Evaluating and Assessing Weather Issues and Vulnerabilities
At the time of writing this, parts of the midwest (most notably Wyoming) are suffering major agricultural/economic impacts due to suffering the worst drought in ten years.
Excuse us? The worst drought in ten years? The worst drought in 100 years – they could be excused for being tripped up by that. Many of us would even allow a person the benefit of the doubt for the worst drought in 50 years. But to be blindsided by something that happens once a decade? That’s imprudent planning and greed on the part of the farmers, who simply gambled that they’d have a good year this year, and so over-extended their water consumption, and now have had their calculated gamble, based on greed, turn around and bite them.
It is hard to feel too sympathetic for such people. But at least they may qualify for various assistance programs, and if they have to end up selling off livestock, they have people to sell the livestock to (potentially located thousands of miles away), and will be able to buy new livestock next year (and potentially source them from a considerable distance). The current infrastructure of the country, its economy, and even its social support mechanisms all act to minimize the still unfortunate impacts on these farmers at present.
But in a Level 2/3 situation, and with any farmer’s market being reduced to a very local region, there’d not be such broader resources to rely upon. Quite the opposite – the market for cattle would become massively depressed, and next year, there’d be precious little in the way of breeding stock to make up the shortfall. This begs the question – how much weather risk is it prudent to accept – not so much in the present day situation, but with an eye to a less forgiving future scenario?
The answer is obviously that we can’t accept any risk that would threaten our ongoing viability. So how much weather risk is too much risk? This is very hard to establish, in large part because weather isn’t a ‘constant variable’ (a concept that sounds like an oxymoron to start with). Let us explain.
Short Term Randomness in Weather
There are many factors impacting on the weather we experience each day. For our purposes, most of these factors can be perceived as semi-random in the medium term. For sure, in the very short-term, we can make a reasonable guess about tomorrow’s weather based on what we know of the weather today, what the barometer tells us, the direction of the winds, and how we read the sky. We can make a somewhat less accurate guess about the day after tomorrow’s weather, too. But the degree of accuracy continues to erode with each extra day into the future we look.
Will it rain on Tuesday in five weeks time? Maybe, maybe not. Will next year’s growing season be shorter or longer than normal? And so on. Apart from making a statistical guess based on past rainfall patterns, it is hard for us to otherwise give an accurate prediction based on any factual modeling of how the weather will act between now and then (although weather forecasting services, with super-computers and data inputs from weather stations all around the world make attempts to answer these types of questions, and with varying and not always impressive results). So, for this purpose, the weather becomes more or less random, within the constrains of certain probabilities.
Another way to think of this is like rolling a dice. You are playing a game where the rules are that for the first 15 minutes of each hour, you’ll win if the dice shows a 1 or 2. For the second 15 minutes, you’ll win if the dice shows a 1, 2 or 3. For the next 15 minutes, you’ll only win if the dice shows a 6. And for the last 15 minutes, you’ll win if the dice comes up 1,2,3,4 or 5.
So depending on the time of each hour, while you still have a totally random chance of winning, your overall chance of winning changes from very favorable to very unfavorable. You can think of, eg, rainfall in a similar manner. While there’s no guarantee about levels of rainfall on any day, the chance of rainfall goes more or less predictably up and down depending on the season.
This randomness applies in the ‘short-term’ which in this context we consider to span periods of five to ten years, more or less.
Longer Term Cyclical Variations in Weather
Now that you understand the random nature of weather on a short-term basis, we can now move on to considering other factors that also impact on weather, but in the long-term. There are various cycles that see regions go through periods of predominantly ‘good’ and ‘bad’ weather, cycles that can last for ten, twenty and even more years. Indeed the length of these cycles from the start of one complete cycle to the start of the next can extend out as long as 50 – 75 years or so.
Here’s an interesting web page with a fairly bewildering array of charts and graphs, but if you scroll much of the way down (and you don’t really need to read or understand everything that is being presented), you’ll see an interesting graph headed ‘The PDO + AMO cycles are not in phase:’ and immediately below that, four fascinating maps of the US showing drought conditions over the course of the cycles, and right at the very bottom, a 500 year time series based on tree growth and clearly showing cyclical variations.
These longer cycles are creating part of the uncertainty about alleged global warming. Was the observed warming trend of a couple of decades ago the result of manmade activities, or was it a normal cyclical thing? And is the lack of global warming for the last 15 years also cyclical, or is it significant?
These graphics give a very good indication of how weather is not only random in the short-term, but also follows longer term cycles.
These long cycles can create a major trap for people trying to understand what type of weather to expect in the future. If people only sample years that are at one part of the overall cycle, they get an erroneous impression of the future. What should be obvious and predictable based on a long enough historical time view so that you can clearly see the cyclical nature of the weather variations, instead is perceived to be unusual, unexpected, and exceptional, even though in truth it is none of those things.
Clearly, in understanding the weather, you need to understand not only the short-term semi-random variations but the overall longer term cyclical impacts on the range of short-term variations.
Working With Long Term Data to Set Acceptable Risks
We said above to steer well clear of even a conservative 100 year flood plain – the good news part of that is that someone else has already calculated what the 100 year flood plain area is likely to be. But for some other measures of possible weather extremes, you’ll probably have to do your own figuring.
Clearly you need to be conservative in assessing the acceptable level of risk for weather extremes like droughts, because if you guess wrong, you could be endangering the livelihood and survival of you and your entire community.
There’s no sense to setting yourself a ‘less than once every hundred years’ target for flood avoidance, but accepting a ‘once every ten years or so’ level for drought avoidance.
Our point, however, in this context is that in order to make such decisions, you must have more than five or ten years of historical data. There are statistical techniques that can analyze shorter periods of data to project the probability of what longer data series would reveal, and maybe you need to get an expert to do such things for you (we offer these types of services ourselves), but this analysis can not factor in the presence of cyclical impacts.
The best thing to do is not guess, but instead to use as many years of weather data as possible.
But be careful in doing so. As the whole climate change controversy indicates, weather data is subject to interpretation. Maybe, over an extended period of many decades, the weather station location was changed. Or maybe other things changed around the weather station – maybe it went from being in the middle of nowhere to now being in the middle of a medium-sized city. Maybe it changed its sampling methodology so that the same weather now results in different numbers being recorded, compared to some years previously.
And, whether due to man-made causes, the influence of the sun, regular cycling, or just random variations over time, most people will accept that climate conditions have changed over the last 50 years or more – there was a steady period of increase, and then – at least through 2011 – temperatures started trending down somewhat again.
So if you can’t get a full 100 years of weather data, that’s perhaps not a great loss; and indeed, if you did, you probably should attach less statistical weight to old data compared to new data.
See also our analysis in our water storage calculations article about combining the worst year’s results, month by month, compared to the worst year’s results for multiple months in a row, as a way of further stressing your projections.
So, while the historical numbers might seem very exact and certain, interpreting them for probable and worst case future outcomes becomes a very subjective undertaking. However you do it, you’re sure to arrive at better final numbers if you have more raw data to start with.
To summarize all of the above, don’t, for example, base your rainfall expectation for a location on only the last two or five or ten years of data. Maybe the region was going through a very rainy cycle, and maybe, as soon as you move there, it will flip over to a very dry decade, making the land you thought to be fertile and well watered suddenly become dry, arid, and expensive/impossible to work.
The more data you have, the more informed your decision will become, and the less risk you’ll be confronting in the future.