Speaker
Description
Key words: climate change; heat stress; milk yield; adaptation; cooling technology.
Introduction:
Climatic variability and change are posing serious threats to global food security (FAO, 2015). This has motivated a rapidly growing body of research examining how climate influences economic outcomes. This “climate-economy” literature uses panel methods and high-frequency variation in weather to estimate the causal effects of changes in climate on outcomes of interest, notably agricultural output. While methodological advancements and greater data have improved the understanding of how extreme weather affects economic production, knowledge gaps persist about under-studied sectors and the potential for adaptation.
The dairy sector is an important segment of the agricultural sectors of many countries, including Israel, and its importance is expected to increase as demand for fresh and processed dairy products continues to increase. Heat stress is one of the main threats on milk production. As overall temperatures and the incidence of extreme heat continue to increase, the dairy sector is likely to face increasing stress and milk production may suffer significant declines.
The daily resolution of milk production data, as opposed to the yearly resolution of crop data, provides a useful setting in which common assumptions made in the literature can receive closer scrutiny, and new insights potentially derived.
Objective:
The objective of this research is to estimate the contemporaneous and delayed impacts of humid heat on milk yield. In addition to contributing new evidence on the impacts of humid heat on milk production, our work makes methodological contributions to the literature. The existing empirical literature on the impacts of climate change on agriculture has addressed counts of extreme heat events, but not how these are spread across time, e.g., the potential of a non-linear cumulative effect during hot spells, or how they interact with high levels of humidity. In addition, we conduct novel estimations of the effectiveness of adaptation strategies, and discuss how our estimates can be used to project the impact of climate change on milk production.
Method:
We use daily data on the milk production of 130,000 cows over 12 years in Israel. The data include the total daily amount of milk produced by each cow, as well as the start date of the given lactation cycle, the number of calvings - which is a reliable proxy for the cow’s age - and the number of milkings per day. We construct two separate weather datasets of hourly temperature and relative humidity at the location of each farm, for the full period 2009-2020. In 2020-21, we attempted to survey all dairy farms in Israel in order to collect information about their basic operational characteristics, and in particular, the type of cooling technologies, if any, that they use to deal with heat stress, when they had installed them, and why. We also asked about additional adaptation strategies.
Our primary empirical model relates milk production by a given cow i in a given day t to same-day heat stress in the farm, represented as a vector of weather attributes Hit:
yit ≡ log(milkit) = G(Hit) + X’itδ + αi + ωt + ϵit (1)
where X is a vector of cow time-varying attributes, αi are cow fixed effects, and ωt are time fixed effects.
As a proxy of heat strain, we choose the wet-bulb temperature (Twb). We conduct robustness tests in which we replace Twb with THI indices.
The distribution of H in the course of the day is represented by a single summary indicator Twbit, such as its mean, maximum or minimum value. The function G(·) remains completely flexible. We approximate it with the non-parametric form:
G(Twbit) = Σh βhI(Twbit = h) (2)
where I(·) are binary indicators of whether the summary Twb of that day falls within a certain interval of values of h, and the summation occurs over interval bins encompassing all such possible values. The parameters βh represent the impact on milk production of each bin of daily values of Twb. We also estimate a model which takes into account the full distribution of Twb in the day of observation, rather than a single value.
Our survey data provides us with information about when and in which farm various cooling technologies were installed. This allows us to estimate the impacts of these technologies on the response of milk yield to heat stress. Specifically, we estimate a regression:
yijt ≡ log(milkijt) = G0(Hjt) + Cjt × G1(Hjt) + X’itδ + αi + ωt + ϵijt (3)
where i is a cow index, j is a farm index, and the variable Cjt is a binary indicator of whether farm j had a cooling technology in place in date t. In this specification, G0(·) is the estimated response of milk to heat stress in the absence of any cooling systems, and G1(·) is the change in the response associated with having a cooling system installed. Note that the parameters of G1(·) cannot be causally interpreted, since variation in Cjt may be endogenous. Nevertheless, the observed correlations constitute a first step towards the assessment of the performance of these technologies.
Findings:
Humid heat exerts highly nonlinear negative effects reaching up to a 10% decrease in milk production on extreme days, and effects persist for up to 10 days after direct exposure. Moreover, the adoption of simple cooling technologies, shifts in birth timing or changes in feed practices may be able to reduce less than half of the impacts of extreme heat exposure.
Conclusion:
Given the technological advancement, long-standing exposure to heat, and climatic diversity of the Israeli dairy system, the results suggest that common adaptation strategies may hold limited potential to avert the impacts of climate change in this nutritionally and economically important sector. More research is needed to quantify the actual performance and cost effectiveness of a broader range of adaptation approaches.
