|
|
|
Site-Specific Long Range Weather Forecasts are an essential tool for the Weather Derivatives, Risk Management, Seasonal Planning, and Trading industries. Historical Data and qualitative seasonal outlooks are no longer suitable for the needs of these growing markets.
Announcement: Weather 2000, Inc. -- The only firm to accurately and precisely predict the back-to-back-to-back harsh and snowy winters.
Case Study 1a: Boston - Winter 2004/05
|
In the wake of the over-hyped El Niño of 2002-2003 (see case study 2a below), many were hesitant to take a stand on what influence El Niño would have on the 2004-05 winter season. Government outlooks and predictions by other forecasting firms incorrectly looked to the super-strong El Niño of 1997-98 for their "high confidence" Winter 2002-03 forecasts for the Megalopolis between Boston and Washington, D.C. As a result, many were gun-shy and a broad brushing "Equal Chances" forecast of normal or below normal temperatures was issued by CPC and others. These fears were further fueled by warm weather that developed in November and early January. Nonetheless, we assured our clients to remain steadfast in light of our total season HDD predictions for a cold Northeast winter.
In the late summer, one of our clients came to us concerned that Boston, Massachusetts could possibly have its third consecutive cold and snowy winter season. We indicated to our client that we believed October 2004 through March 2005 would once again fall on the colder side of normal. Specifically, our forecast distribution indicated a mean of 4% colder than normal, and as much as 10% below normal. Our commentary throughout the Autumn reiterated the likelihood for a cold, and volatile winter, and our site-specific analysis allowed our client to know specifically how the winter would unfold. However, others in the Trading, Risk Management, and Retail communities were still skeptical and indecisive.
The winter of 2004-05 will be remembered by many, not only for the cold weather, but also for the many significant winter storms. Boston had its fourth snowiest winter season on record with 86.6 inches of snow, and it finished 6%* colder than normal, as indicated by the graphic to the left. * - Refers to Nov-Mar non-rounded HDD accumulations. |
For more verifications of client-requested forecasts prepared by Weather 2000 over the last several seasons, please see our Accuracy Statistics page.
Weather 2000 would be happy to customize Site-Specific Long Range Forecasts to help your business: contact us to discuss your needs today.
Case Study 1b: Mid-Atlantic - Summer 2004
|
The summer of 2004 is a great example of why broad brushing West-versus-East-versus-Central forecasts are potentially dangerous, especially during the summer. Localized effects such as sea breezes, evapo-transpiration rates, soil moisture fluxes, and station micro-climates can all produce different seasonal results for two cities that are geographically nearby. Utilization of our site-specific forecasts can unmask these idiosyncrasies before the season gets underway. The mid-Atlantic region experienced above normal precipitation for most of the Spring and Summer of 2004. Much of the sun's energy went into evaporating the excess soil moisture instead of heating the surrounding air. As a result, temperatures during the day stayed in the 80's for most of the summer, and very few 90 degree days were accumulated, even under the influence of the Bermuda High. The lingering excess moisture also contributed to very high humidity levels, making the afternoons very oppressive despite the cooler temperatures. |
||||||||||||||||||||||||
|
These high humidity levels had a nuanced effect at night. Locations both farther inland and less prone to the Urban Heat Island Effect typically see nighttime lows that drop off quite considerably. The muggy weather prevented low temperatures from falling and increased the CDD totals relative to normal in those locations. Since the coastal and highly urbanized areas typically see warmer nights, the warming effect of the humidity was not as pronounced. We correspondingly advised our clients of these fine details via quantitative analyses. The results for the four site-specific locations were remarkable, considering only about 125 miles separate the two farthest stations, but a scenario our clients were prepared for.
|
Case Study 2a: Cincinnati - Winter 2002/03
|
El Niño was the buzz-word of this winter season's forecast right from the beginning. El Niño conditions did form in the Pacific Ocean, but in the end, status conditions were weak at best. Government outlooks and predictions by other forecasting firms incorrectly looked to the super-strong El Niño of 1997-98 and assumed 2002-03 would turn out to be an easy winter to call -- widespread warmth throughout the Eastern two-thirds of the country. Before astronomical winter even arrived, the Eastern third of the nation was blanketed by snowstorms and entrenched in an unrelenting pattern of bitterly cold air, on cue with advisories we gave to our clients months ahead of time: In the late summer of 2002, one of our clients approached us with fears that cold weather would spill into the Midwest despite high confidence warm projections by other services. They asked us to make a site-specific forecast for Cincinnati, Ohio. Our research concluded that not only was cold weather possible for the Winter of 2002-03, but it was likely. |
 Government Forecast issued November 21, 2002 |
The widely read NWS outlook (at the right) was issued three months later than ours, in the midst of the heating season, and had the following imprecise and generic breakdown:
colder than normal: ~ 28.3%
near normal: ~ 33.3%
warmer than normal: ~ 38.3%
We indicated to our client that we believed the October through April period would have a much greater chance of falling on the colder side of normal than on the warmer side of normal. Our commentary throughout the Fall reiterated the likelihood for a cold winter, and our site-specific analysis allowed our client to know specifically how the winter would unfold; however, others in the Trading, Risk Management, and retail communities were left completely unprepared.
