6.1 Experimental monitoring products and climate forecasts

CDC research has lead to the development of a number of experimental climate monitoring and forecasts products, which are systematically updated and evaluated along with operational products. Collectively these efforts provide the scientific research community and non-traditional users access to enhanced climate information, current climate conditions, and climate forecasts.

6.1.1 Value Added Diagnostics and Visualization Products for Weather and Climate Monitoring

The CDC WWW-based maproom (http://www.esrl.noaa.gov/psd/map/) provides a focal point for monitoring and prediction products on climate variability, climate impacts and climate-weather connections. The CDC maproom encompasses a broad range of diagnostic products and enhanced visualization tools for weather and climate monitoring. A vast majority of these diagnostic and value-added visualization products were designed initially to support in-house research. These are now active research areas themselves, with goals being to develop enhanced climate and weather monitoring products and to evaluate operational and experimental climate and weather forecasts. The core of the maproom consists of analyzed and predicted fields of climate variables that are updated on a daily basis, and that provide the raw data for both conventional and experimental products. The maproom is also a host site to enable a similar scope of studies by non-CDC scientists.

The maproom supports CDC's research, diagnostic and assessment activities by serving as: a) a testbed for new products related to climate monitoring and prediction; b) a means to interact with users of climate and weather information; c) a way to familiarize climate researchers with ongoing climate anomalies; and d) a tool to interpret the current state of the climate with emphasis on coherent modes of variability such as ENSO, MJO, NAO, etc. Weekly maproom briefings provide a forum to discuss the maproom products and to monitor the performance of new products.

Several maproom improvements have been undertaken since 1997. Some of these have involved streamlining scripts, consolidating products, and ensuring timely access to data streams. Specific improvements and additions related to maproom content include:

  1. Applying Java animation scripts to climate variables for time averages from daily to seasonal,
  2. Use of the NCEP reanalyses as the basis for global atmospheric products rather than the NCEP operational data,
  3. Monitoring coherent modes of tropical OLR variability, including the MJO,
  4. Monitoring storm tracks based on a sea level pressure algorithm (storm track data from 1958-present also available),
  5. Prediction of seasonal anomalies based on the historical response to SSTs in an ensemble of AMIP runs,
  6. Monitoring atmospheric angular momentum budget for intraseasonal and longer term variations,
  7. Monitoring the states and ensemble predictions of Northern Hemisphere teleconnection patterns,
  8. Implementing 500 mb height anomalies and regional plots over North America from the NCEP ensemble predictions,
  9. Implementing probabilistic Week 2 predictions from the NCEP MRF/AVN ensemble.

CDC's interest in attribution of climate anomalies, and comments from users of climate information (including EMC and CPC), helped drive the development of many of these products. Development of new methods in climate prediction within CDC, on both weekly and seasonal time scales, has resulted in an increase in experimental prediction products for the maproom.

Web usage statistics for the maproom show that the NCEP MRF ensemble predictions to 15 days are among the most popular products. CDC displays both conventional variables (e.g., 500 mb geopotential height) and experimental products (e.g., forecast probability of a 1 or 2 standard deviation anomalies in 850 mb temperature) in the maproom. There is substantial interest in these products from national energy companies which monitor long-range weekly forecasts. In the future, CDC will continue to develop new products that will help identify the variability and trends in the earth's climate. User interactions will help CDC identify existing products and how they may be improved.

6.1.2 Multivariate ENSO Index (MEI)

In addition to providing easy access to interannual indices and maps of climate variability, CDC provides a near-real time monthly updated Multivariate ENSO Index (MEI) based on six COADS atmosphere-ocean variables over the entire tropical Pacific basin: sea-level pressure (P), zonal (U) and meridional (V) components of the surface wind, sea surface temperature (S), surface air temperature (A), and total cloudiness fraction of the sky (C). One week after the end of each month, the MEI is extended based on near-real time NCEP marine ship and buoy observations summarized into COADS-compatible 2-degree monthly statistics. During the summer of 1997, the MEI was put on the WWW and quickly gained popularity as a monitoring product with monthly updates on the 1997-1998 El Niño (Fig. 6.1). In the past five years, the time series of the MEI has been distributed to hundreds of scientists and many other interested parties, and included in many publications, both scientific and popular.

Evolution of El Niño according to the MEI

Fig. 6.1 The seasonally normalized MEI documents the evolution of individual El Niño (shown here) and La Niña events, allowing for direct intercomparison of the relative strength of ENSO events since 1950 (http://www.esrl.noaa.gov/psd/~kew/MEI/).

