Matt Lanza

The complex, nuanced reality of California and wildfire

I wasn’t quite sure where to go with this post, but we’re a weather blog, entrusted by a lot of people to give “to the point,” hype-free information. And we’d be remiss if we didn’t at least chime in on what’s happening out West. It’s tough to watch what has been happening in California without some sort of visceral reaction. This doesn’t feel right. It isn’t normal. This is our new reality.

What is “normal,” anyway?

But it’s important to step back and look at what has happened in California and try to understand it and try to make sense of living with risk. California has a climatological normal, apparently. There is a “normal” value of rain or snow or temperature in California. But that normal is really just taking years of extremes and averaging them together. The normal in California is and has always been extreme. A former colleague of mine used to describe “average” by using the analogy of the river being 4 feet deep across “on average.” Which is to say, at some points, that river may be 20 feet deep. At others, barely six inches deep. In that sense, California has average or normal weather.

Is a winter fire season “normal” in California? Not really. But it’s not unheard of either. The front page of the L.A. Times is below from December 28, 1956. This was a rough wildfire episode in the Santa Monica Mountains.

Front page of the Los Angeles Times, December 28, 1956. (LA Times via newspapers.com)

From that December 28th, 1956 edition, “Wind, humidity, drought, and other factors have combined to make the Santa Monica Mountains fire almost impossible to combat with usual methods, firemen reported yesterday.” Further, “existing firebreaks are simply jumped by spot fires which pop up 100 feet and more from the main mass of flame.” Does this sound familiar? It’s important to understand these things so we don’t lose sight of the fact that just because perhaps there were a number of years where something didn’t happen, it probably wasn’t more than dumb luck. Fires in the dead of winter, while uncommon, can, do, and have historically happened.

It also brings me to a challenging, difficult point to make. Let’s be clear: Climate change is real, and it is making fires burn hotter, more intense, and more frequently in these traditionally more typical offseason periods. Would we have seen the type and intensity of fires we see today under the exact same weather conditions in 1956? Probably not. But climate change is not at all the sole “cause” of these fires being as bad as they are. The reality is always nuanced and difficult and messy. We’re not going to get into all the reasons here but vegetation management? Kind of important! Political decision-making? Important. Water supply constraints and infrastructure? Important. Population growth and sprawl? Important. Regulations and required environmental reviews? Important.

Climate change? Important. It’s all important, but to box it in as one issue, neat, tidy, and clean is a misnomer. And as disaster expert Samantha Montano put it “if you minimize the cause to just climate, you prevent us from being able to address the full spectrum of causes.” (emphasis mine)

All this to say that despite some of the obvious reasons like climate and population growth that have led to fires worsening in California, it remains an extremely complex issue with complex causes that are not singular or simple to solve.

What set this event apart?

So what made this specific event so bad? Let’s start with the fundamental problem: Drought.

California has seen a totally split weather regime so far this wet season: Rainy north, bone dry south. (High Plains Regional Climate Center)

This winter has been borderline ridiculous in terms of how split the difference is in regime between northern California and southern California. San Francisco is sitting at 10.39″ of rain downtown for the season, compared to a normal of 9.44″ to this point. That’s about 110 percent of normal. Downtown Los Angeles? 0.16″ of rain so far this season, compared to an average of 4.76″ typically. That’s about 3.3 percent of normal. Southern California has had absolutely nothing this winter so far. Things have dried out, and there have already been two or three decent Santa Ana wind events to help exacerbate the drying out of fuels in this region. In other words, this is about as bad a scenario as you could ask for without any external triggers.

Unfortunately, the external trigger came on Monday and Tuesday in the form of strong Santa Ana winds.

A classic and unusually strong Santa Ana wind event setup across California early this week, peaking on Tuesday evening. (Tropical Tidbits)

Winds gusted as high as 100 mph in the mountains above Altadena (where the Eaton fire impacted), in the 80s above Malibu, and in the 60s in many other areas. These were unusually strong winds by Santa Ana standards, likely exacerbated by what we call mountain waves. As winds from the north hit the east-west oriented San Gabriel Mountains, the winds are forced up over the mountain and then thrust down into the populous valley below. A lot of times, you’ll see more of a northeast type wind usher in the Santa Anas, with typical isolated pockets of strong winds. But when you get this level of north push, you can basically create a standing wave over the mountain which just leads to constant wind being pushed down below. While the Santa Anas typically come with strong winds, events like this in these specific areas are somewhat rarer, particularly having such a strong storm as was setup over Baja. And it was exceptionally well forecast ahead of time, which is why dire warnings were issued days in advance of this happening.