For more verifications of client-requested forecasts prepared by Weather 2000 over the last several seasons, please see our Accuracy Statistics page.
Weather 2000 would be happy to customize Site-Specific Long Range Forecasts to help your business: contact us to discuss your needs today.
Case Study 2b: Northeast - Winter 2003/04
|
After having been shocked by the long, drawn out Winter of 2002-03, people were afraid, yet skeptical, of the same thing happening again in 2003-04. We advised our clients that although we would once again see brutally cold temperatures in the Eastern third of the country, they should not expect the same drawn out, consistent cold seen in 2002-03. In the late summer, we made the analogy that if 2002-03 was the "Tortoise" of winter seasons, then 2003-04 would be the "Hare". Although net cold would result in the Eastern-third of the nation, extreme volatility would cause moderately chilly temperatures to be followed by record breaking cold. After a relatively temperate November, the first half of December 2003 proved to be extremely hostile, with record breaking snowfall from the Mid-Atlantic to New England . Despite a brief respite, some locations in the Northeastern United States then went on to record the coldest January in over a century! Those put to sleep by early season volatility and others' low-skill forecasts found themselves in a precarious position during the heart of the winter. |
||||||||||
|
Caution: The Winter of 2003-04 is a great example why trends and averages over the last five, 10, and even 30 years must be used with extreme caution! While people may have short term memories, mother nature reminds us that she doesn't pay any attention to contemporary extremes. Although some may have been lulled to sleep by the mild winters of the 1980's and 1990's, emerging research indicates that multi-decadal patterns favor more arctic plunges the next few years and decades. These patterns, similar to those of the frigid post-World War II era, will render impotent forecasts put out by the government and other forecast services that are based mostly on recent trends. |
|
||||||||||||||||||||
For more verifications of client-requested forecasts prepared by Weather 2000 over the last several seasons, please see our Accuracy Statistics page.
Weather 2000 would be happy to customize Site-Specific Long Range Forecasts to help your business: contact us to discuss your needs today.
Case Study 3a: San Francisco - Winter 2002
Prior to the 2002 Winter Season, many questions surfaced about upcoming U.S. weather patterns: What would be the temperature outcomes of an ENSO-Neutral Event? Would large-scale ridging and troughing dominate certain parts of the Continent? The prevailing consensus among government Climate Outlooks (via CPC), and other forecasters was that high pressure and subsequent widespread warm temperatures would be prevalent across the Western parts of the Nation.
 Government Forecast issued November 15, 2001 |
In the Fall of 2001, one of our clients approached us with a site-specific Winter forecast request for San Francisco, CA. Considering the fears and potential prospects of abnormal warmth, the client was concerned that excessive heat would negatively impact this particular California Weather Risk structure.
Upon analyzing the meteorological and oceanographic factors which would be affecting Coastal California and San Francisco specifically, we believed warm temperatures would not be present during the 2002 Winter Season. A very cool pool of Pacific Ocean waters was persisting offshore and would dictate a powerful sea-breeze and marine influence for several more months. Quantitatively our research concluded that as many as 1360 Heating Degree Days could be accumulated during the January - March 2002 period. Based on our probabilistic forecast distribution we called for a 90% chance of at least 1021 HDD's being tallied. This was looked upon as a bold and surprising prediction in light of the other warm forecasts and San Francisco typically accumulating only 1083 HDD's.
We advised our client that the heart of our forecast distribution was centered on 1161 - 1191 HDD's, and in fact San Francisco ended up with a chilly 1171 HDD's for that season. Both the skewness and precision of our prediction exemplified the importance and value of site-specific forecasting to the Trading and Risk Management communities.
For more verifications of client-requested forecasts prepared by Weather 2000 over the last several seasons, please see our Accuracy Statistics page.
Weather 2000 would be happy to customize Site-Specific Long Range Forecasts to help your business: contact us to discuss your needs today.
Case Study 3b: Nebraska - Summer 2000
|
After a very hot Summer 1999 across the Northeast and comparatively much cooler temperatures experienced across the central U.S., many believed Summer 2000 would feature much of the same. By January 2000, the National Weather Service (CPC) seasonal outlooks reaffirmed these beliefs with a general hot-East / cool-Central Summer 2000 forecast.
As can be seen from these outlook graphics on the right, CPC indicated an increased likelihood of colder than normal temperatures during the June - September 2000 period centered around Nebraska. Most of the reasoning behind their cool forecast was the application of temperature cooling trends over the last 10 years (OCN) observed in the center of the Nation.
