This project was devoted to understanding the variability of rainfall on scales that are not easily captured by use of traditional instrumentation. Although radar is extremely useful for a variety of applications, the use of radar to estimate rainfall accumulations necessarily surrenders high precision for large areal coverage. For example, typical NEXRAD radar typically only reports a single value for every 5 minutes and uses that single value to represent all of the information associated with rainfall for a several-city-block-sized area. Anyone who has tried to "wait out a storm" while trying to run to your car knows that rain can vary over substantially shorter timescales and smaller spatial scales than a radar pixel, so we endeavored to try and better understand this variability by taking a "brute-force" approach. We purchased and deployed many high-resolution rain sensing instruments all within a tiny fraction of a single radar pixel -- partially to demonstrate how much variability can really occur on these space and time scales, and partially in an effort to understand how to account for rain variability in general.

Over the course of the project, we learned a lot of useful things related to rain measurement. Here's a partial list of the things we explored and published on:

Rainfall can be highly variable not only over a full radar pixel, but even on scales smaller than the sensing area of a single detector (only a few inches across); this can matter if we make uneducated assumptions about how measurements of rainfall over larger spatial scales downscale to very specific spatial points; the very spatially localized results can have much more relevance to phenomena like erosion which may not be well characterized by only using large spatial measurements.

The "normal" way of analyzing rainfall data (by dividing accumulations into intervals of equal time duration) may not be optimal if certain types of statistical analysis needs to be conducted. It may end up being more scientifically useful to break the data into unequal time intervals where things like drop arrival rate is measured to be approximately constant.

Averaging rainfall over time and averaging it over space are two very different things that will give very different results; rain changes as it moves and acknowledging that when we study it in detail is important.

Not all raindrops fall at the same speed. We have known this for many decades, but generally it is assumed that if you have two drops that have the same volume they generally will fall with the same speed. Our measurements reveal that this is an oversimplification and that two drops that have the same volume and exist in the same basic meteorological environment can still end up falling at different speeds. We're still trying to figure out why, but the fact that they don't all fall at the same speed is important because some of the ways the scientific community has historically measured rain has relied on an assumption of the speed of a drop being well predicted by its size.

Reliable measurement of precipitation depends not only on how much total rain volume is measured but how many drops are measured over what total time interval. The interplay of these variables -- along with how many instruments are used and where they are placed -- can be reasonably tricky to disentangle.

Some of the well-respected instruments used to measure rain on a drop-by-drop basis still have persistent issues associated with the individual drop measurements, which can yield slight errors. Although these errors are not expected to be significant enough to bias our estimates of total rainfall, individual drops or measurements gathered over small time intervals can be significantly in error based on using the instrument's measurements without careful re-processing of the data.

Perhaps the biggest challenge in all of atmospheric science is the interactions between widely separated spatial and time scales. For example, it is currently thought that if particles have a slightly larger chance to be clustered closer together than random chance would allow, light traveling through such a collection of particles would -- on average -- get further through such a clustered medium than a random one, even if you are talking about spatial scales much, much farther than the typical separation distance between particles. Since many atmospheric particles are known to cluster together (likely due to turbulence), this means that there remains the possibility that the way light interacts with atmospheric particles is a bit more complicated than we generally perceive. This deals not only with how much light gets through a cloud, but also on how radar pulses interact with suspended precipitation. Fortunately, it seems like the effect in this realm is relatively minor -- but we're still trying to figure out how large the deviation from our expected interaction might be.

Last Modified: 12/24/2020
Modified by: Michael L Larsen