; a review of the evidence. Glob Food Sec. 2021;28:100488. pmid:33738188
- 19. Zhou M, Groot Koerkamp PWG, Huynh TTT, Aarnink AJA. Evaporative water loss from dairy cows in climate-controlled respiration chambers. J Dairy Sci. 2023;106(3):2035–43. pmid:36631318
- 20. Bohmanova J, Misztal I, Cole JB. Temperature-humidity indices as indicators of milk production losses due to heat stress. J Dairy Sci. 2007;90(4):1947–56. pmid:17369235
- 21. Pinto S, Hoffmann G, Ammon C, Amon T. Critical THI thresholds based on the physiological parameters of lactating dairy cows. J Therm Biol. 2020;88:102523. pmid:32125999
- 22. Campos IL, Chud TCS, Oliveira HR, Baes CF, Cánovas A, Schenkel FS. Using publicly available weather station data to investigate the effects of heat stress on milk production traits in Canadian Holstein cattle. Can J Anim Sci. 2022;102(2):368–81.
- 23. Brügemann K, Gernand E, König von Borstel U, König S. Defining and evaluating heat stress thresholds in different dairy cow production systems. Arch Anim Breed. 2012;55(1):13–24.
- 24. Igono MO, Bjotvedt G, Sanford-Crane HT. Environmental profile and critical temperature effects on milk production of Holstein cows in desert climate. Int J Biometeorol. 1992;36(2):77–87. pmid:1634283
- 25. Bouraoui R, Lahmar M, Majdoub A, Djemali M, Belyea R. The relationship of temperature-humidity index with milk production of dairy cows in a Mediterranean climate. Anim Res. 2002;51(6):479–91.
- 26. Vitali A, Segnalini M, Bertocchi L, Bernabucci U, Nardone A, Lacetera N. Seasonal pattern of mortality and relationships between mortality and temperature-humidity index in dairy cows. J Dairy Sci. 2009;92(8):3781–90. pmid:19620660
- 27. Georgiades P, Economou T, Proestos Y, Araya J, Lelieveld J, Neira M. Global projections of heat stress at high temporal resolution using machine learning. Earth Syst Sci Data. 2025;17(3):1153–71.
- 28. Hersbach H, Bell B, Berrisford P, Hirahara S, Horányi A, Muñoz-Sabater J, et al. The ERA5 global reanalysis. Quart J Royal Meteoro Soc. 2020;146(730):1999–2049.
- 29.
National Research Council. A guide to environmental research on animals. Washington DC, USA: The National Academy Press; 1971.
- 30. Thrasher B, Wang W, Michaelis A, Melton F, Lee T, Nemani R. NASA global daily downscaled projections, CMIP6. Sci Data. 2022;9(1):262. pmid:35654862
- 31. O’Neill BC, Kriegler E, Riahi K, Ebi KL, Hallegatte S, Carter TR, et al. A new scenario framework for climate change research: the concept of shared socioeconomic pathways. Climatic Change. 2013;122(3):387–400.
- 32.
Wilks DS. Introduction. International Geophysics. Elsevier. 2011. p. 3–5.
- 33.
Georgiades P. Temperature humidity index gddp-nex-cmip6 ml projections. 2024. https://doi.org/10.26050/WDCC/THI
- 34. DiCiccio TJ, Efron B. Bootstrap confidence intervals. Statist Sci. 1996;11(3).
- 35.
Livestock Conservation Institute. Patterns of transit losses. Omaha, NE, USA: Livestock Conservation Institute; 1970.
- 36.
Hahn G, Gaugthan J, Mader T, Eigenberg R. Thermal indices and their applications for livestock environments. In: DeShazer JA, editor. Livestock energetics and thermal environmental management. St. Joseph, MI, USA: American Society of Agricultural and Biological Engineers; 2009.
- 37. Arias R, Mader T. Determination of potential risk of heat stress of cattle in four locations of Central and Southern Chile. Arch med vet. 2010;42(1).
