Here are the current weather related datasets:

• annual-precip.json
• climate.json
• co2-concentration.csv
• seattle-temps.csv
• seattle-weather.csv
• sf-temps.csv
• weather.csv
• weather.json
• windvectors.csv

What were you thinking with respect to consolidation?

from vega-datasets.

I think we can probably merge (after looking at examples and the data more carefully).

• seattle-temps.csv
• seattle-weather.csv
• sf-temps.csv
• weather.csv

weather.json is similar but contains a prediction and is used for a specific chart so let's keep it.

from vega-datasets.

Thoughts on weather data

There seems to be one set of daily records and one set of hourly records.

Daily Weather

seattle-weather.csv contains daily weather information.

weather.csv contains the same data from seattle-weather.csv but has data from New York as well.

Note: The dates are slightly different formats.

Hourly Weather

seattle-temps.csv contains hourly weather information.

sf-temps.csv also contains hourly weather information with the same dates.

They can be easily concatenated with the addition of a location variable.

Proposal

I think a sensible consolidation would be to combine the hourly and daily datasets respectively

• weather-daily.csv
• weather-hourly.csv

Concerns

Removing seattle-weather.csv would break many examples and tutorials. If the data is combined then any example would have to filter before being visualized.

I haven't checked the usage of the datasets but there would be similar issues.

from vega-datasets.

Thank you for the analysis @eitanlees. I agree that removing the seattle-weather dataset would break too many examples so let's keep it. I like that you can quickly import it and create a demo visualization without having to filter/facet by location.

seattle-temps.csv and sf-temps.csv are not used that much and only contain temperature information. The source also says (which I know you created) "30-year temperature averages recorded hourly from the Seattle Tacoma International Airport weather station". But the dates are all for 2010. I'm inclined to just remove the two datasets and we can add seattle-weather-hourly.csv instead.

The data for Seattle is used in https://vega.github.io/vega-lite/examples/trellis_area_seattle.html and https://vega.github.io/vega/examples/annual-temperature/ and https://vega.github.io/vega/examples/heatmap/.

from vega-datasets.

The temperature data comes from a 30 year observation period and the average temperatures were reported. The study ended in 2010 and that is where the date comes from. This type of measurement is called an Hourly Normal, rather than a direct temperature measurement at only one time. It's used for studying broader trends in climate sciences. For more information see the NOAA Hourly Normal Documentation

I agree with the recommended changes.

from vega-datasets.

Oh, I see. Maybe it would be nice to join the normal temperatures (and other values) into the dataset then so you can compare actual to normal temperatures.

from vega-datasets.