Being on parental leave has given me the opportunity to catch up with Sesame Street after 30 years. In a recent episode, I learned that Bert has always dreamed of being a TV weatherman. Well, I’m here to be the first to officially invite Bert, and any prospective students who have always dreamed of forecasting the weather for a living, to apply to the UC Davis atmospheric science major. With a little hard work in our atmospheric science classes, Bert, I know you can get over your stage fright and forecast the weather with confidence!
July 6th marks the first day of my family leave over the summer. I will be checking email and Twitter but much less frequently than usual. If you’re awaiting a slow reply to contact, entertain yourself with the newest UCD Cloud Library additions.
Davis has not seen any precipitation this month and will not see any tomorrow. The rumor is that this will be our first ever rain-free February. I looked back through the 50 years of data I’ve used for a couple of these recent posts, and sure enough, in those 50 years, every previous February has seen precipitation. The remarkable thing about our rain-free February in 2020 is just how much of an outlier this is. Not only have all previous Februarys seen rain, but the previous minimum number of rainy days in February is 7!
Warning: This is terrible use of statistics! The climatological probability of precipitation for February days in 36% (i.e. it rains about 1 in every 3 days normally). Given the 36% likelihood of precipitation on any given day, the likelihood of a rain-free February is just 1 in 229,688 (i.e. (1-.36)28). If that February has 29 days (like 2020), the likelihood drops to 1 in 356,951. So, it’s significantly less likely a leap year would have a rain-free February, not just a little less likely as may be intuitive. Of course, to do this calculation well, I really need a lot more data and some way of accounting for the autocorrelation of precipitation…
As a fun check of the data I do have: Februarys with 28 days have an average monthly rainfall of 3.50” and Februarys with 29days have an average monthly rainfall of 3.67”, or almost exactly 1/28 more. Cool.
I thought it was time to add a little intrigue. Below is my favorite picture from the the ever-growing stack of cloud photos (see right…). We’ll call it the unofficial 2019 photo contest winner. This one is of a rain shaft falling from a thunderstorm over New Jersey taken by Lea Tong. What I think is especially interesting is all the subtle structure in the rain shaft. What? Don’t see much structure that you think warrants any explanation? Ah, just wait for my next “forgotten cloud physics” paper to come out some time in 2020! It’ll “blow” your mind.
NJ Downpour © Lea Tong
Happy Thanksgiving, everyone.
A few days ago, a colleague from UCD Mathematics, Joseph Biello (of MJO fame), and I published a paper in npj Climate and Atmospheric Science. It’s available here. We attempted to use well understood aspects of the statistical properties of precipitation to derive a simple model for global precipitation. Then we populated model parameters with data from satellites and from cloud simulations. The nice thing is that our model is analytic, and because its constructed from physically understood pieces, it’s very easy to use as a sandbox. In the paper, we ask what possible future states of precipitation may look like given some arbitrary surface warming. The cool (or not so cool depending on your perspective) result is that many possible future hydrologies are possible. Yet models seems to predict a limited range of possible responses in the properties of precipitation. This could mean two things. 1) There are some consistencies among models that are physically based that we have not identified or 2) the suite of models is not large enough to span the possible responses. My gut feeling is that its probably a little of both. What I think is neat about this paper is that we have created an analytic, data-driven model of the climatology of precipitation properties. We’re hardly the first to do such a thing, but this one is mine.
I just wanted to point out all the new cloud photos on this website (look right). Special thanks to ASGG students Katherine Chin, Jenae Clay, and Megan Schmiedeler for contributing some excellent photos from all over the world. And more thanks to John (dad) and Jen (sister) for their contributions. If you’d like to contribute your own photos to the gallery, send me an email. My plan is to continue to grow this gallery with photos of interesting and unique cloud formations. My next mission is to reach 100 photos. Please use these photos with proper acknowledgement to the copyright owner if one is listed.
This is more than just a pet peeve. In many papers I review (I won’t attempt to blindly guess what fraction, but it’s high!), authors plot ratios of terms poorly. Let’s say you want to know the relative importance of the magnitude of longwave radiative cooling of a cloud top to the shortwave radiative warming. A reasonable way to assess such a thing would be to take their ratio and plot it up (over space or time or whatever). The common way (and I will argue the wrong way) one might visualize this is to plot up the ratio on a linear scale.
I’ve plotted up some synthetic data below. A and B are MATLAB vectors initialized with 100 random numbers between 0 and 1. The histograms of A and B are shown in the top row below. They’re nearly uniformly distributed. Now if I want to know something about the ratio of A to B, I might plot it up like I did on the bottom left. And, I might ask what the mean ratio is. It’s 4.3. This combination of information might lead me to conclude that the ratio is characterized by something greater than 1. Seems reasonable. But remember, I have two randomly initialized vectors; I would expect their properties to be about the same. I’m more likely to appreciate that with the figure on the bottom right. By taking the log, I give equal voice to things less than 1 and greater than 1. Excursions above the dashed line (at 1) and below occur with about the same frequency and magnitude. The mean of log10(A./B) is 0.06; When I raise 10 to that power (i.e. 10^0.06), I realize that my average ratio is better characterized by 1.1 (Or, about even odds of my random variables being greater or less than one another).
I’m getting tired of harping on this in reviews.
Yesterday’s rain was pretty noteworthy. I got 0.93″ at my CoCoRaHS gauge. That would make yesterday the 3rd rainiest May day of the last 50 years here in Davis. It also instantly jumps us to being the 10th rainiest month of May over the last 50 years. With more rain predicted today and next week, this could easily end up being one of the top 5 rainiest Mays ever.
Below is a bar graph showing the sorted daily rainfall totals for May days over the last 50 years. Notice a few things: 1) the logarithmic axes, 2) X-axis values start at day 1412 because 1411 of May days over the last 50 years have seen 0″, 3) the values nicely fall along a line in the log-log plot (except for the very heaviest days). This last point is an example of ubiquitous “power-law scaling” in the Earth system in which intense occurrences are very, very rare but also very much more intense than common events (think something like the great San Francisco Earthquake compared to the weak shaking SF feels commonly). Yesterday’s rainfall didn’t amount to the quake of 1906…more like the Earthquake of 1989.