As these winds get forced downward, through the high deserts, then down mountains to the coastal plain, the air dries out further as well, which is why you end up with such a perfect recipe for fire danger in these wind events.

Will it get any better?

The short answer is not really. If anything, there will be continued offshore winds and fire risk as additional cold air gets pushed east of the Continental Divide next week. This likely means another period of borderline critical fire danger in SoCal next Monday and Tuesday. Rainfall over the next 10 days from the European operational model looks paltry at best.

Meager, minimal rainfall is expected in Southern California over the next 10 days. (Pivotal Weather)

Winter precipitation patterns can change in a hurry on the West Coast, and obviously the major scarring from these fires means that there will be a major sensitivity to heavy rainfall if and when it does occur again. Be it this year or next year. So, this area will not see dry conditions improve over the next week or two in all likelihood. Let’s just hope that forthcoming offshore wind events lack much bite.

Anatomy of a minor Minnesota forecast ‘bust’ from last week

At The Eyewall, our ultimate goal is to make this the premier site or newsletter for weather and forecast information. How we get there involves some experimentation more than just forcing a post for post’s sake. Today we’re going to tackle a relatively small forecast bust that occurred in the Upper Midwest last week that speaks to the difficulties of forecasting sometimes.

In brief: Below, I show a case in Minnesota last week where the forecast sort of misplaced the higher-end snow totals. While modeling had been generally supportive of the higher snow totals along and north of I-94, which goes through Minneapolis, the highest totals actually ended up between Minneapolis and Mankato, south of I-94. It wasn’t a busted forecast in a classic sense, but it offered an opportunity to assess how models did. And wouldn’t you know it, the European AI modeling seemed to be the most consistent in identifying the riskiest areas for highest precip totals as being south of I-94.

If you read every one of our posts every day throughout hurricane season (bless your heart and thank you if you did), you would have noticed a bit of a change in “tone” or frequency of us discussing things like AI modeling or the ICON model. One post in particular stands out regarding Hurricane Francine.

I’ve been professionally forecasting weather for over 20 years now, and much like a generation of forecasters before me witnessed in their careers, I firmly believe we are witnessing one of the most consequential forecasting “revolutions,” if you will in a very long time. There have always been new or tweaked models intended to improve forecast output. Things change, they get better, the science moves forward. But what we are seeing now is what I believe to be a legitimate disruption. AI, machine learning, and more efficient computer power are leading to a new wave of models, new ways of developing models, and a turnaround time on these new tools that’s lightning fast compared to history. The European AIFS model is barely 18 months old, and it’s already become an absolutely essential tool in the toolkit. Every few weeks a company like NVIDIA or Google or a smaller firm comes forward and announces some breakthrough they’ve made in forecasting using AI and machine learning. Of course, science by press release tells us nothing useful. But as some of this stuff gets integrated and made available to forecasters, it seems clear that the hype has some legitimacy.

Snowfall analysis (interpolated) from southern Minnesota last week. Note the patch of 8 to 12 inches south and southwest of Minneapolis. (NWS Twin Cities)

So what does this have to do with Minnesota? Well, I wanted to walk through the anatomy of a fairly minor forecast “bust” of sorts that occurred in Minnesota last week. I’m not going to throw anyone under the proverbial bus here; the NWS and others did a fine job overall messaging how this system would impact the region, and for the most part that’s what happened. But there were some surprises in how this unfolded, which were fairly small in terms of mathematical error or impacts but may have led to the perception of a fairly meaningful forecast bust.

Last Thursday, a storm system passing quickly through the state was expected to produce a widespread accumulating snow, which it did. The general thinking was that the highest totals would be north of I-94, which runs from Fargo through Alexandria (a favorite place of this author’s) and St. Cloud into the Twin Cities. There had been signs this was coming several days in advance. Last Monday, the NWS in the Twin Cities wrote this in their forecast discussion.

