Do you ever find yourself expecting rain but see only sun? How about the other way around? Welcome to springtime in Texas, when cold fronts that once pushed through decisively -- in a way a meteorologist can time within the hour -- end up stalling juuust north of the region, as warm air wins out. When stationary fronts setup to our north, we generally stay dry, but can get absolutely soaked ... and often, while we know, "why", we don't know if or when.
Here's why: When fronts stall, strange things can happen within 200 miles of that front. Rain can "run away" from the front unexpectedly or may not move at all! Often in these cases, the weather features which make it all happen are so small, our computers (and our people) don't, "see them". You can't forecast what you can't see.
Computer models see the earth and its weather in little chunks. The highest resolution operational computer models see chunks at 4km by 4km resolution, or 16 square km. (These are called, "4KM models".) In other words, for 4KM models to accurately gage what's about to happen, a feature needs to be larger than 16 square kilometers. For perspective, Jersey Village is about 9 square km in size, so even if a weather feature was almost twice as big as that Houston suburb, the model would still miss it.
4km models with 16 square km chunks is considered state-of-the-art, super high resolution. The best global models (EURO and GFS) are 18km resolution, so a system must be 324 square km in size to be detected then.
Here's where the problem lay with spring systems: Often, when it rains for a while in one spot --- like along a stationary front -- colder air can, "pool" behind each individual thunderstorm. This cold air pooling can eventually act to, "push" storms in one direction or another. Unfortunately, often these cold pools are less than 16 square kilometers in size and are thus, "invisible" to the models and impossible to forecast using even the best computer models of today. (That's where we step in with human eyes and brains, but even then, unfortunately for these types of short-range forecasts, while we understand the physics of the atmosphere, it's humanly impossible to calculate the chaos to know where it may go ... exactly ... and when.)
These small features are often referred to as, "mesoscale" or, "medium scale." Models are generally too dumb to see mesoscale and make a forecast based on that. When you hear the weather person mention, "mesoscale", you know it's inherently unpredictable based on today's tech.
Now, back to the beginning when I was talking about how rain from fronts that move in a decisive way are much easier to forecast than stationary fronts... A cold front or warm front (a boundary that is either pushing in colder or warmer air) resides in a scale no less than 10,000 square km. That means the front will fill at least 31 of those 324km squares. It's therefore easily inputted into a forecast model.
In an example of how this type of unpredictability can have devastating impact was with the recent Memorial Day floods and Tax Day floods. We knew it could rain that day, and that it could be locally heavy, but the elements which contributed to the storms moving were too small for the models to detect accurately. I remember we saw the, "setup" for potential flooding, but that it was impossible to say if that would happen over College Station or downtown Houston. It hit Houston and history was made.
Welcome to our world!