This content was originally posted to the Trek View blog on 2022-07-29. Some of the information found in it may now be outdated.
One of the things I wanted to do was allow for filtering of my images using weather metrics.
For example, searching for sequence containing images with a wide visibility. Or searching for images in snowy conditions.
In order to do this I needed a source of historic reference data to cross-reference against photos.
There are a tons of services that provide access to such historical data to varying levels.
WeatherStack is a nice and affordable option for these requirements.
I used a photo taken in a fairly remote location (Helvellyn, Lake District, UK) because I wanted to test how measurements were reported when far from a weather station.
Here is the metadata in the photo needed to make the requests to the APIs.
exiftool -GPSLatitude -GPSLongitude -GPSDateTime -n GSAL1592.JPG
Exiftool by default prints data as Degrees, Minutes, and Seconds (DMS). Using the -n flag turns DMS into Decimal Degrees (DD), required for both the APIs.
GPS Latitude : 54.5255907
GPS Longitude : -3.003587
GPS Date/Time : 2022:07:13 12:13:07Z
WeatherStack Historical Weather API
My request to OpenWeather Historic Air Pollution API takes the following structure:
https://api.weatherstack.com/historical?access_key={apikey}&query={lat},{lon}&historical_date={date}&hourly=1&units=m&interval=1
For the test photo that gives;
https://api.weatherstack.com/historical?access_key=REDACTED&query=54.5255907,-3.003587&historical_date=2022-07-13&hourly=1&units=m&interval=1
The response contains a few sections that are well suited for my use-case…
- The location of the weather station
"location": {
"name": "Glenridding",
"country": "United Kingdom",
"region": "Cumbria",
"lat": "54.549",
"lon": "-2.952",
"timezone_id": "Europe/London",
"localtime": "2022-08-07 17:49",
"localtime_epoch": 1659894540,
"utc_offset": "1.0"
},
- The historical data for the day
"historical": {
"2022-07-13": {
"date": "2022-07-13",
"date_epoch": 1657670400,
"astro": {
"sunrise": "04:53 AM",
"sunset": "09:42 PM",
"moonrise": "10:21 PM",
"moonset": "03:36 AM",
"moon_phase": "Waxing Gibbous",
"moon_illumination": 97
},
"mintemp": 8,
"maxtemp": 15,
"avgtemp": 13,
"totalsnow": 0,
"sunhour": 15.8,
"uv_index": 3,
- The hourly data
Note, the hourly data section contains blobs of data representing hours during the date specified (&historical_date=2022-07-13&hourly=1&interval=1). The time property (e.g. "time": "1200", below), defines the time of the blob, in this case 12h00m (twelve o’clock in the afternoon). In total there are 24 hourly blobs for each day (0, 100, 200 … 2300);
"hourly": [
{
"time": "1200",
"temperature": 15,
"wind_speed": 25,
"wind_degree": 258,
"wind_dir": "WSW",
"weather_code": 116,
"weather_icons": [
"https://assets.weatherstack.com/images/wsymbols01_png_64/wsymbol_0002_sunny_intervals.png"
],
"weather_descriptions": [
"Partly cloudy"
],
"precip": 0,
"humidity": 71,
"visibility": 10,
"pressure": 1024,
"cloudcover": 37,
"heatindex": 15,
"dewpoint": 9,
"windchill": 13,
"windgust": 28,
"feelslike": 13,
"chanceofrain": 0,
"chanceofremdry": 0,
"chanceofwindy": 0,
"chanceofovercast": 0,
"chanceofsunshine": 0,
"chanceoffrost": 0,
"chanceofhightemp": 0,
"chanceoffog": 0,
"chanceofsnow": 0,
"chanceofthunder": 0,
"uv_index": 4
},
All the fields are described in the Weather Stack documentation here.
I cannot think of another weather measurement I would choose to add – the data is very comprehensive!
You can then either save this response to a database, or add it to each photos metadata using custom tags.