Blog

Demolish Marquette to Improve Forecasts?

Would forecasts improve if we removed all the buildings, ripped out all the concrete and moved everyone to tents situated along Lakeshore Blvd? Short answer: likely.

Too often the forecasted high temperatures for Marquette run well below the measured high temperatures, especially on sunny days when the wind isn’t blowing off the lake. This is true of all forecast agencies I’ve observed to one degree or another. It just so happens the local National Weather Service, our forecast provider, is also guilty of this trend.

[UPDATE 7/4] Today is the 6th consecutive day of under forecasted highs (by greater than 3.0 degrees Fahrenheit). See our forecast accuracy page for more on that (forecasted highs are logged at 9:03 AM daily).

Typically, the argument that is used against relying on private weather stations (like this one) for forecasting or other mission-critical purposes is that they are too often incorrectly sited (placed), are less accurate than professional equipment, and are not maintained to accepted standards.  All of that may be generally true.

However, I’ve gone to considerable lengths to be sure that the instrumentation used for this website are at least as accurate as most official weather stations. Please refer to our about page for further details on this. You can also contact me and I will be happy to tell you more or take you for a tour.

Our measurement location is a city lot between two houses (25-50ft away) that also contains a garage and shed (20ft away). The temperature sensor is upwind from the nearest structures which are to the northeast and southeast (prevailing winds are from the SW, W, NW, N). We are less than 6 blocks from the lake — not too close to be completely dominated by it yet close enough to feel its general effects on the city.  In all, this is a representative location for the city with good exposure to the prevailing winds. Our equipment can sense the weather where people live.

The problem with dismissing all weather stations that aren’t sited according to the official specifications (i.e. at least 100ft away from cement, twice the horizontal distance of the height of the nearest object, etc) is that there are few, if any, such places in the city core. Even if there were and we were fortunate enough to have someone pay to install a weather station there, it wouldn’t measure the conditions most people experience as they walk, shop, work, or relax outdoors in the city. Perhaps you’ve noticed we don’t have an airfield in downtown Marquette?

A basic rule of meteorology is that when measurements and forecasts are in conflict you trust measurements. What happens if you routinely ignore or dismiss the current conditions? Well, you’re likely to see forecasts like the ones this weekend — on 3 successive days — that talk about “steady” or “near steady” temperatures 5-10+ degrees below current temperatures when the forecast was generated. This is despite the fact, in all instances, that temperatures had been clearly rising and hadn’t leveled off. So the forecast only became more wrong as time went on. That is sometimes the outcome of letting the satellites or weather models dictate what’s “happening”.

Primary “aspirated” station in background, backup station in foreground

There are 4 weather stations within the city limits of Marquette on a network (CWOP) that feeds directly into the NWS computer system. All 4 can be found on our area page under the heading “In Town”. One of them, run by the Coast Guard, is right on the lakeshore in a specific climactic zone that is heavily influenced by Lake Superior. Relatively few people live within the thin strip of land to which its measurements apply. The other three are private stations that are usually within a few degrees of each other. They are fair representations of east, central and south Marquette, respectively. In fact, all three (including ours) are the highly regarded Davis Vantage Pro2 weather station. There is absolutely no reason to assume all 3 of those stations are unreliable, especially given their agreement most of the time. In fact, one of those stations is actually run by an NWS employee! Dirty little secret: the local NWS office runs a VP2 weather station too. While it isn’t used for official measurements, its readings are available on their website. So one would logically assume they believe it to be reasonably accurate.

Twenty five years ago I was working as a news director for a radio station in Detroit. Our forecasts came off the Associated Press wire from, you guessed it, the National Weather Service. To be sure, the tools forecasters possessed then were quite primitive compared to what’s available even to non-forecasters now. Given what they had, these guys were legendary. Of course, they made mistakes. Everyone does. Forecasting is, after all, at its essence, a prediction of the future. Predictions are bound to fail at times because we live in a non-linear world with unpredictable dynamics. Even so, I can only recall a handful of times where I refused to give a forecast as written because it was counterfactual the moment it was released based on information available by looking out the window or reading a thermometer. Yes, Detroit is not Marquette. But the principle applies. Don’t state things that every member of the public with two eyes and skin knows is false and keep stating it for hours. That applies from the North Pole to the South Pole.

Part of the problem is the climatology for the city, I suspect. The official station for the city (MRQM4) is so close to the lake (it’s 200 ft from the Coast Guard weather station), it runs cooler in the spring and summer than the city as a whole. So if you feed the last 30 years of its high temperatures into the computer as a basis for the weather in Marquette, you’re going to get forecast guidance with highs that are too low for several months of the year (and perhaps too high in the fall too?).

Another possible culprit is that any collection of buildings and concrete will unnaturally heat temperatures at ground level (2 meters). Hence, why the siting guidelines for weather stations advise staying clear of them. I’m referring to the “heat island” effect. I wonder if this is fully accounted for in the weather models (software used as a guide for forecasting) or only factored into areas where a certain population threshold is met.

As I initially stated at the beginning of this article, the underestimates tend to occur on sunny days with winds away from the lake. These errors do occur, to a lesser extent, in the winter as well. Of course, there are further refinements you can make to the heuristic through careful observation, but that’s the main thrust. With that in mind, perhaps on such days there should be a “fudge factor” applied by adding a few degrees.

Like it or not, the reported temperatures in town are, on the whole, the actual temperatures people feel. It is unwise to dismiss them on a technicality that would require the city to become an open field or be confined to an area within a block of the lake. That’s why, for now, we still have human beings to override computers when they are wrong. The question becomes: can they or will they?

[UPDATE 7/11/18] After 7 straight days of blown highs, we’ve had a string of 5 days of highs within 3 degrees of the forecast! It certainly seems like there’s been some effort made to improve them especially throughout the day. I’m hopeful the lows will follow suit.

Leave a Reply

Your email address will not be published. Required fields are marked *