Many people love to hate the weather forecast, or the weather forecaster! There’s no end to the jokes about meteorologists. For instance, I’m sure you’ve all heard this one: “In what other occupation can you be wrong more than half of the time and still keep your job?”
But as the skies open up and ruin your plans despite the weatherman’s promise of sun, who else are you going to blame? In a New York Times article entitled The weatherman is not a moron, author Nate Silver likens the weather forecaster to a baseball umpire who rarely gets credit for making the right call.
Most people take forecasts for granted. When they’re correct, folks don’t notice but when the forecast is wrong, everyone comments. Instead of putting all your faith in the weatherman, there are some great tools available that will enable you to go well beyond standard public forecasts.
But first, let me give you some background on forecasting. Nowadays, weather forecasts are mainly based on massive computerized numerical weather prediction models. These models draw upon vast datasets and run complex simulations using some of the world’s most powerful supercomputers.
A better prediction can often be achieved by increasing the number of opinions. There are numerous weather prediction models throughout the world and no one model will always outperform in all situations. For this reason, forecasts are based on several models.
For example, The Weather Company’s forecast considers 162 different model results. The relative weighting of each of these opinions is based on algorithms that compare how closely each forecast matches what was recorded at each weather station. As certain models perform better at certain times, their relative weighting increases.
The Weather Company, and its parent company, IBM, have an arsenal of supercomputers on hand to parse through billions of forecasts per day. Most people don’t have those sorts of resources available, but if you’re looking to dig deeper into the various forecast models, check out SpotWX (https://spotwx.com), for a range of weather forecasts for any location or “spot” you choose.
The main feature of SpotWX is that when users select a location on the map, they can view the output from 10 to 12 different numerical weather prediction models, depending on what’s available for that specific location. Viewing different outputs is much like asking for a second or third informed opinion on what is likely to happen. If there’s a consensus of opinions, the outcome is more probable — if the forecasts are all different, there’s more uncertainty.
The three examples on page 29 show forecasts from two Canadian models (HRDPS and RDPS) and one American model (NAM). In this example, the noon temperature and humidity forecasts from the three models show a temperature difference of 0.8° C and a relative humidity difference of six per cent, so there’s reasonable agreement. For important decisions that rely on a weather forecast, consulting more than one opinion is an excellent strategy.
Another helpful insight from SpotWX is information about model resolution. Every model outputs a virtual grid across the region that it covers with a forecast calculated for each grid cell. For example, the HRDPS produces a fine grid of 2.5 km over parts of North America, the RDPS is at 10 km over all of Canada and much of the United States, and the NAM is at 12 km over most of North America. Some models, such as the GDPS, run at 25 km resolution globally.
While a finer resolution is not necessarily more accurate, it will produce a forecast with more detail within a local area. This is particularly useful in areas with varied elevation, terrain, or landuse. For example, the 25 km grid of the GDPS will produce a somewhat regional or averaged forecast for an area whereas the 2.5 km HRDPS will consider local variability and be more likely to produce a farm- or field-specific forecast.
After listening to the weather forecast, it usually begs the question: “How sure are you?” Some weather events are harder to predict than others such as where and when a thunderstorm will occur or the exact path of a frontal system and whether or not these weather events will hit or miss your farm.
The question of forecast certainty makes this next tool quite handy. Climendo (http://climendo.com) is a website and app that tracks and compares forecasters’ accuracy in order to identify the statistically best forecast for a given location. While SpotWX provides the user with output from several numerical weather prediction models, Climendo shows actual forecasts from various providers.
Along with providing five different forecast opinions, Climendo gives a confidence rating from certain to fairly certain to fairly uncertain. This probability is based on the level of agreement between forecasts. Disagreeing forecasts will result in higher level of uncertainty. When an important weather-dependent decision needs to be made, you can factor in the level of certainty associated with the forecast.
Ultimately, everyone wants the most accurate forecast that they can get. An always-right forecast would, of course, be the best solution, but with weather, there will always be errors and uncertainties. While forecasts are improving and technology is getting better at dealing with this uncertainty, people who can dig deeper into the forecast for their area can make more informed weather-related decisions.
Armed with these tools, you can understand more of what goes into the forecast and perhaps question the weatherman when the forecast does not look right. Every farmer knows the high costs of unforeseen weather, so better insight into future weather can be extremely valuable and worth the investment. FF