The general model consensus this morning has this swath setting up along & north of Interstate 94, but a few solutions do displace it totally to the south across southwest Minnesota & northern Iowa. So in summary, confidence continues to be very high in this system developing and for at least these Winter Weather Advisory-level snowfall amounts to occur, but is still low to medium on where the heaviest amounts are most likely across Minnesota & Wisconsin.

On Tuesday, their forecast discussion had not budged much at all, and the models had basically gotten in line with the idea of the highest precipitation totals along and north of I-94.

Models over the last 12 hours have trended slightly farther south, with the consensus track now placing the highest chances for snow along the Interstate 94 corridor. Still, chances still look good to expect 3-6″ of snow along and north of the Interstate 94 corridor of central MN & western WI, with amounts tapering off to around 1″ across southern MN.

There was not a whole lot of change in tone on Wednesday either, other than the potential that a dry slot may cut down totals a bit further in southern Minnesota with some risk of getting into the Twin Cities. But in general, the consensus was for a 3-6″ snowfall in Minneapolis, with lesser amounts to the south. In fact, the Weather Prediction Center’s probability map of 4 inches or more of snow from that Wednesday showed about a 70 percent (high) chance of it occurring, and the geographically astute will notice it was very much along or north of I-94.

NWS WPC probability of 4″ or more of snow last Thursday, issued on Wednesday afternoon. (NOAA WPC)

If you look at their forecast of where the low pressure system itself would track, you would see it going across southwest Minnesota and into northeast Iowa.

Wednesday afternoon’s forecast of the low pressure track in the Upper Midwest. (NOAA WPC)

The colors indicate ensemble track clusters, while the solid black line and L’s indicates the WPC preferred track forecast. Putting this all together, the NWS issued this snowfall forecast below on Wednesday afternoon.

NWS Twin Cities snow forecast on Wednesday evening for last Thursday’s snow. (NWS Twin Cities)

If you looked at this and lived in Mankato or Faribault, you expected just a few inches of snow. Faribault would actually end up closer to 8 inches of snow. Thursday’s forecast update did correctly nudge totals up there a bit.

Thursday morning’s updated snow forecast for Minnesota. (NWS Twin Cities)

When all was said and done, as you can see near the top of the post, the heaviest snows actually fell along and south of I-94, not north of I-94. So why did this happen, and did any models catch this?

Well, first off, the models did pretty well with the storm track forecast. As shown by the WPC above, the low pressure system tracked into northeast Iowa and across northern Illinois as expected. No surprises there.

Let’s take a look at the forecast precipitation from the models versus what verified. These are from 6z runs on Wednesday, meaning this would have been the newest likely data you’d have at your disposal making a forecast on Wednesday, early to mid-morning. I have included the ECMWF classic operational model, the ECMWF AIFS model, and the HRRR model, our higher resolution short-range model, as well as a verification based on radar estimates. Click to enlarge it.

Clockwise from top: Verified precipitation totals based on radar observations, the ECMWF operational model forecast, the ECMWF AIFS model forecast, and the HRRR model forecast. (NOAA MRMS, Pivotal Weather, StormVista Weather Models)

What’s up here? The Euro operational, which is generally anticipated to be the best in these scenarios had the corridor of highest precip totals right along I-94. The HRRR model was displaced way north of I-94. The AIFS model had the highest totals south and west of Minneapolis. The AIFS was actually quite consistent with that story. The ECMWF operational model had wobbled north with the highest totals on Wednesday before correcting back southward prior to the onset of the event on Thursday, with Wednesday evening’s model runs basically right in line with the AIFS model.

One of the things that continues to fascinate me about the AIFS model is that it often locks in on a solution with respect to a storm track and stay fixated on that. Let’s look at the 10 runs up to Thursday afternoon’s low placement. The ECMWF operational model had a couple issues. It was too far south early on, then corrected about 3 tiers of counties too far north before eventually settling back closer to the Iowa/Minnesota state line.

