SUMMER   PREVIEW   -- 25  MAY  2004  





The latter half of   SPRING   2004    bears a striking  resemblance  to the  SPRING   2003.  The  critical question is   whether  the pattern flip flop that occurred in  Mid July 2003 to a   HOT  dry pattern over the Plains and Midwest in 2003  will occur    this Summer... and when.

Given  the  number of   atmospheric  similarities   between  APRIL  / MAY   2003  and  APRIL / MAY 2004  it is  tempting to forecast a repeat  of a similar  type  of  Summer  ( that is the rainy pattern holding  over the  Plains  and Midwest   until   early  July and  a lot of  heat over the  Rockies). However there are some    critical differences   that Must be considered..



Click on the Maps for LARGER   View


JUNE  2004 JULY  2004 AUGUST  2004
JUNE  Jet.jpg (97180 bytes) JUNE  frcst.jpg (77616 bytes) JULY  Jet.jpg (98794 bytes) JUL   frcst.jpg (75816 bytes) AUG   Jet.jpg (98266 bytes) AUG   frcsrt.jpg (69995 bytes)




   Be sure   to check back later    Friday  Friday night  and    Saturday for more  detailed   explanations    forecasts and  ideas  
















1.  REVIEW  OF WINTER 2002-03 2.  USE of  ANALOGS
3.  Is there an El Nino Coming? 4.   Current  ENSO   conditions
7.   MEI 8.   QBO



Many  (but not all)  forecasters use "Analogs"  to help make a seasonal or monthly forecast. In weather forecasting the use of analog is an attempt to understand   the  seasonal forecast period  based (in past) upon similarities between a particular set of parameters.   The idea is by searching   for similarities  with  other  years or seasons    it gives a forecaster some clues as  to how the   monthly or seasonal forecast  may  develop. .

For example one may consider the fact that most  US winters   during   strong  El Nino years bring warm and dry conditions to the Upper Plains   and Midwest   and  rather wet and cool conditions to the Deep South and changeable conditions in the Northeast.   Or a forecaster may look at say the past 6 months... 12 months... or 18 months worth of temperature  and  precipitation data over a certain section of the country and use those similarities to the the current situation to assist them in making the monthly or seasonal forecast.

Done the correct way the Analog method has validity to it.   For example to simply assumed that all El Nino  events  produce   ONLY  one set of conditions in the winter months  is of course folly.    Some of the most severe winters in the last 100 years have been El Nino winters and likewise.... some of the most Mild winters in the last 100 years have also been El Nino winters.  Some forecasters  and   weather hobbyists   hold  the view   that analogs  are   not very helpful since  weather  records are incomplete or not sufficient enough to be considered as an indicator of what the upcoming seasons might hold.      For most locations temperature and precipitation data only extends back 100 or 130 years  and accurate Upper Air maps   only go back 70 years at best . That argument is   Valid  and ANY  seasonal forecast  using analogs  that is based on ONE   parameter is   very risky  and has a high degrees of failure  built within the forecast.

One way of getting around this is to use several factors parameters and cross matching the analogs to look for common ground.

Just   like   the  last several  seasonal    forecasts   I will use these   parameters and in this order of  importance





















Already asked to 10 years there has been some significant improvements in the development of seasonal climate models and their skill level... which basically is a fancy scientific charm that is used to describe the accuracy all the climate models.

Climate models have been in existence  for over 20 years but for most all of their existence they have been of such poor quality  that  long-range forecasters often either discount them completely or give them a very little if any credence.    However since the mid-1990s with the development of ensemble forecasting the climate models had been improving steadily each year by small incremental amounts. As we enter the middle portion of this decade some of the pine models had been showing fairly decent still over a thirty day and 90 day time frame.

(Click HERE   to see  what  ENSEMBLE forecasting is  and why it is  sooooo important  and Useful!)

Even so one must be careful in using climate models. It is a little like giving whiskey   and car keys to teenage boys.... An over reliance on the Climate models can lead one to disaster in terms of weather forecasting and potential trading positions.    It is important that only forecasters with experience in explaining the model bias and tendencies use  these  models.   I have been using climate models for several years and I have considerable skill in this area and I believe it is one of the reasons I've been pretty good with my seasonal forecasts over the last several years.

One of the primary climate models over the last several years has been the AGCM. (AGCM stands for atmospheric global climate model). The
AGCM is run the beginning of every month and a consists of a 20  member ensemble not just one climate model. The AGCM has been in existence and running every month since 1997 but only over the  last few years has the model shown any skill at all.

The AGCM has several very important biases which must be analyzed and taken into consideration before issuing a  ANY  seasonal forecast. One day there may a climate or seasonal model which is accurate and skillful enough so that the biases do have to be pointed out and a more reliable climate forecast can be considered.

The AGCM has a   serious cold and WET bias. This bias exists within the model because the AGCM is essentially an extension of the long-range American model which also is famous for having a cold and wet bias   after  day 7 towards developing large areas of storminess and precipitation.

Here are some older but recent runs of the A CM model which clearly show the
1 )severe cold bias of the model and
2) its inability to show ANY sort of significant heat at ANY time of  the year
3)  the  AGCMs  inability to show any sort of prolonged dryness in either Canada or the United States.


