Author Topic: Weather Forecasting for Builders – Concepts and Links  (Read 52 times)

0 Members and 1 Guest are viewing this topic.

Offline Reninco

  • Junior Member
  • **
  • Posts: 51
  • Welcome to the CP-Forum
Weather Forecasting for Builders – Concepts and Links
« on: March 28, 2021, 04:47:12 AM »

This turned out to be a longer first post than intended – links will be in the second post.

In light of the recent horrific suffering in Texas I feel forecasting is a timely topic for builders and the Country Plans readership.
From personal experience after working many decades I have been subjected to incidents from flying debris, falls on ice and heat related issues. Thinking back all could have been avoided with modern day forecasts that have evolved in the last 7 years. My Hope is the readers will gain more insight into forecasts and not repeat my painful lessons.

The basic concepts - or it helps to know this stuff before you go to the next stuff.
Forecast Concepts
A weather forecast is the prediction of what the atmosphere will be like in a particular place at a particular time.
The prediction or forecast is based on a weather model.
A weather model uses direct observations, scientific knowledge and math stuff in combination with supercomputers to generate an atmospheric prediction in a particular place and time…such as…at 1:00PM it will be 56 degrees near my house.

Particular Place: Definition
A particular place for a forecast model is called a grid.
Grid size can range from .75 km to 45 km per side and are assigned a permanent latitude and longitude location.
Grids can be imagined as a square post extending indefinitely upward.
Grids have selected forecast elevations – imagine 500 foot sections of that post.
Grids have more sections the closer they get to the surface – imagine 200 foot sections in the lower portions of that post.
Common models have over 100 different grid elevations and most extend into the ocean and/or soil.
There are 6 different paths in a cube for weather to flow.  With In-Out flow that gives 12 different types of movement per individual sections of a grid. This image show a single grid then a group of 9 grids.

100 x 12 = 1200 possible movements per grid. 
A lot of calculation of numbers gives the term numerical weather prediction or NWP. I could mention differential equations but I suspect most readers would rather be pulling splinters from their thumbs.
Forecast of single states of the atmosphere are commonly referenced as plots, parameters or elements in weather charting. 
Familiar elements are: temperature, wind and precipitation.
There can be up to 200 different forecasted elements per each grid level.
1200 movements X 200 elements = 240,000 types of flow for each grid for each time the grid is digested for the model forecast.
The reason for so many levels is that weather “flow” is primarily vertical (think sun and gravity pushing/pulling vertically) rather than horizontal.
Grid plots generate one “value” or number per each state of change. Due to the problematic challenge of reading a tiny number in a gridded forecast, the easiest system is to assign a color that is associated with that “number” for visualizing weather flow on a chart. It’s also called colorful fluid dynamics or computational fluid dynamics or CFD. With the advent of color tv and monitors this style of view is now second nature but it is not always the best way to understand a forecast. A little test of visual acuity, do you see the numbers or colors first?

And on a greater scale with many elevations and grids.

Gust forecast grids surrounding Pierre, South Dakota. I have circled single grid cells. The “pixelly” look is typical of all gridded forecasts.

Forecast Time
Particular forecast times for weather models are usually associated with each hour in UTC time.
Due to complexity differences; some models have larger hour steps and smaller models can have quarter-hour steps. *
Ownership
Due to cost, complexity and the support needed for weather models, they are usually developed by nations or large university systems. The US spends around 2 billion per year to create their different forecasts.
Countries that Produce Models
Nations associated with common models: European consortiums, United Kingdom, Germany, United States, Canada and of course the Swiss. Also, many countries copy public models and produce their own regional forecasts from that model. See GFS below.
US Models
Currently the United States (government) generates around 100 different weather models.

The United States models are considered “public” they are free to use, copy or distribute.
For forecasting, please realize…some of the US models are in the process of being discontinued, some have small important legacy features, some are more accurate, some have lost funding, some have gained funding, some are seasonal, many have duplicated plots, and a few struggle with accuracy and should be ignored.
To keep it simple I’ll start with common models that are stable, have funding and have shown to be improving in accuracy in the last 5 years.
Key US Models
GFS – The Global Forecast System, this covers the globe in 13km grids and is produced every 6th hour in 16 day blocks or chunks of time. Due to the recent upgrade made to the GFS model (called upgrade package FV-3) this model is at-times referenced as FV-3…I’ll just keep with the GFS acronym*. 
RAP – The Rapid Refresh Forecast, this covers half of the globe and is focused on weather approaching and departing North America, it has 13 km grids and is produced every hour (hence the rapid acronym) in 21 hour blocks of time*.
HRRR – High Resolution Rapid Refresh Forecast (RAPs baby brother), this covers the continental US in 3 km grids and is produced every hour in 18 hour blocks of time*.
*Size, time and production varies somewhat – these are the default products commonly found on most charts. More in later posts.
HRRR forecast grids surrounding Pierre, South Dakota.

Weather Suite vs Weather Model
A weather suite is a collection of different weather models. Tropical Tidbits and Pivotal Weather are typical weather suites. Suite charting drawbacks are that due to the sheer volume it is problematic to show every single grid elevation of every single model plot with every time step at every grid location for every model. You’d be eating the proverbial elephant in one sitting so most suites just provide the common atmospheric elements.