The European model forecast of low pressure position for the 10 runs leading into 18z (12PM CT) on Thursday. (Tropical Tidbits)

By no means would this be considered a “bad” forecast. You’re talking about a really well done forecast overall. Interestingly, if we look at the AIFS model, there was a lot less bouncing around that occurred. The forecast position of the low stayed basically within the upper 2 to 3 tiers of counties in Iowa.

The AIFS model forecast of low pressure for the 10 runs leading into 18z (12 PM CT) on Thursday (Tropical Tidbits)

Generally speaking, the models seemed to do well with this system, but overall, it gets challenging in storms like this. If the exact track shifts 20-30 miles, you can take the best vertical velocity (or “lift,” rising air) in whatever direction that shift is in. I suspect that despite having a good forecast at a high level, the European operational model’s tendency to bounce around a bit possibly negatively impacted the forecast that had the higher totals generally near and north of Minneapolis as opposed to between Minneapolis and Mankato.

What is impressive to me is the continued performance of the AIFS model. It remains extremely imperfect from a specifics standpoint. In other words, because of how it runs, it’s usually unable to capture fine-scale features like precipitation maxes in major events, peak low pressure in a large hurricane, etc. But as a forecaster, I don’t need it to do that for me. What I need is generally accuracy, and then I can rely on other model data to fill in the gaps. In this instance, the AIFS was rather consistently showing the highest precipitation totals south of I-94, whereas its counterparts either flailed around a bit or didn’t catch on til the end. Even some of the highest resolution modeling we have to handle situations like this took until 12 hours or less before the system hit to adjust these totals farther to the south.

Anyway, to me this is another case where AI modeling could give you an edge. The challenge is feeling confident enough to buy in.

Dreaming of a white “hot” Christmas? Pattern change to put winter on pause for much of the U.S.

For meteorologists, winter technically starts at the beginning of December, not at the solstice which is coming up here shortly. But winter weather can happen all throughout fall too, and so far this season there’s been a bit of it around the country. We’ve seen some decent snows at times in the Sierra of California, where the snowpack is running around 100 to 200 percent of normal for the date.

California snowpack is in good shape for this time of year. (California DWR)

The lake effect snow belts have been crushed this year at times. Erie, PA for example is having their second snowiest start to the season since recordkeeping began there in the late 1800s.

Erie, PA is off to their second snowiest start to a winter on record. (NOAA)

Many people think of Buffalo being the lake effect magnet, and it is at times, but the wind direction has definitely favored some of the Pennsylvania and southwest New York snowbelts. Precip is running about 150 to 200 percent of normal over the last 30 days in Erie versus 75 to 100 percent of normal in Buffalo. You can also see a lake effect snow signal off Lake Ontario too toward Watertown and the notorious Tug Hill Plateau.

The lake effect snow belts have seen significant precipitation over the last 30 days, impacting the area between Cleveland and Buffalo and east of Lake Ontario quite significantly. (NOAA)

There’s been some cold at times, certainly in the northern Plains and Midwest and Northeast. But also the Intermountain West has been chilly at times. Not their coldest start to a winter by any means, but firmly in the middle of the pack or below average.

And we’ve got another blast of cold incoming here over the end of the week and parts of the weekend in the eastern two-thirds of the country, including a modest snow event tomorrow through Friday for parts of the Plains, Upper Midwest and Lakes. More mountain snows are likely out West also. But this pattern is going to change in a very big way after the weekend. A ridge of high pressure in the upper atmosphere that’s expected to shift about two standard deviations from the mean, with even a 20 to 30 percent chance of a 99th percentile event. In other words, the ridge is going to be strong for December. As you might expect, that’s likely to lead to abnormal warmth.

The December 26 through 31 temperature forecast is ugly if you like cold and snow. (Tropical Tidbits)

In fact, it’s not just likely to be warm, the confidence level in a warm forecast is about as good as you’ll ever see at this lead time.

Today's CPC 8-14 Day Outlook is the most above normal (reddest) one they have issued (2008-present) for the Lower 48. The average location in the Lower 48 has a 75.9% chance of being above normal, a 19.9% chance of being near normal, and a 4.2% chance of being below normal for the Dec 24-30 period.🔥

— Brian Brettschneider (@climatologist49.bsky.social) 2024-12-17T01:56:32.108Z

CPC forecasts are not technically intensity forecasts; they’re confidence in above or below normal forecasts. That said, one could probably interpret their map shown by Brian above and the 5 day average for the 26th through 31st above it similarly. Warmer than normal and potentially much warmer than normal.