6.1.3 Linear Inverse Modeling Forecasts of Tropical Sea Surface Temperatures

The development of a linear inverse modeling (LIM) approach to generate low-frequency climate forecasts is a prime example of CDC basic climate research directly contributing to the NOAA goal of providing new and improved climate forecasts. As a service to the scientific community, since the mid 1990's CDC has produced LIM tropical SST forecasts and made the forecasts available on the WWW and in the EFFLB along with appropriate verification statistics. CDC provides monthly predictions of tropical Indo-Pacific sea surface temperature anomalies (SSTAs) with 3, 6, 9, and 12 month lead times using a linear inverse modeling procedure with tropical Indo-Pacific SSTAs as predictors (Fig. 6.2). Anomalies are calculated relative to the standard 1950-1979 COADS climatology. CDC also produces similar LIM monthly predictions of northern tropical Atlantic Ocean and Caribbean sea surface temperature anomalies (SSTAs) with 3, 6, 9, and 12 month lead times. In contrast with tropical Indo-Pacific forecasts, these predictions are generated using global tropical SSTAs as predictors and anomalies are calculated relative to the COADS 1950-1993 climatology.

Linear inverse model predictions of Niño 4 SSTA

Fig. 6.2 Predictions of Niño 4 SSTA (solid blue line) and verification (solid red line). Dashed lines indicate one standard deviation confidence intervals appropriate for the LIM forecast (as opposed to those based on the history of standard deviation of previous forecast errors).

6.1.4 Subseasonal LIM forecasts of Tropical heating and extratropical circulations

For over a year, CDC scientists have used LIM to make medium and extended range (Weeks 2 to 6) forecasts of weekly average anomalous 250 and 750 mb streamfunction and tropical diabatic heating (Fig. 6.3). The procedure (described in Chapter 2) is similar to the method used to predict tropical Indo-Pacific and Atlantic/Caribbean sea surface temperature anomalies, except that streamfunction and diabatic heating are the model variables and the time-scale is weeks instead of months. NCEP MRF model forecasts at Week 2 and current verifications are provided for comparison when available (Fig. 6.3). The CDC wintertime and summertime forecasts are currently made out as far as six weeks ahead. A second forecast, the Combined forecast, is made by combining the LIM technique with the NCEP MRF ensemble mean Week 1 streamfunction forecast. Throughout the winter (Dec. 1-Feb. 28) and summer (Jun. 1-Aug. 31) these forecasts are updated daily. Much of the skill in these LIM forecasts, particularly at the extended range, is not only due to anomalous tropical convection related to ENSO but also to the extratropical response to the MJO. Advances in this experimental climate services forecast product will depend both on continued model development and evaluation and on enhanced monitoring and prediction of MJO activity.

Observations and predictions of 250 mb streamfunction pattern for Feb. 4, 1998

Fig. 6.3 Top: Observed 7-day running mean 250 mb streamfunction pattern for February 4, 1998. Middle: Week 2 linear inverse model forecast of 7-day running mean 250 mb streamfunction verifying against top panel. Bottom: Week 2 MRF forecast verifying against top panel.

6.1.5 Other experimental forecast products

CDC provides monthly experimental CCA forecasts of tropical SSTs, precipitation, 500mb height, and surface temperature for the subsequent 4 overlapping 3-month season (Fig. 6.4, http://www.esrl.noaa.gov/psd/~gtb/seasonal/). Predictions of tropical SSTs are derived from 4 different sources: (1) an inhouse CDC canonical correlation analysis (CCA) model, (2) the NCEP's Climate Modeling Branch (CMB) SST forecast, (3) the International Research Institute's (IRI) forecast of tropical SSTs, and (4) the CDC LIM Indo-Pacific SST forecast. Each tropical SST forecast, plus their linear average, are used as boundary conditions for predicting seasonal climate anomalies. A second CCA model relates the first 28 EOFs of tropical SSTs to a similar 28 EOF basis of an atmospheric predictand. The predictands are seasonal tropical precipitation anomalies, Pacific-North American seasonal temperature and precipitation anomalies, and Pacific-North American seasonal 500-mb height. The predictand data are derived from ensemble simulations from four atmospheric GCMs (NCEP MRF9, ECHAM3, CCM3, GFDL-R30) run for 1950-99 and forced with the monthly varying global SSTs. The atmospheric prediction is made by first projecting the predicted tropical SST anomalies onto the 28 EOF SST basis set for 1950-1999. The CCA model, based on zero-lag relations of this SST and the individual atmospheric predictands derived from the GCMs, is then used to predict seasonal precipitation anomalies and atmospheric circulation.

