Weather data and the supply chain

The predictive power of weather info, as illustrated by cows and La Niña.

My introductory column focused on the broad theme of combining global weather data, sensor networks, and advanced forecasting techniques toward the early identification of weather and climate related hazards.  As discussed, early warning of potential weather-triggered problems will not prevent a physical hazard from occurring, but advance warning, even if only provided with a short lead time, may allow for some mitigation or avoidance measures to be employed ahead of an impact event.

Identifying extreme events is just one area where it’s useful to apply long-range weather intelligence. In addition to the employment of various techniques to avoid human suffering, there are also numerous commercial interests who can apply this information as well. Just as a property reinsurer might want to know how either an acute or a seasonal weather event might affect premiums, growers may also want to assess crop potential, retailers might want to project seasonal product demand, and traders may want to employ a pricing strategy.

Food, energy, and weather

There are two things that each and every one of the roughly 7 billion people on the planet need: food and energy. Weather is central to both.

On the food side, advanced weather outlooks can help assess crop potential, both in terms of actual production/yield, and potential for disease pressure. As noted in my previous post, growers who anticipated crop losses from excessive heat or a lack of moisture (or both) may have purchased crop protection insurance, while manufacturing companies may have secured forward prices or managed their exposures through a hedge.

But even at a more basic level, there is a lot of environmental data available at the public’s disposal. If structured and viewed in the right way, that data can provide insights on the supply side of many basic commodities. Assessing food production potential for basic necessities is what I like to call the first true step in understanding the global agricultural supply chain.

Now, there’s no such thing as a perfect forecast. Even with forward-looking metrics that can be deemed 100-percent accurate, there are variables that go into determining, for instance, how many bushels per acre a particular region’s wheat crop will yield. Forecasting is part art, part science. From the science perspective, the more clean, reliable data that we can obtain and plug into a model, the better we may get at determining production potential, or more importantly, highlighting areas that may be susceptible to a weather risk.

How La Niña affects milk production

Here’s a simple example of the climate/food relationship: Casual weather observers are probably familiar with the El Niño Southern Oscillation cycle (referred to as ENSO). This particular phase of this large-scale physical weather driver often governs the global pattern. While an El Niño dictated much of the US pattern in 2009 (remember snow being trucked in for certain Olympic events in Vancouver?), the opposite La Niña has developed in 2010, which brings its’ own set of variables. The current La Niña is one reason that our January-February outlook at Weather Trends is for a little cooler than last year for the western US. We also can use the ENSO index to assess yield potential for a commercially important commodity: milk.

For some time, we have known about the positive relationship between La Niña-like conditions and milk production for US dairy producing herds.  During an El Niño, conditions for much of the US dairy production regions may tend to be warmer and wetter, and as dairy cows exhibit sensitivity to heat stress and/or muddy fields, these conditions correlate with decreased milk production, particularly during the key months (March-July) in the annual cycle.  As the opposite La Niña pattern has been developing for the last several months, cooler temperatures have been present across much of California (the largest producing state), limiting heat stress and contributing to an active grazing season, both of which are good for milk yields.  This particular La Niña is actually shaping up to be a pretty strong event, and as a result, we have been expecting better US production numbers to follow.  Note that this does not take into account decreased herd size, so the emphasis is on milk yield per cow.

Picture 1.png

To test this idea, we can look closer at the relationship between the Southern Oscillation Index (SOI), which is an ENSO guide, and US milk production over the last decade.  Specifically, we can highlight periods where there has been a stronger trend toward positive-phase SOI in recent months relative to the 6 month moving average; the assumption being a stronger relative acceleration toward positive phase supports better milk production weather. 

Using a simple decision-tree scheme, the time series was split by grouping all months where the more recent period showed stronger positive SOI characteristics (as defined by a quantitative index).  Of this reduced group of months, we then looked at monthly normalized US milk year-over-year (y/y) production to see if stronger numbers may have been related to the index.  Therefore, using a sample size (n) of 60 months, y/y milk yields increased in 52 of these months (87 percent), verifying that a positive correlation exists.

 

Picture 2.png

This is not to suggest that the SOI, or any other weather parameter, is the primary factor in assessing potential milk production.  Remember, a forecast is a blend of art and science, so some subjectivity is involved. But this simple analysis does demonstrate that weather can be a key driver in the amount of milk that is flowing from producers to consumers, and it bears watching as a signal for forward pricing, and assessment of global stocks.

tags: , , , ,
  • Alex Tolley

    It would have been better to explain why you massaged the data as you did to get a result. Also showing a chart of the transformed data would have been helpful too. It is hard to understand why:
    “assumption being a stronger relative acceleration toward positive phase supports better milk production weather.” should be expected, a priori, rather than some other relationship.

  • Michael Ferrari

    @Alex: data not massaged – just put in this graphic to illustrate the broader generalizations that can be made. The example highlights the starting point towards better understanding the relationship between the physiology of dairy cows and weather patterns associated with heat stress/dehydration (& yield). We then build granular models that relate physical scale microclimate processes to the associated biological response, quantifying the relationship.

  • http://www.i-sitekohchang.com/koh-chang-weather/forecast.html jekko

    Michael, in your post you say that there is no such thing as a perfect forecast but i disagree. A forecast is by definition is just that, a forecast. It will change as the data sets changes. It is not a prediction of what will happen but a good indicator to plan for the future.