Forecasts and Models

User avatar
ChrisGreaves
PlutoniumLounger
Posts: 15498
Joined: 24 Jan 2010, 23:23
Location: brings.slot.perky

Forecasts and Models

Post by ChrisGreaves »

In this post Ken points to Ventusky.
Ventusky point out that they are not a forecaster but a modeller.
Nonetheless we can choose "Run" and the historic pattern of wind/rain/temperature slides seamlessly into a portrayal of the model.

Which set me to wondering, exactly what is the difference between a forecast and a model?

Top of my head I would reply that a Model is a mathematical representation of [weather] conditions based on numeric data (readings from a specific place at a specific time), and that a Forecast is the result of a mental process that takes that data and gives a prediction of conditions at some time in the future.

Fifty years ago post offices telegraphed conditions to the weather office and men in white shirts and black ties sipped tea and drew on their vast knowledge, and in their head came up with a pretty good idea of whether it might be sunny or cloudy.

Today we have computerized data logging and computers that can do a fairly good job of modelling the thinking that a man used to do.

The Change By Stealth idea must apply. Tell me the temperature at 0900 hours and I can predict/model/forecast the temperature at 0901 with accuracy. It will be unchanged from that at 0900. My probability of accuracy goes down as we move through 0930, 1000, 1030 and so on. At that point I need to factor in, say, wind direction. For me, wind from the south means warm air from the Gulf Stream waters, so a rise in temperature. The mental process I use to predict/forecast the temperature at 0901 could be programmed by a VBA student on their first day in class. No complex computing here.

As the forecast period lengthens, my VBA model would begin to fail (in accuracy) much faster than the forecast from a seasoned professional with fifty years knowledge of local conditions.

But still and all, isn't a Expert System just an attempt to program human knowledge and strategy into a digital computer? Deep Blue's chess triumph was a computerised model to replace a human's mental ability to play chess.

I have perused the results of web searches for "forecasting versus modelling" and "weather forecasting versus modelling" and do not find an explanation of the difference that could satisfy this ten-year-old mind. For example: "Financial forecasting is the process by which a company thinks about and prepares for the future. Forecasting involves determining the expectations of future results. On the other hand, financial modeling is the act of taking a forecast's assumptions and calculating the numbers using a company's financial statements.". To me this says that "forecasting is a mental calculation whereas modelling is a computerised calculation", which means that the basic difference is the vehicle (man or machine) that performs the calculation.

(signed) "Puzzled" of Bonavista
An expensive day out: Wallet and Grimace

User avatar
HansV
Administrator
Posts: 78236
Joined: 16 Jan 2010, 00:14
Status: Microsoft MVP
Location: Wageningen, The Netherlands

Re: Forecasts and Models

Post by HansV »

Naomi Campbell is a model, but not a forecast...
Best wishes,
Hans

User avatar
kdock
5StarLounger
Posts: 720
Joined: 21 Aug 2011, 21:01
Location: The beautiful hills of Western North Carolina

Re: Forecasts and Models

Post by kdock »

:laugh: :laugh: :laugh: :laugh: Kim
"Hmm. What does this button do?" Said everyone before being ejected from a car, blown up, or deleting all the data from the mainframe.

User avatar
ChrisGreaves
PlutoniumLounger
Posts: 15498
Joined: 24 Jan 2010, 23:23
Location: brings.slot.perky

Re: Forecasts and Models

Post by ChrisGreaves »

HansV wrote:
12 Jul 2020, 10:00
Naomi Campbell is a model, but not a forecast...
Well, she has also begun a career as an actress, so she is presumably in the cast of some film or another.
Fore at last count :groan:
Cheers
Chris
An expensive day out: Wallet and Grimace

User avatar
Jay Freedman
Microsoft MVP
Posts: 1313
Joined: 24 May 2013, 15:33
Location: Warminster, PA

Re: Forecasts and Models

Post by Jay Freedman »

Back on topic (if that's permissible)... In at least some areas that use mathematical modeling to produce forecasts, there are many different models that use different assumptions about which inputs are important, how much weight to give to each of them, what intervals to use between iterations, and other details that cause each model to produce results that are slightly -- or maybe extremely -- different from one another. Producing forecasts usually involves comparing the results from a number of models and selecting what appears to be the most likely real-world outcome.

Examples of this process include the maps of predicted hurricane paths (rightly called "spaghetti diagrams"), macroeconomic studies, and political predictions from polling (cf. www.FiveThirtyEight.com).

LisaGreen
5StarLounger
Posts: 964
Joined: 08 Nov 2012, 17:54

Re: Forecasts and Models

Post by LisaGreen »

Well she is definitely not fo recast!!!

User avatar
ChrisGreaves
PlutoniumLounger
Posts: 15498
Joined: 24 Jan 2010, 23:23
Location: brings.slot.perky

Re: Forecasts and Models

Post by ChrisGreaves »

Jay Freedman wrote:
15 Jul 2020, 00:38
.. In at least some areas that use mathematical modeling to produce forecasts, there are many different models that use different assumptions about which inputs are important, how much weight to give to each of them, what intervals to use between iterations, and other details that cause each model to produce results that are slightly -- or maybe extremely -- different from one another. Producing forecasts usually involves comparing the results from a number of models and selecting what appears to be the most likely real-world outcome.
Thanks Jay.
I am beginning to think that "modelling" implies the use of strictly-defined mathematical models (nowadays executed on computer, buit fifty years ago pencil-and-paper), whereas "forecasting" implies "within the head".
I remember wheat farmers consulting their diary of weather records from time to time and noting that "whenever <this condition> occurs, we get "that condition" within the next two days.
In the southern part of Western Australia this is easy. A succession of low-pressure cells - The Roaring Forties - sweeps across the state on about a six-day cycle. So the footy is rained on for three weekends, then for three weekends you can leave your brolly at home!

The human brain is very good at modelling, so I suspect that the real difference is in crispness of data and process.

Modern technology (Satellite/Radar imaging) helps a great deal. I could sit in a west-facing apartment on the border between Toronto and Mississauga and through the window, see the rain coming at me from perhaps ten miles away, and on my monitor see the same rain storm sweeping across the map.

My head-bound modelling was based on a ten to twenty-mile window, whereas the radar could alert me half a day (or more) before.

The head-bound modelling today is most likely based on data being emitted by a cluster of mathematical models.

Cheers
Chris
An expensive day out: Wallet and Grimace