CDC experimental June 2001 forecasts for
July-August-September 2001 and October-November-December 2001.

Fig. 6.4 CDC experimental June 2001 forecasts for July-August-September 2001 and October-November-December 2001.

6.1.6 Week 2 predictions and the atmospheric angular momentum budget

CDC is working with the Dodge City (KS) NWS Office to link weather and climate phenomena, and to anticipate errors in Week 2 predictions in various forecast models. This collaborative effort contributes directly to the "Special 2 week weather outlook" (http://www.crh.noaa.gov/ddc/wx/2week.txt) issued routinely by the Dodge City NWS Office. These Week 2 outlooks are produced using probability distributions of atmospheric circulation and other variables from Week 2 ensemble model forecasts to develop initial discussions of Week 2 conditions. The initial discussions are then diagnosed and modified to account for model biases using a suite of WWW-based products, including CDC maproom atmosphere angular momentum (AAM) and outgoing longwave radiation (OLR) products, that provide additional information on antecedent and existing conditions.

Synoptic experience indicates that low and high global AAM influences the variability of blocking and Rossby wave dispersion over the Pacific basin region.The Week 2 discussions use such synoptic insights to interpret the ensemble predictions and to determine the likelihood for systematic errors during particular regimes. Cases of good or bad Week 2 forecasts over the Pacific basin region are selected for more intensive study and diagnosis, always within the context of the phase of ENSO, the MJO, the seasonal cycle and Rossby wave dispersion. Additionally, in the last two winters Week 2 predictions of the Pacific basin circulation anomalies by the linear model discussed in Section 6.1.3 have provided unique insights into the role of tropical convection in forcing the extratropics. Aspects from an interesting case of this past northern winter are discussed below.

Figure 6.5 shows that during the last year ~50-70 day oscillations in global relative AAM have been prominent. These oscillations started in July 2000 and continue to the present (July 2001). They are superimposed on a persistent negative AAM anomaly (-1.5 sigma), reflecting weakened subtropical westerlies. The weakened westerlies coincide with cool (warm) sea surface temperatures in the tropical eastern Pacific (Indonesian region) and all have been present since ~August 1998. This easing of the subtropical westerly flow during the last 2 years opposes the observed trend in the NCEP/NCAR reanalysis during the last 30-40 years. The intraseasonal oscillations are closely related to the tropical convective forcing of the Madden-Julian oscillation, although not exclusively. Mid-latitude processes related to momentum transports, eddy feedbacks and mountains also force anomalies in global and zonal AAM and produce additional variability in the time series.

Time series of angular momentum

Fig. 6.5 (Top) Time series of the vertically and zonally integrated relative atmospheric angular momentum from July 2000-July 2001. (Bottom) Global atmospheric angular momentum anomalies since July 2000.

A prominent feature of the 2000-2001 northern winter, seen in the top panel of Fig. 6.5, is a large zonal mean anomaly pattern in the Northern Hemisphere. Starting in mid-November 2000, strong westerly flow develops in mid-latitudes (30-50N) with weaker than normal westerly flow in adjacent latitude bands. The easterly anomalies in the subtropics intensify and help drive global AAM to very low values during the winter period. Minimum values are reached in early February 2001, and are followed by a rapid rise to positive values during February 2001. Two physically significant events precede the rise.

In late January 2001, a strong MJO organized over the Indian Ocean and moved eastward reaching the dateline around 1 March 2001 (not shown). The evolution of the zonal AAM, seen in equatorial regions of Fig. 6.5, is consistent with the composite MJO. As positive convection anomalies move east across the ocean warm pool, zonal mean west wind anomalies develop on the equator and propagate poleward into the subtropics. MJO-related positive frictional and mountain torques contribute partially to the global AAM rise.

At about the same time (late January 2001) a subtle change occurs in the zonal mean flow at ~55N as the easterly zonal wind anomalies transitioned to westerly anomalies. This subtle change initiates a chain of events that is represented synoptically by the composites in Fig. 4.10. Stronger westerly flow over the mountains at 55N produces topographic Rossby waves that move negative mass anomalies equatorward. Simultaneously upper level wavetrains amplify and disperse over the mountains and contribute to meridional momentum and heat transports. The resulting upper level momentum sinks in the subtropics tend to force easterly wind anomalies at the surface, i.e., a positive frictional torque, contributing another portion to the rise in global AAM. We suspect this is typical behavior for intraseasonal (20-90 days) AAM variations during northern winter, i.e., tropical MJO-induced forcing and extratropical transport-induced forcing both contributing to AAM changes.

Back | Forward