Thursday, May 05, 2011

EC predicting a hot dry summer.. or are they?

Environment Canada's preliminary 3 month summer outlook was issued May 1st, garnering some enthusiastic media headlines across the country about how EC was proclaiming a hot dry summer across the country. Stock up on the sunscreen and invest in that inground pool! This summer will be a scorcher according to the headlines. Unfortunately, this was more media hype than reality when taking a closer look at what EC's summer outlooks were really showing..

The image to the left is the outlook map that caused all the excitement. It shows most of Canada glowing red.. like some great fireball had landed on the country and oozed red hot lava all over the country. Only the coasts would be spared from this glowing fireball of a summer, and an inexplicable patch of coolness over southern Hudson Bay. Well, if the map is mostly red, it must mean a hot summer, right? Well, not really.

Let's take a look at the more revealing probabilistic forecasts from the very same model output.

This chart breaks down the odds of a warmer than normal, near normal, and cooler than normal summer for country, in increments of 10% probabilities. It is a far more revealing product than the grossly simplified image above and tells us a lot more information. For example, this map shows that the far north has the greatest chance of seeing an above normal summer, while it's not as likely over the south. It also shows a greater chance of a "near normal" summer over Eastern Canada, and cooler than normal over the West coast. Everywhere else, it's a tossup. This is a very different interpretation from the original map which showed virtually all of Canada above normal. This more detailed chart is known as the probabilistic forecast, and is a far more valid product than the oversimplified and misleading deterministic product that is so misinterpreted by the media and public. Simply put, the deterministic map is a poor reflection of the probabilistic output from the climate models, and should not be taken at face value without consulting the probabilistic forecast for additional interpretation.

What about precipitation? Again, here's the simplified deterministic map given to the media..

Blue areas denote drier than normal, and red denotes wetter than normal (shouldn't the colour scheme be the other way around when talking about precipitation?) Note that the Prairies and NW Ontario are shown as wetter than normal, while the West Coast and the Arctic are drier than normal, as well as parts of Ontario. The probabilistic precipitation forecast is much more informative (as well as having a more logical colour scheme) showing near normal pcpn over much of the country, drier than normal over the Arctic, and wetter than normal over BC and Alberta. Again, a vastly different story than what the oversimplified deterministic forecast here shows. It should be noted however, that summer precipitation forecasts have little or no skill, (less than 35% accuracy as national average) which means they're pretty much useless over much of the country.

Put it all together, and the forecasts are actually pointing to near normal temperatures and precipitation for much of the Prairies.. hardly the hot dry summer being advertised.

So how are these long range seasonal forecasts produced? Seasonal 3 month outlooks are produced by the Canadian Meteorological Center's supercomputer in Montreal by running an ensemble of 4 global climate models initialized by weather patterns over the last 10 days of the month. The models run simulated weather patterns for the next 90-120 days, and an average temperature map for the country over that time period is produced. This temperature map is compared to "normal" temperatures for the country for that 3 month period, and the outlook map shows areas of warmer than normal , colder than normal or near normal temperatures according to the model ensemble. This outlook is totally model driven with no human intervention or augmentation whatsoever. Since it initializes with the last 10 days of weather patterns, long range outlooks are heavily weighted towards climate anomalies over the past couple of weeks. It does not take into account expected changes in ENSO patterns or other major climatic influences and as a result, the seasonal outlooks predictive skill is very low.

In contrast, seasonal outlooks from NOAA's Climate Prediction Center (CPC) incorporate expertise and input from human forecasters who have experience in long range climate patterns. As a result, CPC seasonal forecasts tend to be more realistic and more accurate. Even here though, predictive skill with seasonal outlooks is limited, with the best skill noted during major climate episodes like El Nino or La Nina. For what its worth, CPC is predicting a cooler and wetter summer over the Prairies this year (see image above), at least through June, with a trend towards warmer and drier conditions towards the latter half of summer.

So, still think we're going to have a hot dry summer?

25 comments:

  1. Rob, this is a fantastic explanation and I can only thank you for the expertise you offer your readers.My simple question is this: why do we continue to suffer from inferior forecasting in this country i.e. why is there not some impetus for Canada to "up its game" and offer the same level of expertise that NOAA provides?

    ReplyDelete
  2. Thanks Rob. I think it should be recommended reading for any journalist!
    What do you think of the European model that Brett uses? It gets a lot of praise but back in 2009, it was dead wrong on its prediction of a hot summer for the prairies.

    ReplyDelete
  3. Great post Rob....yes required reading for all media folks.

    We get the same questions down here.

    Our local climate guy is indicating lingering La Nina conditions early on with a cooler and wetter May and June than normal then turning a bit drier with at least normal or a bit above toward August-September.

    ReplyDelete
  4. Amen to that explanation and the comments..

    About the >50mms predicted in the next 5 days.. is that the coloring influence of QPF (on my name)

    ReplyDelete
  5. WOW!
    That is the longest ESSAY...I mean post that Rob has ever put together!!!

    ReplyDelete
  6. Mark..

