Written by Sig Silber
We published the NOAA Long-Term Forecast Part I on July 18, 2020. Now we compare the NOAA forecast for Alaska and CONUS with the JAMSTEC forecast. It is easier to see the disagreements by comparing the maps which we show side by side in a table with a brief summary of the comparison. Obviously, the farther out you look, the less confidence you have in the forecasts and thus the differences in the forecasts. Also provided are the JAMSTEC World Forecasts. In Part I, we showed the differences in the assumptions with respect to ENSO (JAMSTEC is more convinced than NOAA that we will have more than a borderline La Nina) which may explain some of the differences in the forecasts.
I thought of providing the updated assumptions but both forecasts are based on the assumptions at the time they made the forecast so I have decided it would confuse the issue if I provided an update on the assumptions. I may do that at the end of the Month when NOAA updates their August forecast or I may wait until NOAA updates the ENSO Status on August 9.
Please share this article – Go to the very top of the page, right-hand side for social media buttons. Please feel free to send this article to anyone who you think might find it interesting and or useful.
C. Comparison of the NOAA and JAMSTEC Forecasts
Below is the comparison of the NOAA and JAMSTEC temperature and precipitation forecast maps for three time-periods and from left to right the NOAA forecast for Alaska and CONUS (the contiguous mid-latitude U.S) and then JAMSTEC for North America (which includes Canada and Mexico). The NOAA forecast maps can be clicked on to enlarge. The JAMSTEC maps in the table are not set up to be clicked on to enlarge (because we have no larger version of them). We have concluded that these smaller images work fine for comparison purposes. Later in the article, we show the World Forecasts.
JAMSTEC works with three-month seasons: Fall SON, Winter DJF and Spring MAM. Out of each three months, there is one where the months in the two forecasts align perfectly for the first time period. This is not one of those months so for the first period we are comparing SON for JAMSTEC against ASO for NOAA. It is not ideal but ok as long as you keep in mind that it is a slightly different period but pretty much Fall in both cases. We could have used the NOAA SON forecast which would then line up perfectly. But we thought the ASO forecast since it is closer to today’s date would be of more interest.
In addition to the value of comparing the JAMSTEC and NOAA forecasts, the JAMSTEC forecast by showing North America provides more context for the Alaska and CONUS Forecasts as the temperature and precipitation patterns cover North America, not just Alaska and CONUS.
Map Comparisons and our Comments
Temperature
| NOAA Alaska Plus CONUS | JAMSTEC North America | |
| FALL NOAA ASO 2020 JAMSTEC SON 2020/2021 |  |  |
WINTER DJF 2020-2021 |  |  |
SPRING MAM 2021 |  | |
Precipitation
| NOAA Alaska Plus CONUS | JAMSTEC North America | |
FALL JAMSTEC SON 2020 |  |  |
WINTER DJF 2020-2021 |  |  |
SPRING MAM 2021 |  |  |
JAMSTEC World Forecasts
This month our comments are taken directly from the JAMSTEC discussion (the translation may be a bit stilted). Now that JAMSTEC has a number of models we are not exactly sure of the best one to display and they vary somewhat. We assume that the JAMSTEC discussion considers their various models and provides their assessment having considered their suite of models.
Fall which is SON 2020
| Temperature |
 |
| Precipitation |
 |
Winter which is DJF 2000/2001
 |
| Precipitation |
 |
And Spring which is MAM 2021
| Temperature |
 |
| Precipitation |
 |
D. Conclusion
As usual, there is substantial disagreement between NOAA and JAMSTEC. It is difficult to relate the differences in the forecast to differences in assumptions on ENSO. But JAMSTEC tends to consider other factors that may not be considered by NOAA. On the other hand, the new versions of the JAMSTEC model are early in their usage and may not have been fully calibrated. I do not think I have gone into it in detail but when it comes to models we need to recognize that there are limitations and NOAA and JAMSTEC use very different approaches which I have summarized in the below table.
| Shorter Term | Intermediate-Term | |
| NOAA | Deterministic | Statistical |
| JAMSTEC | Deterministic | Deterministic (may also use statistical methods) |
| Assessment | Generally Considered to be reliable for 14 to 28 days | Errors build up in deterministic models and statistical models generally have insufficient historical data to be reliable |
So it is kind of a pick your poison choice. But both agencies have great skill at employing approaches that have inherent limitations.
