What is the weather?

Today I’ll show you how in few simple steps build global temperature map, similar to Ventusky is offering. I’ll use data provided by - National Oceanic and Atmosfpheric Admistation, which is common source in case of similar visualizations.

When comes to global temperature map first thing we should talk about is so called resolution. It determines the number of nodal points for which the value of the interpolated parameter is assigned. To clarify what I’m talking about take a look on image below.

grid

Vertical lines represents meridians, horizontal lines - parallels, dots - nodal points. In such case (please note numbers assigned) resolution is equal 0.25 degree. Another commonly used resolutions are 0.5 degree and 1 degree (please note that smaller resolution results in more accurate model). Knowing that 1° of latitude is equal around 111 kilometers we can easily calculate that in case of abovementioned grid, nodes are approximately 28 km away. You might argue that it is high resolution already, but nowdays we can easily observe that temperature across one city might vary by several Celsius degrees, therefore I would be reserved with such judgments. Let’s choose 0.25 degree for this tutorial.

I’m visiting NOAA GFS page (GFS stands for Global Forecasting System), all other models and resolutions are availabe here. You should see there 10 subdirectories (state for Nov 21, 2023), for the last 10 days. They release forecasts each 6 hours, starting from midnight - inside selected day directory you should find another subfolders named after hours: 00, 06, 12, 18. Keep in mind it’s UTC time (aligned with meridian 0°), not your local one. The new forecasts appears typically dozen of minutes after corresponding time (this is the case with today’s data, for previous days you should see always all 4 folders). In next page you have to choose atmos subdir (probably it’s shorcut for atmoshpheric). Now you should finally see GRIB (General Regularly-distributed Information in Binary form) filter configuration

GFS setup

Starting from top of the page you can select one of files: *.anl, *.f000, *.f001, *.f002, …, *.f384. ANL file refers to an analysis file. The analysis represents the initial state of the atmosphere at the beginning of a model run. On the other hand, the f000, f001, f002 etc., represent forecast hours. These files contain the model’s predictions for different hours into the future, starting from the initial analysis time (in such case f000 means model prediction for now, last one - f384 - 16 days from now). I’m selecting current hour (.f000)

Next you can select model options. I haven’t deciphered all the possible options yet, but I think description is quite straightforward. In case of global temperature map following parameters are needed:

  • level desired: 2m above ground (this is what we as humans feel)
  • variables desired: TMP (temperature)
  • subregion: left longitude: -180°, right longitude: 180°, top latitude: 90°, bottom latitude: -90° (entire world)

At the bottom of the page we check Show the URL only for web programming and click Start download. We’ll be redirected to a page which shows a link to direct download. I’m copying the link since I preffer to encapsulate entire process in bash script.

#!/bin/bash

day=$(date -u '+%Y%m%d') # use UTC time
current_hour=$(date -u +'%H')

if [ $current_hour -lt 6 ]; then
    rounded_hour="00"
elif [ $current_hour -lt 12 ]; then
    rounded_hour="06"
elif [ $current_hour -lt 18 ]; then
    rounded_hour="12"
else
    rounded_hour="18"
fi

BASE_URL="https://nomads.ncep.noaa.gov/cgi-bin/filter_gfs_0p25.pl"
DIR="dir=%2Fgfs.${day}%2F${rounded_hour}%2Fatmos"
FILE="file=gfs.t${rounded_hour}z.pgrb2.0p25.f000"
VAR="var_TMP=on"
LEVEL="lev_2_m_above_ground=on"
REGION="subregion=&toplat=90&leftlon=-180&rightlon=180&bottomlat=-90"
URL="${BASE_URL}?${DIR}&${FILE}&${VAR}&${LEVEL}&${REGION}"
echo "\nDownloading data from noaa - ${URL}\n"
curl -L $URL -o temp.grib2

There is no magic here, just one thing to remember is to round current hour into hours predefined by NOAA: 00, 06, 12 or 18. Also I’m defining each search parameter in the new line to make script more readable. Last but not least is to fetch data with cURL - I want to save file as temp.grib2. After getting data I want to process the file. For this reason we will need CLI tool called GDAL (Geospatial Data Abstraction Library) - translator library for raster and vector geospatial data formats. On MacOS it can be downloaded with brew. For other operating systems download instruction can be found here. Once installed I’m checking file with gdalinfo command.