    Over the past decade or so, Environment Canada has undergone significant budget cuts which has had a profound impact on our weather service and its priorities. With more limited resources, the weather service has shifted its focus towards high impact weather within the next 24-48 hours. Beyond that time frame, EC has increasingly used automated forecasts from direct model output including Day 3-7 forecasts and seasonal forecasts. They can only do so much with what they have.. and their top priority is alerting Canadians to the potential for severe weather in the near future, which is what a weather service should be doing. So in a way, EC has "upped its game" somewhat at least in terms of forecasting severe weather in the short term. Unfortunately, other weather products of lesser importance (including seasonal forecasts) have to get by with more limited resources. And I don't think that will change unless more Canadians feel mid and long range forecasts are a high priority for them..

    ReplyDelete
  7. Andrew.. I don't follow Brett Anderson's forecasts (I assume that's who you mean..) but the European model has a good reputation for medium amd long range forecasts. However, even with their superior long range forecasts models, even the Brits have acknowldged that seasonal forecasts don't have enough skill to warrant issuing them on a regular basis.

    ReplyDelete
  8. Thanks for everyone's comments.. the main reason for this post was to give some background information on EC's seasonal forecasts, which can be quite confusing and offer little insight into how the forecasts were produced or what they are implying. This confusion is only amplified when the product is given to an uninformed media who are forced to jump to their own conclusions much of the time. The result is bad information all around, which only damages EC's credibility when these forecasts don't work out, or change. I'm just trying to clarify things a bit to help put these forecasts into perspective..

    ReplyDelete
  9. Jim..

    Not looking good for next week.. looks like an unsettled pattern coming in over southern MB.. especially Monday through Wednesday. Ensembles are showing potential for 15-25 mm over Winnipeg over that time, with an 25% chance of 50 mm or more over us.

    ReplyDelete
  10. High of 26 C and showers on tuesday???
    Is a warm front with risk of strong thunderstorms coming this way???

    ReplyDelete
  11. 26C Tuesday? with showers? Highly unlikely with main surface low over Dakotas and warm front to our south. GGEM pushes warm front furthest north, but still has it along or south of the US border. That high of 26C is being affected by the GGEM model seeing lots of warm air aloft moving over us, however we won't see it at the surface if we have showers/tstms and east or NE winds. Ensembles suggest highs of 10-15C more likely Tuesday with surface low moving to south and east of us which makes that 26C a noticeable outlier. It could happen if the low is further northwest and warm front pushes north of us, but confidence in that happening is very low at this point. This is the problem with automating forecasts.. an experienced forecaster would never issue such a bold forecast in this situation.

    Good setup though for elevated thunderstorms ahead of the warm front with areas of heavy rain possible over portions of southern MB. Could be an active period through early next week.

    ReplyDelete
  12. These computer driven predictions aren't worth the paper they're printed on and one has to wonder why EC even bothers to publish them???

    ReplyDelete
  13. Anonymous..

    Not all automated forecasts are bad. In fact, much of the time, automated forecasts are reasonably good (at least day 3-5 forecasts) and can go out "as is" with little value that a human forecaster can add. There are situations however where the model does not handle a pattern well. In these cases, a human forecaster can use their experience and pattern recognition to adjust and improve the forecast. Those are times that I can highlight in my blog to give additional background info.

    Overall though, improvements to automated forecasts need to come not only with better weather models, but with smarter and more sophisticated programming.

    ReplyDelete
  14. Ligthning strikes coming out of cell near Brunkild heading towards Sanford/south Winnipeg within next hour or so..

    ReplyDelete
  15. CWB Weatherbug site at Brunkild reporting 10 mm of rain past 20 minutes.. peak rainfall rate of 129 mm/hr..

    ReplyDelete
  16. Just called my parents in Elie, they got a good thunderstorm with heavy rain, pea size hail!

    ReplyDelete
  17. Thunderstorm over south/southeast end of the city moving east.. heaviest core south of Perimeter with 8-12 mm of rain past 20 minutes towards St Adolphe area..

    ReplyDelete
  18. The last thing some people wanted to see was heavy rain in the forecast!! Areas in in RM of Cartier and ST Francois are still dealing with very high river levels!

    ReplyDelete
  19. Nice little thundershower just passed through the St. Vital area here.

    Is that further convective development I'm seeing to the west of the Winnipeg area?

    ReplyDelete
  20. BTW.. my weather station is not updating due to loss of internet connection at home. Looks like a squirrel chewed through a cable line. Shaw can't come up til Tuesday to repair, so no Rob's Obs weather data until then. Dumb squirrel..

    ReplyDelete
  21. Nothing really much in St. James...only a light sprinkle a little while ago.

    ReplyDelete
  22. Tomorrow night could be interesting for elevated convection streaming up from stateside!
    Environment Canada is calling for up to 25mm of rain!

    ReplyDelete
  23. Rob....Should have gone with MTS...
    Squirrels don't like MTS cable lines....LOL

    Just joking, I'm with Shaw also :>)

    ReplyDelete
  24. False alarm.. it was just a tripped circuit breaker. Rob's Obs back on line. Apologies to all the squirrels out there..

    ReplyDelete
  25. Guess that hot dry summer happened after all :)

    ReplyDelete