JAMSTEC Discussion
We provided the full NOAA Discussion in Part I. The much shorter JAMSTEC Discussion was published on July 17, 2020 Prediction issued on July 1, 2020 and we included it in Part I but we are repeating it here also.
ENSO forecast:
Observation shows that a weak La Niña is developing at present. The SINTEX-F predicts that the weak La Niña-like condition will persist in the latter half of this year.
Indian Ocean forecast:
Observation shows that the tropical Indian Ocean is warmer-than-normal at present. The ensemble meam suggests that the warmer condition will persist in autumn, then will return to a neutral-state from winter. However, there is a large uncertainty in the prediction as some of the members (specially, of the SINTEX-F2-3DVAR system) actually predict a weak negative IOD event in autumn.
Regional forecast:
On a seasonal scale, the SINTEX-F predicts that most part of the globe will experience a warmer-than-normal condition in boreal autumn, except for southwestern Australia, India, and some parts of northern Africa. In boreal winter, the model also predicts that most part of the globe will experience a warmer-than-normal condition, except for western Brazil, western Australia, India, and some parts of the Indochina Peninsula.
As regards to the seasonally averaged rainfall in boreal autumn, a drier-than-normal condition is predicted for some parts of the South American Continent, southern Africa, West Africa, China, some parts of Europe. In contrast, most part of Northern America, Mexico, India, East Africa, Southeast Asia, Philippines, Indonesia, northern Europe, and northern Russia will experience a wetter-than-normal condition. In boreal winter, a wetter-than-normal condition is predicted for northwestern coastal area of Canada, a northern part of the South American Continent, northern Australia, western Europe, Philippines, eastern Indonesia, and Madagascar. In contrast, southern U.S.A., southern part of the South American Continent, central Africa, Spain, Portugal, western Indonesia, eastern China will experience a drier-than-normal condition.
The model predicts most part of Japan will experience slightly warmer-than-normal condition in autumn as a seasonal average. In winter, most part of Japan will experience warmer- and drier-than-normal condition.
E. JAMSTEC, NOAA and other Agency Indices
We showed these in Part I which can be accessed here.
F. Description of the New JAMSTEC Model
Tropical climate variations such as El Nino/Southern Oscillation (ENSO) and ENSO Modoki in the tropical Pacific and the Indian Ocean Dipole Mode (IOD) have enormous impacts on the global climate and the human societies. Therefore, there is a significant benefit to our societies if these climate events are predicted sufficiently ahead of their occurrences. Since the mid-1980s, many research institutes and operational centers have developed various numerical prediction models for ENSO forecast. Also, the IOD and ENSO Modoki predictions are attempted by some of the leading modeling groups recently. The numerical prediction of weather has been proven to be very useful to us now days because of the superb advancements made in that area of research during last several decades. Such weather forecast systems mostly employ standalone atmospheric models on the assumption that the oceans do not change in the relatively short prediction period (~1 week). However, such standalone atmospheric models are not ideal for predictions of climate phenomena like ENSO, ENSO Modoki and IOD that strongly depend on the ocean-atmosphere interactions. Application of ocean-atmosphere coupled model is naturally a proven approach to overcome the shortcoming of the standalone atmospheric model and to realistically simulate the climate phenomena. For our climate predictions, we have developed the SINTEX-F1 ocean-atmosphere coupled general circulation model under the EU-Japan research collaboration. Based on this seasonal prediction system (“F1”), we have performed climate predictions at least 1 year ahead and distributed the prediction information on JAMSTEC website since 2005 (LINK). We