Available CMIP6 Variables

The full list of CMIP6 variables that simulations might output is quite large. And most of them only have meaning to Earth systems scientists. So, we’ve collected a substantially smaller subset of them for the Snowflake data marketplace. Specifically, we’ve collected:

Near Surface Temperature

Those graphs you see in the New York Times describing how the Earth will warm up due to climate change are probably based on the Near Surface Temperature. This variable is what the simulation thinks the air temperature will be roughly 2 meters above the surface of the Earth, or basically at eye level. If you’re interested in measuring “how hot will it be in my home town” this is pretty much the variable you want. A few things to be aware of: this variable is measured as an average across the time unit of the data set. So, for example, in our monthly data sets, this temperature is the average across all the days and all the nights across the entire month. In the peak of the afternoon in this ficticious, simulated future, actual humans will experience hotter temperatures than what this variable shows. And, likewise, in the deep of the night in this simulated future, actual humans will experience colder temperatures than what this variable shows. But it is still quite a useful variable to compare the temperature from the same season in, say, 2015 to the same season in, say, 2040 to see how much hotter, overall, things will feel.


In the raw data, precipitation is measured as a kind of “average precipitation velocity” meaning, basically, “on average, across this entire month, how forcefully was snow/rain/sleet falling.” In CMIP6 this is measured in “kilograms per meter squared per second” units. This is not a way a typical person thinks about precipitation, so in the post-processed data tables we’ve converted this for you to “how much precipitation fell on the ground across this time period” as in “milllimeters per day” or “inches per day” that you’re used to hearing about in weather reports. Bear in mind that this is still an average and represents only one day. So, for example, if you’re looking at a monthly data set, and we’re showing a value of “4.56 millimeters per day” for January, and you want to show “how much rain will fall in January 2045 in total” then you need to calculate 4.56 mm/day * 31 days in January = 141.36 mm total for every square meter of Earth.

Near Surface Relative Humidity

As with Near Surface Temperature, the Near Surface Relative Humidity is also meant to represent the humidity at roughly 2 meters above the surface of the Earth (i.e. eye level). Humidity is a trickier beast to understand than you’d think. The relative humidity is the one you’re used to hearing on weather reports and when people say things like “argh, I was so miserable in South Carolina this summer when it was 90 degrees and 99% humidity!” Here’s the tricky part, though: 100% humidity at 90 degrees F does not represent the same “amount of water in the air” as 100% humidity at 40 degrees F. The reason is that relative humidity represents: “compared to how much water the air could hold at this temperature, how much is it actually holding.” And colder air simply isn’t capable of holding as much moisture as hotter air, in the first place.

The other thing to know is that, especially in colder climates, humidity can be very “diurnal”, meaning it’s one value early in the morning and a wildly different value in the late afternoon and yet a third value in the evening. Imagine, for example, a June morning in San Francisco. As the fog rolls in and shrouds the streets in the morning, the humidity is fairly high due to the fog (say, 90%). But, then, as the sun burns off the fog in the sunny, late afternoon, then the relative humidity plunges to being delightfully low (say, 20%). If you average those two measurements out to give one number for the entire day, then you end up with a number that splits the difference (say, 45%) and doesn’t, actually, represent a humidity that humans really experienced at any point in the day.

Nevertheless, it’s still useful to compare the average relative humidity for January 2015 to, say, all the Januarys after 2040, to see larger trends. So, we’ve included it here in our data sets.

Near Surface Specific Humidity

As with other “near surface” variables, the Near Surface Specific Humidity is meant to represent the humidity at roughly 2 meters above the surface of the Earth (i.e. eye level). Unlike relative humidity, specific humidity does not change depending on the temperature or pressure of the air, which is why it’s so useful to meterologists. In other words, a specific humidity of 0.34 on a 30 degree day represents the same “amount” of water as a specific humidity of 0.34 on a 70 degree day.

One of the difficult things to understand about specific humidity is that it has no units. It’s actually a ratio. So, if you hear someone say the specific humidity is 0.34, you’re left listening for more, as in “0.34……of what?” Well, it’s just “0.34”, period, that’s what. What the 0.34 represents is the ratio of the mass of water vapor vs the mass of everything else the air is made of, usually a ratio of the number of grams of water vs the number of kilograms of everything else. So, if you have a specific humidty of 0.34, and you’ve captured a bunch of air with a mass of 1.0 kilogram in a container, then you know that 0.34 grams of water is suspended somewhere in that 1.0 kg of air you’ve captured.

Snow Area Coverage

If you’re trying to figure out what ski season in Tahoe is going to be like in the year 2050, then snow area coverage is the variable for you. This variable simply measures “what percentage of the land in this area is covered by snow”. So, a snow area coverage of 100% means there’s snow everywhere. And a snow area coverage of 50% means half the land in that particular grid cell is covered by snow, and the rest of the land is bare land.

There is only one cautionary note to this variable for grid cells that contain both ocean and land (i.e. they’re a coastline). For those grid cells, the percentage is only referring to the land part of the cell. So, if there’s a 100km square that’s 70% land and 30% ocean, and the snow area coverage variable says 100%, then what that really means is that (100% snow * 70% land) = 70% of that complete 100km square is covered in snow. In practice, this probaby doesn’t matter to you because you’ll be comparing the same grid cell in one time period to the same grid cell in a later time period to answer questions like “how much less snowy will this place get”, in which case you’re talking about the same place and so a 30% in one year and a 22% in another year is still a meaningful comparison, even if the cell is partially ocean.

Have suggestions for More Variables?

Like we mentioned before, the full list of available variables is huge. So if there’s something you were interested in and it’s not listed here, please get in touch at info@pollen.io.