- 38. Hubbard KG, Stooksbury DE, Hahn GL, Mader TL. A climatological perspective on feedlot cattle performance and mortality related to the temperature-humidity index. Journal of Production Agricult. 1999;12(4):650–3.
- 39.
Gilbert M, Nicolas G, Cinardi G, Van Boeckel TP, Vanwambeke SO, Wint G. GLW 4: Gridded Livestock Density (Global - 2020 - 10 km). Food and Agriculture Organization of the United Nations (FAO). 2024. https://data.apps.fao.org/catalog//iso/9d1e149b-d63f-4213-978b-317a8eb42d02.
- 40. Gilbert M, Nicolas G, Cinardi G, Van Boeckel TP, Vanwambeke SO, Wint GRW, et al. Global distribution data for cattle, buffaloes, horses, sheep, goats, pigs, chickens and ducks in 2010. Sci Data. 2018;5:180227. pmid:30375994
- 41.
World Meteorological Organization. Guidelines on the definition and characterization of extreme weather and climate events. Geneva: World Meteorological Organization; 2023.
- 42. Nawaro J, Gianquintieri L, Pagliosa A, Sechi GM, Caiani EG. Heatwave definition and impact on cardiovascular health: a systematic review. Public Health Rev. 2023;44:1606266. pmid:37908198
- 43. Wedler M, Pinto JG, Hochman A. More frequent, persistent, and deadly heat waves in the 21st century over the Eastern Mediterranean. Sci Total Environ. 2023;870:161883. pmid:36736407
- 44.
I P C C. Annex VII: Glossary. Climate Change 2021 - The Physical Science Basis: Working Group I Contribution to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press; 2021. p. 2215–56.
- 45.
Hahn GL, Mader TL. Heat waves in relation to thermoregulation, feeding behavior and mortality of feedlot cattle. In: Proceedings of the Fifth International Livestock Environment Symposium. 1997. p. 563–71.
- 46. Carvajal MA, Alaniz AJ, Gutiérrez-Gómez C, Vergara PM, Sejian V, Bozinovic F. Increasing importance of heat stress for cattle farming under future global climate scenarios. Science of The Total Environment. 2021;801:149661.
- 47. Rahimi J, Mutua JY, Notenbaert AMO, Dieng D, Butterbach-Bahl K. Will dairy cattle production in West Africa be challenged by heat stress in the future?. Climatic Change. 2020;161(4):665–85.
- 48. Rahimi J, Mutua JY, Notenbaert AMO, Marshall K, Butterbach-Bahl K. Heat stress will detrimentally impact future livestock production in East Africa. Nat Food. 2021;2(2):88–96. pmid:37117410
- 49. Suarez-Gutierrez L, Müller WA, Li C, Marotzke J. Hotspots of extreme heat under global warming. Clim Dyn. 2020;55(3–4):429–47.
- 50. Garry FK, Bernie DJ, Davie JCS, Pope ECD. Future climate risk to UK agriculture from compound events. Climate Risk Management. 2021;32:100282.
- 51. Grossi G, Goglio P, Vitali A, Williams AG. Livestock and climate change: impact of livestock on climate and mitigation strategies. Anim Front. 2018;9(1):69–76. pmid:32071797
- 52. Herrero M, Henderson B, Havlík P, Thornton PK, Conant RT, Smith P, et al. Greenhouse gas mitigation potentials in the livestock sector. Nature Clim Change. 2016;6(5):452–61.