So what’s the deal? Is winter over? No, it’s December 17th. Winter is not over. However, we need to see some changes to dislodge this developing pattern next week. It would be very helpful to see the pattern over Alaska change. The animation below shows forecast height anomalies 20,000 feet up over the next 2 weeks or so. Above normal heights will often correlate to colder weather being dislodged from Alaska and northwest Canada and deposited south and east into the continental U.S. Indeed, we see above normal heights in the near-term, followed by a swift transition to deep blue, below normal heights next week. Good news for Alaskans who like cold weather. By the end of the model run we maybe see some changes over Alaska in terms of rising heights. This could help lead to downstream changes over the Lower 48, but whether that happens or not just as we cross into 2025 remains to be seen.

A pattern over Alaska hospitable to Lower 48 cold transitions to one that is decidedly hostile to it next week before some relaxation of the pattern late. (Tropical Tidbits)

The bottom line: It’s more likely to be a warm Christmas than a white one in many places. Next week looks ugly for winter lovers across the Lower 48. There may be some changes by about New Years, but exactly how fast any transition back colder occurs remains very much TBD at this time.

As the 2024 Atlantic hurricane season officially ends we review the forecasts and what actually happened

Hurricane season ends today, officially, although it’s been quiet for the last 10 days or so. Offseason storms have occurred. The last wintertime tropical system was an unnamed storm in January of 2023 off the coast of New England. Hurricane Alex occurred back in January 2016 out in the middle of the Atlantic. An unnamed storm formed near the Azores in December of 2013. The 2007 Atlantic season continued into December with Tropical Storm Olga, which racked up $45 million in damage in the Caribbean.

An analysis of the 2024 Atlantic hurricane season through Hurricane Oscar in October. (NOAA NHC)

The 2024 Atlantic season will likely be remembered for the Appalachian flood disaster due to Hurricane Helene, as well as the relatively quick follow through in Florida from Hurricane Milton. Helene will almost certainly be retired in the post-season, the first “H” storm to get that treatment since 2017’s Hurricane Harvey. I suspect Milton will also be retired, which would make it the first “M” storm since Hurricane Michael in 2019 to be retired from the list. The “B” storms are particularly tough to retire, given that they’re usually weaker and earlier in the season. The last “B” storm to be retired was 33 years ago, Hurricane Bob (which this author remembers from being sideswiped in New Jersey as a kid). Beryl will also almost certainly be retired, less so for what happened in Houston and more for what it did to some of the Caribbean islands as a category 5 storm.

The “A” storms are a bit easier to retire, with notable intense storms like Alicia, Allen, Andrew, Anita, and Audrey on the list. Flood events like Allison and Agnes have also been retired.

Storm names are generally retired because they were especially memorable or catastrophic in terms of property damage or loss of life. It’s part of why naming works for hurricanes; if it was an especially bad storm, that name will never be heard again except in relation to that specific storm.

The total damage from this season is likely to top $100 to 150 billion in damage based on various reliable estimates. There’s a much deeper story to be told here. Between various disasters and inflation in recent years, the cost of insurance has risen dramatically. This year likely did not help those problems. Folks like Steve Bowen below, as well as Kelly Hereid and Susan Crawford’s “Moving Day” are very good resources to follow this ongoing story.

Most people tend to focus on the hazard portion of weather / climate risk. Which is important.The next step is connecting the dots so everyone understands that this is a real and growing economic / pocketbook risk.It could lead to a future financial crisis.www.redfin.com/news/florida…

— Steve Bowen (@stevebowen.bsky.social) 2024-11-20T20:40:04.086Z

How did the preseason forecasts do?