have achieved great successes in these years and SINTEX-F1 has become one of the leading models of the world for predicting the tropical climate variations in particular the IOD, the ENSO and the ENSO Modoki. (publications) To improve prediction of extratropical climate, a upgraded CGCM called SINTEX-F2 has been developed; the new system is a high-resolution version with a dynamical sea-ice model (“F2”). For the tropical climate variations in the Pacific and the Indian Ocean, the SINTEX-F2 preserves the high-prediction skill, and sometimes even shows higher skill especially for strong events, as compared to the SINTEX-F1. In addition, it has turned out that the new system is more skillful in predicting the subtropics, particularly, the Indian Ocean Subtropical Dipole and the Ningaloo Niño. The SINTEX-F1/F2 seasonal prediction systems adopts a relatively simple initialization scheme based on nudging only the sea surface temperature (SST). However, it is to be expected that the system is not sufficient to capture in detail the subsurface oceanic precondition. Therefore, we have introduced a new three-dimensional variational ocean data assimilation (3DVAR) method that takes three-dimensional observed ocean temperature and salinity into account. This system (“F2-3DVAR”) has successfully improved seasonal predictions in the tropical Indian and Atlantic Ocean. “All” shows mean of all ensemble members of the three systems. The 12-member F2-3DVAR system is recently upgraded to increase the ensemble size to 108-members “108mem” SINTEX-F1 system (Luo et al. 2005) We adopt the SINTEX-F1 atmosphere-ocean coupled general circulation model, which was developed under the European Union-Japan research collaboration. The SINTEX-F1 consists of the atmospheric component ECHAM4 and the ocean component OPA8. The ECHAM4 has the horizontal resolution of T106 (~100km) with 19 vertical levels. The OPA8 has the resolution of 2° Mercator mesh (enhanced to 0.5° in the latitudinal direction near the equator) with 31 vertical levels. The atmosphere and ocean components in the model interact every 2 hours via OASIS2 coupler without any flux corrections. SINTEX-F2 system (Doi et al. 2016) The SINTEX-F2 coupled model has been developed for better representation of several physical processes and to resolve relatively small-scale phenomena in the ocean. The atmospheric component, ECHAM5, has a horizontal resolution of 1.125° (T106) with 31 vertical levels. The horizontal grid used for the oceanic component, OPA9, is on the ORCA05 configuration, which has a horizontal resolution of about 0.5° × 0.5° with 31 vertical levels and without any further refinement over the tropics. The Louvain-la-Neuve Sea Ice Model, version 2 (LIM2) is embedded. SINTEX-F2-3DVAR system (Doi et al. 2017) This system is a upgrade version of the SINTEX-F2 system in terms of the ocean initialization. In this system, OGCM SSTs are strongly nudged toward the observations in the coupled run continuously from January 1982, which is similar to the simple SST-nudging scheme used in the F2-system. In addition, 3DVAR correction is conducted every 1st day of each month using subsurface ocean temperature and salinity observation. The set of in situ observations consists of all types of ocean profiling instruments that provide temperature and salinity (when available) from the expandable bathythermographs (XBTs), mooring buoys, sea stations, Argo floats, etc. The details of the 3DVAR scheme used here such as formulation and specification of observation and background error covariances are shown in Storto et al. (2014) SINTEX-F2-3DVAR 108-members ensemble system (Doi et al. 2019) The 12-member F2-3DVAR system is recently upgraded to increase the ensemble size to 108-members using the Lagged Average Forecasting (LAF) method. Based on this new system, we have conducted the prediction runs with a four-month lead-time from the eight initialized dates (1st-9th) of each month during the period from 1983 to 2019 (6-month lead time forecast is also available from some key months). SINTEX-F Family ( F1 + F2 + F2-3DVAR) system (Doi et al. 2020) |
 |
