Good afternoon,

I'm working on an application makes calls to a weather API for different coordinates across the United States.

Instead of getting the weather at each and every point, I'm planning on clustering coordinates together based on how close they are to each other. That way, I can take an average of coordinates and get a single weather forecast that is representative for the group.

This raises the question, how close do these points need to be? Would a cluster of radius 50 Miles be sufficient? What about 250?

I'm really only gathering these features from my Weather source:

  • Temperature
  • Category (Rain, Cloudy, Hail, etc.)
  • Visibility

I don't really need a high degree of precision. In fact, the coordinates I'm using are going to have a 25-50 mile variation from their "true position."

I'm not a climate scientist, so I apologize if this question is a little rudimentary. I just wanted a second opinion on the topic so I can balance the trade-off between number of API Calls and fidelity to regional weather.

  • 1
    $\begingroup$ Topography & proximity to coastlines would be factors. Over a large plain weather can be "constant" over a large area. With hills & mountains, it gets more complicated. Coastal regions can be different from inland regions. $\endgroup$ – Fred Nov 29 '20 at 7:29
  • 1
    $\begingroup$ @Fred and then you have different parts of a cricket stadium have rain and clear weather at the same time. $\endgroup$ – gansub Nov 29 '20 at 12:45
  • $\begingroup$ @Fred earthscience.stackexchange.com/questions/18594/… $\endgroup$ – gansub Nov 29 '20 at 12:51
  • $\begingroup$ Mountains can make rainfall vary by 500% over a 35 mile distance, and weather prediction also uses super precise cloud albedo and temperature satellite data. The average can probably obscure data from areas with high altitude variablity. $\endgroup$ – DeltaEnfieldWaid Dec 3 '20 at 9:20

Weather is subject to Tobler's First Law of Geography: Everything is related to everything else, but near things are more related than distant things.

Temperature- this changes not only by the feature, but also the day of the year. The winter months, as well as the fall and spring, will have much more drastic temperature changes. This may be due to things like fronts, but is ultimately caused by heating differences from the sun. However, a few miles can yield quite a drastic change. Things become even more complicated when elevation is involved.

Precipitation- This is even more variable. Snow can be happening on a mountain, but have rain in a valley. Even the side of a mountain can influence how much rain is seen. I have seen it rain on one side of a building, but not the other. Precipitation is a tricky thing to forecast.

Visibility There are several factors that influence visibility. Fog and precipitation are probably the most common meteorological factors. Elevation can really influence your line of sight. Obviously, visibility is also influenced by the presence of smoke as well as dust.The smaller the particle, the more likely it will obscure vision. And snow tends to be more reflective than rain. Something else that may influence visibility is wind speed, as aforementioned particles can be kicked up from the surface. So visibility will change depending on how all of these change. It may even change depending on the direction you look.

All of this to say is that it really depends and there is rule of thumb.


I’m a pilot and deal with meteorology on a day to day basis and have just started to write my first App, albeit with very slow progress -so this is a really interesting question. I’m not sure if you’re aware but virtually all weather observations come from airports and this collective data along with computer modelling forms the basis of all weather reports and forecasts. It will be this data that your API is based on. If you wish to present ‘actual’ conditions your data points need to be close to where the observations are being made for obvious reasons. If your app is for a forecast area then the data points can be much further apart perhaps 30 to 50 miles with any error being time related not necessarily weather. You are most unlikely to be able to cover local effects so some users of your app may be disappointed. If you are interested in gaining specific weather information either actual or forecast in a specific area just type the nearest airport followed by TAF into google. This will give you the forecast, it will also bring up a METAR which are the actual conditions updated hourly.