NOVEMBER   AGCM   for  the period  of  N-D-J  2002-03

As you look at these AGCMs keep in mind that the TOP map is the Actual depiction of what was going on in the three-month period .... And the bottom map is the model depiction or forecast of what the three-month period was supposed to be like.

*** click  on  the maps   and they will enlarge ***

200212_US_sfc_air_temp.gif (31061 bytes)  200301_US_Precip.gif (25494 bytes)  200301_US_sfc_air_temp.gif (12478 bytes)




For those of you that cannot recall the months of July and August 2002 placed in the top 10 hottest July August months on record for the Midwest. As you can see from these examples JUNE and JULY 2002 -- the three month period of JUNE JULY AUGUST 2002 was a hot and dry month across most of the Rockies... the central & Upper Plains and into the Midwest. Yet the June and JULY AGCM runs showed no HEAT of any kind of across the CONUS.

As you look at these  AGCM models keep in mind that the TOP map is the Actual depiction of what was going on in the three-month period .... And the bottom map is the model depiction or forecast of what the three-month period was supposed to be like.


*** click  on  the maps   and they will enlarge ***

200207_US_sfc_air_temp.gif (31860 bytes) 200208_US_sfc_air_temp.gif (36340 bytes)  

Do You see a pattern developing here with this Model? You should.

SUMMER   2003.


In this next set let take a brief look at  the AGCM from LAST SUMMER.....

For those of you that cannot recall the months of JUNE and JULY 2003 were very hot over the Western US with a record dry and Hot  period that lead to the worst Fire season ever  while over the Eastern US it was rainy and cool. Then in late JULY the pattern switched and AUGUST 2003 turned out to be one of the driest AUG over the Midwest ever.

The AGCM run from the month of MAY 2003 depicted NO heat at all ANYWHERE across the CONUS when in fact there was a large area of Above Normal temperatures over the the western US. This pattern continued through July of 2003... and even more so in the month of August. We can see in actuality that August 2003 showed huge areas of much above normal over the entire western United States by the model forecasted almost no significant deviations from above or below normal with regard to temperatures anywhere in the US or Canada.


*** click  on  the maps   and they will enlarge ***

 200306_US_sfc_air_temp.gif (25912 bytes)   200307_US_sfc_air_temp.gif (27935 bytes) 200308_US_sfc_air_temp.gif (29716 bytes)



Because of the  AGCM's appalling cold bias forecasters have been working on a new climate model over the last two years. This model has been showing significant improvement in its ability to forecast conditions reasonably accurate out to 3 and 6  months.

Below is a comparison of the April AGCM and its temps depiction for the period for the Summer and Autumn 2004... and in the right column... the New improvement climate model   (called the SFM). I have highlighted a PURPLE line over the maps to show the models depiction of the Jet stream.

As you can see there are 4 maps -- the Upper left map is MJJ-- the Upper right map is JJA --the Lower Left is JAS -- . the Lower right ASO. The beginning of the summer -- MJJ ( may June July) appears to show a large area of  Below Normal temperatures covering much of the Upper Plains   Midwest and into the Ohio valley ( ECB) s which implies a fairly deep trough over the middle portion of the nation. This pattern appears to continue into the heart of the summer (JJA).

Looking at the new climate (the SFM) in the right column we see a much hotter pattern developing in the first half of the summer over the heart of the Midwest which continues into the JJA period. In fact the model implies that there may be some sort heat dome developing over the Midwest by middle portions of the summer.


*** click  on  the maps   and they will enlarge ***

apr4cfm.gif (41427 bytes) apr4sfm.gif (32951 bytes)


However last Friday a significant change occurred in the climate models . 

Here is the June forecast from both the May AGCM and new SFM climate model. The AGCM   show a    very hot for the month  for June and it is arguably the hottest summer month I have ever seen from this model since I have began   following   this model  back in 1998. The more reliable SFM model is not quite that hot but it is significantly hotter than it was and the region of cool temps have shrunk and shifted to the Upper Plains.

MAY  AGCM   for 
JUNE   2004
NEW  MODEL  --MAY   FOR  JUNE   2004
SFMmon1_ts62.gif (35894 bytes) CFSmon1_62.gif (47203 bytes)



Looking at the new climate model  of May  2004 for the  SUMMER    the May  2004 AGCM suddenly made a dramatic shift in its depiction of temperature patterns across the Plains and Midwest  for the Summer of 2004 . As I stated above the March and April AGCM show a fairly cool and wet pattern across the Plains and Midwest  this summer. However the new MAY 2004  AGCM  showed a complete reversal of that idea with large areas of Above Normal temperatures developing across the Midwest. In  fact the MAY 2004 AGCM forecast for the heart of the Summer is the WARMEST summer pattern this model has ever depicted -- however keep in mind that the AGCM only goes back to 1998.

We see that the new Climate model is not only consistent with what   it  WAS showing in March  and April but the area of Above Normal temperatures over the Midwest and eastern Plains is in fact increased -- in terms of area and intensity.


