After billing the round of downpours that would move through New England as a "heavy rain and energetic thunderstorm" event, I was heartily disappointed in my forecast - or perhaps I was disappointed in nature's failure to live up to my forecast - with a verification of heavy rain but a notable absence of thunder.
The rain certainly was quite impressive, with storm total amounts of one to two inches across Northern and Central New England, and convective downpours producing even higher amounts of 1.5" to 2.25" in Southern New England. This rain was timed rather inconveniently for the morning drive, and resulted in traffic accidents across New England and airport delays across the Northeast. So, from the public impact perspective, the forecast verified well, and the broadcaster in me knows that the public was adequately prepared for a rainy morning and will view the forecast as successful. The meteorologist in me (and for many of you who read this site) focuses on the science, though, and I'm unimpressed with the convection from this morning with regard to its lighting productivity. Any good scientist must go back and review what went wrong in the experiment (in this case, the forecast!) in hopes of improving upon it the next time.
In this case, I remarked in my forecast discussion here yesterday that vertical motion forecasts indicated air would be lifting to about 300 mb, where the temperature was a frigid -36 degrees Celsius. This was plenty cold enough for effective and frequent lightning production, for which I usually key in on the -20 Celsius isotherm. The first problem was that forecasts for vertical motion in the guidance products appear to have been overdone. Here's a time-height cross section, with time advancing from left to right, from the NMM guidance product's Thursday 12Z cycle, courtesy of coolwx.com's graphical interface. Notice the lift extending to 300 mb. When cross-referenced with the same time-height cross section of temperature, the extension of the lifting motion into the cold air is quite apparent. Though I haven't included it here, I did check the relative humidity forecasts, too, to ensure we had cloud production throughout the entire layer of lift.
Contrast that forecast with the next set of time-height cross sections: the NMM Friday 06Z cycle, run just prior to the arrival of the heaviest rain, and there's been a substantial decrease in how deeply the lift extends into the sky - by about 100 mb, now lifting to about the 400 mb level.
When cross-referenced with temperatures at this level, they're much warmer than 300 mb and just barely reach - or perhaps fall just short of - the threshold of -20 C.
So, why the difference in the forecast? Though I'm tempted to pin it on the lack of surface based instability, I think that's inaccurate. I knew yesterday that I was discounting surface-based convective parameters in my forecast, and I always worry that perhaps confidence has built too strong, leading to a weak arrogance, and I wrongly overlooked a parameter that was either lacking or present. I don't think that's the case here, though, because the key was the mid-level dynamic lift and assistance from the tremendous isentropic lift we examined near the 850 mb level. The latter definitely verified, and verified well, hence, the tremendous rain and rainfall rates. The former - the dynamic lift - evolved a bit differently than predicted, as the vorticity broke into several pieces, thereby speeding up arrival of the first rain band on Thursday night and also meaning the primary shortwave to traverse New England on Friday morning was weaker. This led to less mid-level upward vertical motion, meaning the cloud top fell shy of the all-important threshold. As for what caused this erroneous forecast of vorticity, it appears largely an issue with convective parameterization in the guidance products. Changes are always being made to the equations run in the guidance, and recent changes have rendered the NMM and GFS (especially the GFS) vulnerable to significant convective feedback. Essentially, this is when the model spins up thunderstorms, then allows those thunderstorms to continue redeveloping, even if the atmosphere wouldn't necessarily favor it. Thunderstorm clusters release latent heat of condensation, and in the model they almost become their own heat engines, much like hurricanes do in real life. This causes the model to keep cranking out more and more thunderstorms, stronger and stronger, far longer into the forecast period than it should. This has been a bigger problem here in the Northeast this spring than in the past *several* years, and is something we're struggling with behind the scenes literally everyday. In fact, I turned to one of our interns yesterday and said, "How are we even getting forecasts right?! We should be failing every time!" Of course, we're not failing every time, but that's hugely because we're keeping our eye on the bigger picture and trying to decipher details from the overall weather pattern, rather than specific model details.
This weekend will be an excellent test-run. The large-scale pattern shows a pattern favorable for shortwaves to traverse New England with a trough axis just to our west. This would support periodic scattered showers and thunderstorms ahead of the passage of each shortwave. The problem if you look too closely at the details is that there are actually DOZENS of small vorticity maximums crossing the region, which is why the NMM, for example, keeps virtually constant precipitation going through the interior all weekend long. The best model of late has been the convective-friendly WRF, and that indicates a good part of New England will be rain-free for quite awhile on Saturday, with scattered afternoon storms that will be most numerous in Northern New England, and across Connecticut. This matches well with the lower-resolution ECMWF, which - thanks to its lower resolution - is not falling prey to the same amount of convective feedback. The result *should* be a Saturday that allows for sun to burn through morning clouds and fog for most spots, bringing temperatures into the 70s before any of the scattered afternoon convection fires. Of course, there's always the possibility of isolated or widely scattered activity sooner near any convergence boundaries, but for most of New England, I still think Saturday looks like a good day with a decent amount of sun after a gray start for some, especially where there's a renewed storm Friday evening that will leave fresh low-level moisture. The axis of the vorticity maximum doesn't swing thru until Sunday 18Z, so this means Sunday will be a bit more convoluded with the potential for showers and thunder to linger in the cyclonic vorticity advection ahead of the shortwave, and especially near convergence boundaries, Saturday night. Convection probably wouldn't wait long to refire on Sunday - perhaps as soon as mid-morning - and would grow more numerous with diurnal heating until the vort axis moves east Sunday afternoon and anticyclonic vorticity advection moves in. Another shortwave follows for Monday.
This is the type of thought-process that must be conducted each day, then repeated each time the forecast is updated. A focus on more precise details from the models will lead to certain disaster in the forecast, because of the recent degradation of the guidance to adequately handle convection/thunderstorms.
Comments