  • 1
    $\begingroup$ "virtually all weather observations come from airports" What about satellite imagery of pixel precise rainfall and cloud albedo, altitude maps and ground temperature sattelites? Those contruibute huge datasets that go into numeric weather prediction models. In France there is a specialized weather station for every region of about 75 miles radius, which is more weather stations than airports. Locally the rainfall can vary by about 300% on average, on east and west side of a mountain, 35 miles distance. $\endgroup$ – DeltaEnfieldWaid Dec 3 '20 at 9:17
  • $\begingroup$ “What about satellite imagery of pixel precise rainfall and cloud albedo, altitude maps and ground temperature satellites?” You are quite correct this is precisely the equipment meteorologists use for forecasting along with observations primarily from major airports. $\endgroup$ – Rob Wilkinson Dec 3 '20 at 19:49
  • $\begingroup$ Are you able to confirm your calculation “In France there is a specialized weather station for every region of about 75 miles radius, which is more weather stations than airports” A specialised weather station every 75miles radius would cover an area of 17,671 Sq Miles. France has a total area of 248,572 Sq Miles. This would equate to 14 Specialised Weather Stations. There are 170 airports in France of which half receive commercial traffic. $\endgroup$ – Rob Wilkinson Dec 3 '20 at 19:59
  • $\begingroup$ Hey they said that there are professional weather ststions from the radome project every 30km, so 554 in France and 74 on overseas zones. I dunno precisely if the radome is very frequently updated. Also there are 42 synop stations on radio towers which are updated every 20 minutes. The uk is samish at 200 automated weather collection stations from the m.e.t. $\endgroup$ – DeltaEnfieldWaid Dec 4 '20 at 8:00
  • $\begingroup$ It's different here in every region because of the dense population, this website sais that there are 1100 stations in the official meteorology program including Radome which updates from 24 to 240 times daily: donneespubliques.meteofrance.fr/… $\endgroup$ – DeltaEnfieldWaid Dec 4 '20 at 8:19

It very much depends on what you are modelling. For example, it's well known that tornados require much more precise modelling than say rainfall.

In fact, good data on tornado formation is scarce given that they are unpredictable and highly localised phenomena.

Even the nationwide scan in the USA that does a scan every five minutes is problematic. Artifacts from this dataset has suggested that they are top down formed whereas recent data from a highly mobile monitoring equipment has suggested it is exactly the opposite - they are formed bottom-up.

  • 1
    $\begingroup$ Providing links can sometimes help give power to your message... it might be helpful to give one about the revelation on radar scan artifacts data? $\endgroup$ – JeopardyTempest Dec 3 '20 at 3:31
  • $\begingroup$ @jeopardyTempest: These days when search engines are ubiquitous all it takes is someone to enter a few key words in to locate useful referemces ... but I'll think about it. Thanks for the advice. $\endgroup$ – Mozibur Ullah Dec 3 '20 at 3:34
  • 1
    $\begingroup$ @JeopardyTempest: Well, I typed 'tornado formation down-up' into Duck Duck Go and the first search result mentioned this. I also got very similar results on Google. $\endgroup$ – Mozibur Ullah Dec 5 '20 at 21:10
  • 1
    $\begingroup$ I usually don't down vote good answers and especially not for new users, but I'm leaving a temporary and quickly-reversible donwvote here to further encourage you to update your own post and bring it up to the level of a proper Stack Exchange answer. SE works differently than other Q&A sites, and supporting statements in answers with citations and/or links is strongly encouraged in science and fact-based SE sites. I see that you are very active in many sites and in most of them (and especially in things like Mathematics) and answer can be seen to be self-evident. $\endgroup$ – uhoh Dec 6 '20 at 2:54
  • 1
    $\begingroup$ But here in Earth Science SE there is some history of problematic and unsupported answers and so we're trying to be vigilant and as much as possible maintain a supported-answer practice. So in this case "Feel free to update my post with the links..." and "These days when search engines are ubiquitous..." is probably not the best response here, and down voting and eventual voting to delete an unsupported answer and perhaps converting it to a comment may ensue. Welcome to Earth Science Stack Exchange! Please consider editing your post, or deleting it and posting a comment instead. Thanks! $\endgroup$ – uhoh Dec 6 '20 at 2:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.