*** click  on  the maps   and they will enlarge ***

may4cfs.gif (38331 bytes) may4sfm.gif (34356 bytes)


This is why the May 2004 climate Models are significant.    First even the colder bias AGCM is significantly warmer than what the Model was showing in March and April.   Second the MAY 2004 AGCM is the hottest summer forecast this model has ever produced.    Third the AGCM has Moved Towards the new and better climate model ( the SFM)... which tells me that the SFM is again outperforming the AGCM (which it should)! Fourth the Jet stream -- which again I have superimposed on the maps with the Thick purple line  -- shows a trough over the West Coast of US
and the Ridge  over the eastern Rockies Plains and Midwest. This is supported by the developing drought conditions over the Rockies and deep South as well as the developing sea surface temperature patterns in the Pacific Atlantic Ocean. In other words this may 2004 forecast makes sense... whereas the March and April 2004, models were not.

Here is the temperature forecast from the May 2004 AG CM for the summer. It has a large area of very dry conditions over the Middle Atlantic region into the southeastern states and normal rainfall over the Plains and Midwest. However as are shown above the model always under does the dry areas because of the cold and what bias. Thus I suspect the area of  Below Normal rainfall over the southeast to be much larger and press into a   good portion of the Midwest. In fact the rainfall forecast from the new climate Model shows a drier pattern over the Lower Midwest.

xprecip_us_current.gif (16955 bytes)   


The  sitaution  has    to be watched  very carefully.... and  it has major implications  for   the Grains and   Energy.  BUT   just like last Summer it is imperative that  one does NOT  Jump the Gun.   For example early  Last summer     both in the original  Summer forecast   and the Mid Summer Update  I went for   a WET    and seasonal JUNE and   JULY    and a much drier   & Hotter  AUGUST.   I received a steady stream of calls  from   folks wanting to know   when the  pattern shift was coming... almost beggiing to  go public with the pattern shift forecast  coming.

The point is that IF   the  wet pattern of  May 2004 does end    It Might  end  by  Memorial Day  or  June 3 or June 13 or  June 23rd.  I will watch  it closely  and stay on top of it.... I believe  that  when and if  this summer pattern does develop....    No oine will get you that weather info  faster   than I will.  






Weather models are based on something called GRID points.  In other words   weather models  are  essentially a elaborate game of CONNECT the dots!.
Each of these grid points or DOTs represents data that goes into the weather model's equation and programs. The closer the "dots"  or  grid points   are
together...  the better the data and the more likely it is   that  the Weather model to be close to "REALITY"  .

This data   in the  Grid points is derived front  several sources....   such as weather balloons pilot reports satellite data etc

That is why we call them MODELS. The  KEY point YOU must understand   is that there is ALWAYS  going to be some data that does NOT get into the computer and model equations. So the Models have to ESTIMATE what the missing data is.    Sometimes these estimates are really good.... and sometimes they are really bad. For instance over the Pacific Ocean there are few land stations to report Real data so the models have to use satellite reports and estimated data to fill in the blanks or gaps.

Now suppose one day the American model shows Big storm developing over the northern Pacific Ocean   over the next 2 or 3 days and   has the Low pressure area hitting Seattle in 6 days. The first model run is called the OPERATIONAL model.    No matter what happens the weather forecaster must ask himself    "what is the reliability of weather scenario this model  is depicting ? "    The best and oldest known way is to ask if the weather model "solution" makes   meteorological sense..  So if the model says it is going to snow in Chicago   3 inches in early May  -- that solution is NOT likely to be correct.

But back in the 1990s with the rise of advanced computers it was realized that you can use the new computing power to run the same model over and over  again. The result is a spread of solutions NOT  just one.  Hence the term    "ENSEMBLE" .    So lets say  you have 20 versions or MEMBERS of the American or Canadian weather  model along side    the operational Model run.

Now lets go back to that hypothetical storm  that the model shows hitting Seattle in 6 days.   With the Ensemble   a forecaster can now compare  the Operational Weather Model  to the other members of the  Ensemble.   This gives a forecaster a degree of certainty and probability that did NOT exists before.     Thus in our   case....   while the Operational Model shows the Big Low hitting   Seattle in 6 days but 10 other  versions of the ensemble show NO storm at all... and 8  versions show only a weak cold front... the forecaster can say with Much higher confidence that it is unlikely a Big storm will hit Seattle in 6 days.

Or on the other hand  the  Operational American weather Model  is   famous for developing  whacky  weather  scenarios  after day 7     in North America.  It may show a  Hurricane  suddenly   developing in the Gulf of Mexico.   By using the  ensemble  the forecaster  can get a  good idea of whether  such a  scenario     even possible or  likely.   

The American model ENSEMBLE has 21 members and is run twice a day out to 16 days... but the grid point resolution  ( the  DOTS)  are  spaced fairly far apart.  This model is undergoing an upgrade.   The Canadian Model also has an ensemble which  features 17 members   that run out to 10 days but it is only run once in the early morning hours. The European weather model ensemble has 51 members and has the best resolution of any of the Medium range weather models and is available in the pre-dawn hours     and again around   11am  CDT.     The Navy   model has a 9 member ensemble out to 10 days  and is run once in the early morning