> gdalinfo temp.grib2 
Driver: GRIB/GRIdded Binary (.grb, .grb2)
Files: temp.grib2
Size is 1440, 721
Coordinate System is:
GEOGCRS["Coordinate System imported from GRIB file",
    DATUM["unnamed",
        ELLIPSOID["Sphere",6371229,0,
            LENGTHUNIT["metre",1,
                ID["EPSG",9001]]]],
    PRIMEM["Greenwich",0,
        ANGLEUNIT["degree",0.0174532925199433,
            ID["EPSG",9122]]],
    CS[ellipsoidal,2],
        AXIS["latitude",north,
            ORDER[1],
            ANGLEUNIT["degree",0.0174532925199433,
                ID["EPSG",9122]]],
        AXIS["longitude",east,
            ORDER[2],
            ANGLEUNIT["degree",0.0174532925199433,
                ID["EPSG",9122]]]]
Data axis to CRS axis mapping: 2,1
Origin = (-180.125000000000000,90.125000000000000)
Pixel Size = (0.250000000000000,-0.250000000000000)
Corner Coordinates:
Upper Left  (-180.1250000,  90.1250000) (180d 7'30.00"W, 90d 7'30.00"N)
Lower Left  (-180.1250000, -90.1250000) (180d 7'30.00"W, 90d 7'30.00"S)
Upper Right ( 179.8750000,  90.1250000) (179d52'30.00"E, 90d 7'30.00"N)
Lower Right ( 179.8750000, -90.1250000) (179d52'30.00"E, 90d 7'30.00"S)
Center      (  -0.1250000,   0.0000000) (  0d 7'30.00\"W,  0d 0' 0.01"N)
Band 1 Block=1440x1 Type=Float64, ColorInterp=Undefined
  Description = 2[m] HTGL="Specified height level above ground"
  Metadata:
    GRIB_COMMENT=Temperature [C]
    GRIB_DISCIPLINE=0(Meteorological)
    GRIB_ELEMENT=TMP
    GRIB_FORECAST_SECONDS=0
    GRIB_IDS=CENTER=7(US-NCEP) SUBCENTER=0 MASTER_TABLE=2 LOCAL_TABLE=1 SIGNF_REF_TIME=1(Start_of_Forecast) REF_TIME=2023-11-20T12:00:00Z PROD_STATUS=0(Operational) TYPE=1(Forecast)
    GRIB_PDS_PDTN=0
    GRIB_PDS_TEMPLATE_ASSEMBLED_VALUES=0 0 2 0 81 0 0 1 0 103 0 2 255 0 0
    GRIB_PDS_TEMPLATE_NUMBERS=0 0 2 0 81 0 0 0 1 0 0 0 0 103 0 0 0 0 2 255 0 0 0 0 0
    GRIB_REF_TIME=1700481600
    GRIB_SHORT_NAME=2-HTGL
    GRIB_UNIT=[C]
    GRIB_VALID_TIME=1700481600

We can see couple important informations here. First of all image size is 1440 x 721 pixels. It is saved on 1 band, using Float64. In metadata section we can see that Celsius is used as a temperature unit. That’s enough I have to know. I’m adding 2 new lines to my script now:

gdal_translate -of GTiff -sds -a_srs EPSG:4326 temp.grib2 temp.tif
gdaldem color-relief temp.tif palette_celsius.txt temp_color.tif -alpha

gdal_translate does exactly thing as it says - converts raster between 2 different formats. By adding a_srs argument we change projection into EPSG:4326. As an output we get TIFF image. Second command converts grayscale image into colorful image using palette_celsius.txt. This simple file contains 4 columns: temperature value (in Celsius) as well as three color components R, G, B associated with this value. In each next row we define color for different value in ascending order. I’m borrowing colors and thresholds from Ventusky website. Entire file content you can see below.

-40 238 238 238
-30 255 170 255
-20 145 9 145
-15 36 24 106
-10 85 78 177
-5 62 121 198
0 75 182 152
5 89 208 73
10 190 228 61
15 235 215 53
20 234 164 62
25 229 109 83
30 190 48 102
40 107 21 39
50 43  0  1

If you get lost somwhere around the road, you can see entire bash script here. Keep in my mind that in corner case which I did not cover, script will not generate image cause NOAA folder does not exist yet (folders are not created punctually, there is always some delay). In such case you can manually overwrite hour with previous one. I’m running the script and view the image. Now I know what is the weather 😅.

Global temperature map
-40
-30
-20
-15
-10
-5
0
5
10
15
20
25
30
40
50

Pretty cool as for tiny bash script don’t you think ? Just to make things clear typically RGB image is not send to the frontend application. Usually app fetches grayscale image, which is later drawn on a canvas using WebGL shaders, where color pallete is applied. But the script I’ve prepared is perfect for home made visualizations. NOAA gives you limitless possibilities , you can check temperature on different levels (even below the ground!). Also other parameters are available like relative humidity, pressure , wind speed. It’s very good place to experiments. That’s all what I wanted to share today. I hope next time you’ll think about the weather You just run the script 🤣. See you soon!