Back in May, I described the upcoming hurricane season as likely to be an “arduous slog.” It ended up that way in the end, sort of, but a bizarrely silent August threw a massive wrench in forecast expectations this year. NOAA’s official hurricane outlook called for the following:

  • An 85 percent chance of an above normal season (verified)
  • 17 to 25 named storms (18 storms, just verified)
  • 8 to 13 hurricanes (11, verified)
  • 4 to 7 major hurricanes (5, verified)
  • Accumulated cyclone energy (ACE) 150 to 245% of the median (About 120-125% of the 30 year average, fell somewhat short)

So overall this was a mixed bag of a seasonal forecast. There were some good calls and some overdone calls. In general, NOAA has actually had a bit of an under-forecast bias, with 4 of the last 5 years verifying above their forecast range.

  • 2023: Forecast 14/7/3 and verified with 20/7/3
  • 2022: Forecast 18/8/5 and verified with 14/8/2
  • 2021: Forecast 17/8/4 and verified with 21/7/4
  • 2020: Forecast 16/8/5 and verified 30/14/7
  • 2019: Forecast 12/6/3 and verified 18/6/3

So, they met the target this year, though the ACE forecast was overdone. Much of this season’s absurdly high forecasts were driven by excessively warm water temperatures in the entire Atlantic basin, as well as the projected development of La Niña in the Pacific Ocean. The warm water temperatures held on all season.

Water temperature chart of the Atlantic main development region shows that 2024 (blue) was variably at record levels throughout the season. (University of Arizona)

Notably for 2025, we remain near record levels across the Atlantic, except in the Gulf which has mercifully fallen back to just “above normal” levels. So the water temperature forecasts were accurate.

So what about La Niña?

Though it was a little sluggish to get there, the May forecast (red lines) for La Niña essentially verified (dotted line) within the ensemble spread that was forecast in May. (ECMWF)

Well, this one is a bit more complicated, but the answer is that it partially verified, yes. The ensemble spread shown by the multiple members in red lines above did a nice job capturing the realistic spread in possible options. And through August, the La Niña seems to have struggled to develop. We seem to be getting there now, sort of, though we are not officially there yet and we may not officially get there. We’re basically ENSO neutral, leaning negative. So while the forecast technically verified within the plume, it did not get to La Niña, which probably had some impact on reducing potential storms.

So what hurt August and why did this season “feel” like it really underachieved? Well, if we look at the variable of “velocity potential” as a proxy for “rising air,” where negative values (blue and purple) indicate more rising air than usual, we can see an interesting comparison to an active stretch like August 15 through September 15, 2020, which produced some hefty storms. Use the slider to see the differences.

A comparison of velocity potential between August 15 and September 15 in 2024 (less active) and 2020 (more active). (NOAA)

Both seasons featured generally significant rising air over Africa and/or the Indian Ocean, much like you would want to see in an active stretch. In 2020, however there was significant rising air over the Atlantic as well, whereas in 2024 we saw significant sinking air present in the Gulf, Caribbean, and parts of the western Atlantic. Interestingly, if you look at the rate of precipitation this year, you see a major difference in *where* in Africa the above normal rainfall occurred.

Comparison of precipitation rate between August 15 and September 15 in 2020 and 2024, showing a marked shift north in where the heaviest rain fell in Africa. (NOAA)

There was a good deal of talk this year about how the Sahara and Sahel were experiencing significant rainfall, and indeed there was a massive greening in the Sahel this year. Does this entirely explain why the 2024 Atlantic hurricane season was a bit less frenetic than expected? No, but it does offer a possible partial explanation and avenues for research that can help with future hurricane season outlooks.

Overall, there is a lot to unpack about this season. From a damage standpoint it met expectations. From a statistical standpoint, it met some expectations and fell short in others. The goal with this post was not to really explain everything, but I was hoping to at least offer some insights as to how what was predicted actually verified and what else may have been at play.

From this point, our site will go quieter. We will be monitoring the weather and likely chime in during some intense storms. This week’s lake effect snow probably would have offered an opportunity, and maybe on Monday we’ll push something. That said, thanks for your support this hurricane season! Keep spreading the word. If you feel so inclined, since we don’t have a “tip jar” here yet, you can donate through Monday at Space City Weather, our parent site in Houston to support our work! Thanks as well to our sponsors this season. Their support helped offset basically all our server costs and offered a small incentive for us to post. We are open to ideas, partnerships, etc. for the 2025 season, so feel free to reach out if you’re interested through the contact link. Thanks again!