Several of our clients had concerns regarding the cool Plains outlook by CPC, and asked Weather 2000 to examine the situation more closely. While conducting our research for an Omaha, NE site-specific forecast, we noticed that drought and low sub-soil moisture would exacerbate during the several months leading up to the Summer season. Among other factors, we believed such minimal precipitation and moisture / cloud cover would support ample sunshine and heat waves. Our forecast for Omaha (April - October 2000) therefore concluded that temperatures would actually be above normal. Specifically, our forecast distribution indicated a mean of 5% warmer than normal, and as much as 22% above normal.
Summer 2000 was in fact quite warm across the central U.S., and specifically was 11%* warmer than normal in Omaha, NE, as can be seen in graphic to the left. * - Refers to Apr-Oct CDD accumulations. |
For more verifications of client-requested forecasts prepared by Weather 2000 over the last several seasons, please see our Accuracy Statistics page.
Weather 2000 would be happy to customize Site-Specific Long Range Forecasts to help your business: contact us to discuss your needs today.
Case Study 4: Texas - Winter 1998/99
To further illustrate the value of site-specific long range forecasts, we will next examine a San Antonio, Texas heating degree-day (HDD) statistical forecast that Weather 2000 prepared for a client.
 |
As can be seen in the graphic to the right, San Antonio historically has a very large range in HDD accumulations during the winter season. At the time, of this forecast request in August, 1998, the Climate Prediction Center (CPC-NCEP/NWS) had issued their regular seasonal outlook for the upcoming winter season. Although the CPC forecast for Texas indicated an increased likelihood of warmth, their probabilistic distribution was so wide that sizable chances of near-normal and colder than normal temperatures were also included. Specifically, their forecast had the following breakdown:
colder than normal: ~ 13.3%
near normal: ~ 33.3%
warmer than normal: ~ 53.3%
This type of forecast is far too vague for the needs of the industries previously mentioned. Through research and analysis, Weather 2000 constructed a much more precise forecast for San Antonio. All forecasts are made using a fixed-confidence scheme of 80%, which means weather outcomes will fall within our forecasted bounds 80% of the time. In this example, the forecast was not only heavily skewed favoring warmth (virtually excluding any chance of cold), but was also extremely precise, encompassing merely 42% of the historical distribution (see graphic above). Based on this distribution, the client was provided with the following information from the Weather 2000 forecast:
colder than normal: ~ 1%
warmer than the historical average (<1582): ~ 95%
exceeding record warmth: ~ 8%
It should be noted that Weather 2000's forecasts account for record-breaking events, which because of their rarity, do not typically behave in a statistically even fashion. At the conclusion of the winter season, seven months after the forecast was made, San Antonio had accumulated 1098 HDDs, 34 away from the record and within our forecast bounds.
For more verifications of client-requested forecasts prepared by Weather 2000 over the last several seasons, please see our Accuracy Statistics page.
Weather 2000 would be happy to customize Site-Specific Long Range Forecasts to help your business: contact us to discuss your needs today.
Case Study 5: Pennsylvania - Summer 1999
  |
Frequently, large portions of the nation are not forecasted for at all by CPC (see graphic on right). They label this type of non-forecast as "Climatology" or "CL", which reverts back to uncertain and equal chances of anything happening. People often mistake these blank areas as a "near normal" forecast, whereas in actuality, it signifies that no forecast was made. That is one of the main differences between determinate and probabilistic forecasts, the latter of which is used for seasonal outlooks.
This was the case in February, 1999 when CPC had issued their regular seasonal outlook for the upcoming summer season. Since there was no forecast made for Pittsburgh, PA (or anywhere in that region), the user was left with a proverbial coin-flip of possible outcomes with the following breakdown:
colder than normal: ~ 33.3%
near normal: ~ 33.3%
warmer than normal: ~ 33.3%
Through research and analyses, from the hemispheric to the micro-scale, Weather 2000 constructed a precise forecast for Pittsburgh. The forecast was very narrow, and shifted toward warmth for the period June - September 1999. Statistically, we were 80% confident the cumulative CDDs would fall within the range 547 - 769. Based on this cumulative distribution (see graphic on right), the client was provided with the following information from the Weather 2000 forecast:
|
colder than normal: ~ 10% warmer than the historical average (>601): ~ 75% |
In line with Weather 2000's predictions, the June - September 1999 summer season not only verified warmer than normal, but the exact 747 CDD total was captured within the forecasted bounds.
For more verifications of client-requested forecasts prepared by Weather 2000 over the last several seasons, please see our Accuracy Statistics page.
Weather 2000 would be happy to customize Site-Specific Long Range Forecasts to help your business: contact us to discuss your needs today.
|
WEATHER 2000 is a registered trademark of Weather 2000, Inc. Copyright © 1997-2006 All Rights Reserved |
|