Other Models
Often referenced on the weather channel and now by many other “weather broadcasters” is the “Euro” model. Named European Centre for Medium-range Weather Forecasts or ECMWF, it is privately owned by about 30 different European countries. Due to private ownership the forecasts are limited or downscaled for the public side but I’ll give ideas in the next post on how to get this valuable forecast. Due to the computer size, and nature of model production, it is a slightly more accurate long-term model than USA’s GFS model…and unless the United States commits more funding it will always be more accurate. 

Quite a lot to digest – I’ll give general tips and links in the next post.

Offline Reninco

  • Junior Member
  • **
  • Posts: 51
  • Welcome to the CP-Forum
Re: Weather Forecasting for Builders – Concepts and Links
« Reply #1 on: March 30, 2021, 03:43:19 PM »
One last step is an altitude conversion as most weather charts use the air pressure reference. Air pressure vs altitude will fluctuate during the course of the day but this table will be close enough for builders. Shown are the common references in millibar equivalents seen on weather charts…consider the equal sign as an “close-enough” sign. If you live “between” elevations its best to use the higher elevation as the better weather reference.
Sea Level = 1013
350 ft = 1000 mb
2500 ft = 925 mb
5000 ft = 850 mb
10000 ft = 700 mb
18000 ft = 500 mb
30000 ft = 300 mb
Other chart abbreviations
An element or plot will also be given at meters above surface level, these are typical abbreviations.
 2m = 2 meters
10m = 10 meters
80m = 80 meters
Charts can be quite busy; a clearer way is to get regional charts that show finer details as most browser zooms will not reveal a better detail in a colored chart. A global chart for finding a temperature in your home town is a lesson in frustration.
To develop or pick a forecast these are the starting steps
1. Model
2. Region
3. Element or parameter of interest
4. Elevation (that is close enough) along with a surface parameter
5. Time of day
So, for Pierre South Dakota establishing a temperature for any part of the day I would use these selections:
1. HRRR Model
2. North Central Region
3. 2M, Skin (earth temp), 925 mb temperatures (these three will be very close to each other)
4. Surface, 925 mb
5. I would not focus on the exact time, you will be more accurate with the trend…”the temperature will quickly pass 32 degrees between 10AM and 11AM.” Looking at looping charts helps to clarify trends.
6. Repeat with the RAP model and Euro model

Links: Models and Suites
Models
HRRR and RAP links are directly from the production team – they have little or no latency.
HRRR:  https://rapidrefresh.noaa.gov/hrrr/HRRR/Welcome.cgi?dsKey=hrrr_ncep_jet
RAP:  https://rapidrefresh.noaa.gov/RAP/
Typical HRRR chart selection with features labeled. There are close to 150 plots with this chart so scrolling is necessary.

GFS is a 13km resolution, this is an experimental but almost operational version with 3 km grids of continental US that forecasts out to 48 hours. https://www.emc.ncep.noaa.gov/mmb/mpyle/hrefv3/00_exp/main_conus.php
Alternate GFS: https://weathermodels.com/index.php?r=site%2Fpreview&mode=animator&set=GFS%2050-STATES%20USA&area=South%20Dakota&param=2-m%20Temperature&offset=0&thumbs=1

Weather Suites
Typical weather suite selection…

SpotWx (best of the bunch, has exact grid finder, shows simple line graphs of key elements) https://spotwx.com/
Weather Models (Euro access to private maps for .50 a day, second best of bunch) https://weathermodels.com/index.php?r=site%2Fpreview&mode=animator
Tropical Tidbits: https://www.tropicaltidbits.com/analysis/models/
WeatherNerds: https://www.weathernerds.org/models/
Pivotal Weather (pay site of free models) https://www.pivotalweather.com/model.php
National Weather Service (has grid finder, gives short written summaries for those that like words rather than colors, is slow to locate and does not update often enough and has large time steps, after 1 day model defaults to larger grid or to model of lower accuracy, has no elevation correction, a warning multipage disclaimer with hard to locate link for that page, picture may show woods while location is in an inner-city, graphic with warnings seem to stop at State lines rather than a obvious geographic feature) https://forecast.weather.gov/MapClick.php?lat=44.37&lon=-100.35

Wind Visualizer Suites
Ventusky (best of bunch has full HRRR and archive functions and will show grid values) https://www.ventusky.com/
Windy TV: https://www.windy.com/
Earth: https://earth.nullschool.net/
HintFM:  http://hint.fm/wind/
NOAA visualizer (will not load at times,  balky selections, link provided just as a comparison) https://www.nnvl.noaa.gov/weatherview/index.html

I do not use Accuweather or Weather Underground as they have too many scripts running in the background and too many ads in the foreground. 

SpotWx gives all the major plots as line charts – here is the Spot line chart for wind…and yes it was pretty windy this day.

SpotWx front page with grids surrounding Pierre South Dakota. Lower page shows Models, grid size, time and date. Note the grid size - larger is better for long range trends smaller for accuracy.

« Last Edit: March 31, 2021, 04:19:37 AM by Reninco »


 

Templates: 5: index (default), Ads (default), Portal (default), Display (default), GenericControls (default).
Sub templates: 12: init, html_above, adsheaders_above, body_above, adsindex_above, portal_above, main, portal_below, adsindex_below, body_below, adsheaders_below, html_below.
Language files: 3: SPortal.english (default), index+Modifications.english (default), Ads.english (default).
Style sheets: 1: portal (default).
Files included: 37 - 1131KB. (show)
Cache hits: 13: 0.00265s for 40,676 bytes (show)
Queries used: 27.

[Show Queries]