- 53. Bishop-Williams KE, Berke O, Pearl DL, Hand K, Kelton DF. Heat stress related dairy cow mortality during heat waves and control periods in rural Southern Ontario from 2010 -2012. BMC Vet Res. 2015;11:291. pmid:26610334
- 54. Díaz RF, Galina CS, Aranda EM, Aceves LA, Sánchez JG, Pablos JL. Effect of temperature - humidity index on the onset of post- partum ovarian activity and reproductive behavior in Bos indicus cows. Anim Reprod. 2020;17(1):e20190074. pmid:32399065
- 55. Díaz RF, Galina CS, Estrada S, Montiel F, Castillo G, Romero-Zúñiga JJ. Variations in the temperature-humidity index and dorsal fat thickness during the last trimester of gestation and early postpartum period affect fertility of bos indicus cows in the tropics. Vet Med Int. 2018;2018:2360430. pmid:30140424
- 56. Carabaño MJ, Díaz C, Ramón M. Assessing heat tolerance through productive vs physiological indicators. Data from dairy sheep under on-farm conditions. Animal. 2022;16(11):100662. pmid:36327798
- 57. Thornton P, Nelson G, Mayberry D, Herrero M. Increases in extreme heat stress in domesticated livestock species during the twenty-first century. Glob Chang Biol. 2021;27(22):5762–72. pmid:34410027
- 58. Hempel S, Menz C, Pinto S, Galán E, Janke D, Estellés F, et al. Heat stress risk in European dairy cattle husbandry under different climate change scenarios – uncertainties and potential impacts. Earth Syst Dynam. 2019;10(4):859–84.
- 59. Renaudeau D, Collin A, Yahav S, de Basilio V, Gourdine JL, Collier RJ. Adaptation to hot climate and strategies to alleviate heat stress in livestock production. Animal. 2012;6(5):707–28. pmid:22558920
- 60. Galán E, Llonch P, Villagrá A, Levit H, Pinto S, Del Prado A. A systematic review of non-productivity-related animal-based indicators of heat stress resilience in dairy cattle. PLoS One. 2018;13(11):e0206520. pmid:30383843
- 61. Misztal I, Brito LF, Lourenco D. Breeding for improved heat tolerance in dairy cattle: methods, challenges, and progress. JDS Commun. 2024;6(3):464–8. pmid:40458150
- 62.
I P C C. Framing, context and methods. Climate change 2021 - The physical science basis: Working Group I contribution to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press; 2023.
- 63. Stevens B, Bony S. Climate change. what are climate models missing?. Science. 2013;340(6136):1053–4. pmid:23723223
- 64. Iacono MJ, Delamere JS, Mlawer EJ, Shephard MW, Clough SA, Collins WD. Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J Geophys Res. 2008;113(D13).
- 65. Schütz KE, Rogers AR, Poulouin YA, Cox NR, Tucker CB. The amount of shade influences the behavior and physiology of dairy cattle. J Dairy Sci. 2010;93(1):125–33. pmid:20059911
- 66. Tucker CB, Rogers AR, Schütz KE. Effect of solar radiation on dairy cattle behaviour, use of shade and body temperature in a pasture-based system. Applied Animal Behaviour Science. 2008;109(2–4):141–54.
- 67. Blackshaw J, Blackshaw A. Heat stress in cattle and the effect of shade on production and behaviour: a review. Aust J Exp Agric. 1994;34(2):285.
- 68. Finch VA, Bennett IL, Holmes CR. Coat colour in cattle: effect on thermal balance, behaviour and growth, and relationship with coat type. J Agric Sci. 1984;102(1):141–7.
- 69. Wang X, Bjerg BS, Choi CY, Zong C, Zhang G. A review and quantitative assessment of cattle-related thermal indices. J Therm Biol. 2018;77:24–37. pmid:30196896
- 70. Zanetoni HHR, Tinoco IFF, Barbari M, Conti L, Rossi G, Baeta FC. Alternative form to obtain the black globe temperature from environmental variables. Agronomy Research. 2019;17:900–6.
- 71. Ravagnolo O, Misztal I, Hoogenboom G. Genetic component of heat stress in dairy cattle, development of heat index function. J Dairy Sci. 2000;83(9):2120–5. pmid:11003246
- 72. Kummu M, Heino M, Taka M, Varis O, Viviroli D. Climate change risks pushing one-third of global food production outside the safe climatic space. One Earth. 2021;4(5):720–9. pmid:34056573
- 73. Weindl I, Popp A, Bodirsky BL, Rolinski S, Lotze-Campen H, Biewald A, et al. Livestock and human use of land: Productivity trends and dietary choices as drivers of future land and carbon dynamics. Global and Planetary Change. 2017;159:1–10.