|
|||||||
|
|
|
|
|
|
|
|
| |
 |
|
TIP 6 Science reports | |
|
|
|||||
 |
 |
||||
|
|||||||||||||
|
Science reports The ENSO Observing System and its Impact on ECMWF Seasonal Forecasting (D. Anderson, T. Stockdale, and O. Alves ECMWF)
At ECMWF a comprehensive coupled atmosphere-ocean model has been developed which is used to make seasonal forecasts on a global basis, where appropriate. Both the atmospheric and oceanic components of the coupled model are global. The atmospheric resolution is T63 in spectral space (approx 2 degrees in physical space). The ocean model resolution varies, being quite highly resolved in the meridional direction near the equator but less well resolved at high latitudes. Initial conditions for coupled integrations are obtained by forcing the ocean with the fluxes of momentum heat and fresh water from the ECMWF operational atmospheric analysis system, and relaxing strongly to the Reynold's SST fields. All subsurface ocean data are assimilated in order to produce an ocean analysis. This is done weekly in a 'quasi-operational' environment. From the ocean analysis, three coupled forecasts are made out to 6 months every week. A parallel analysis is also performed which is similar to the first except that sub-surface ocean data are not assimilated and three forecasts are made from that analysis too. It is necessary to have a significant ensemble size in order to sample the chaotic behavior of the coupled system. By having two sets of forecasts, from two differently produced ocean analyses, it is possible to sample uncertainty in ocean initial conditions to some degree and to assess the impact of the ocean observing system on forecasts. For the 1997 event the presence of TAO data has been very beneficial, both for analyses and for forecasts. Forecasts of the current 1997 event were shown. The research version of the model for which the forecast period was one year, predicted a warming of ~2K in the Niño-3 region from as early as Oct 96 but did not indicate the major El Niño which was to follow. However, the 'quasi-operational' version of the model with a 6 month forecast lead time correctly predicted the timing of the onset of the 97 ENSO from November 96. Although the model results are encouraging, there are also well-known deficiencies. Full, as opposed to anomaly, coupling is employed between the two components which allows some model drift. This is corrected a posteriori. Developments to the coupled system to improve the resolution in both atmosphere and ocean and to improve the atmospheric physics are in hand. Because of the large amplitude of the present El Niño, there is the possibility of making useful forecasts for mid latitudes and other parts of the world in which chaotic processes play a significant role. To extract a coherent signal from the chaotic background requires a significant ensemble size and so the ensemble size has recently been increased to one six month forecast per day, giving an ensemble size of around 30 per month. An assessment of the realism of the atmospheric response predicted by the coupled system is made difficult by the small number of years for which good quality initialization of the forecast system is possible (primarily because TAO has not been deployed for very long). Much more is known about the ability of uncoupled atmospheric models driven by observed SST. To try to relate our knowledge of uncoupled integrations to the coupled system, a parallel set of experiments has been run. In one set, a 19 member ensemble of coupled integrations was run four times a year in the period 1991 - 96. In a second set, a 19 member ensemble with an identical atmosphere model was run, with the same atmosphere initial conditions but using prescribed observed SST. Comparisons were shown of rainfall anomaly distributions predicted by the two model ensembles. Perhaps surprisingly, many aspects of the distributions were the same, suggesting that coupled model drift and anomaly forecast error were not excessively disrupting the predictable signals. Nonetheless, some of the differences could be highly significant on a regional scale, suggesting that the coupled system still needs substantial development. The relative importance of model drift and errors in the forecast SST anomalies has yet to be established. Impact of the TOGA Observing System on Ocean Initialization and ENSO Prediction (M. Ji, NCEP) Prediction of seasonal to interannual sea surface temperature (SST) variability in the tropical Pacific, primarily variations associated with the El Niño-Southern Oscillation (ENSO) phenomenon, has been a major objective of the climate research community. A major achievement from the recently completed 10-year international research program, tropical ocean global atmosphere (TOGA, 1985-1994) is the development of prediction capability for ENSO using dynamical coupled models and the deployment of the TOGA observing system to support the prediction efforts (McPhaden et al. 1997; National Research Council, 1996). The NCEP coupled model is a coupled general circulation model. It attempts to takes full advantage of the availability of realtime in situ and remotely sensed observations for the tropical Pacific to initialize the ocean component of the coupled system in order to achieve useful skill for El Niño forecasts. In this study, we carried out three parallel hindcasting experiments using the NCEP coupled ocean-atmosphere forecast model. The three experiments differ in the method used for ocean initialization. The initialization for the first experiment, denoted as SST, is accomplished by running the coupled model in a simulation mode for 1981 1996. Only observed SST are inserted into the model during the simulation. The initialization for the second experiment, denoted as XBT, is similar to the SST experiment, however, observed subsurface temperature data from XBTs and TAO moorings are assimilated into the coupled model during the simulation. In both cases, no observed surface wind stress is needed because the coupled model predicts the wind stress forcing for itself during the simulation. The third experiment, we denote it as ASM, uses the standard NCEP ocean data assimilation system which assimilates observed temperature data into an ocean GCM driven by observed wind stress forcing for ocean initialization. One year hindcast experiments are carried out using the same coupled forecast model with the three different sets of ocean initial conditions. Due to the computational intensive nature of the coupled general circulation model, only four hindcasts for each year are run for the period of 1981 - 1996. The hindcasts are initiated in January, April, July and October in order to take into account of some seasonal dependency of prediction skill. The skills for hindcasts are evaluated for two separate period: 1981 - 1989 and 1990 - 1996. Since the hindcasts are one year in length, the corresponding two periods for validating the predictions are 1981 - 1990 and 1990 - 1997. Evidence from comparison of skills for the two periods suggests that for the 1980s, comprehensive ocean data assimilation provided very little improvement to the prediction skill. However, for the 1990s, with the full ocean data assimilation, i.e., utilizing observed winds, SST and subsurface temperature data, the prediction skill achieved by the ASM is much higher than that for just SST. The 1980s is dominated by several strong ENSO episodes (The warm episodes of 1982-83, 1986-87, and the cold episode of 1988-89). During this period, the strong oscillatory ENSO mode prevailed. A coupled model that captures the dominant quasi-periodic oscillation of the coupled climate system such as the quasi-biennial and quasi-quadrennial mode (Jiang et al. 1995) potentially is capable of making skillful predictions. This is probably the reason for many simple coupled models to be successful in hindcasting ENSO episodes of the 1980s (Ghil and Jiang, 1997). However, in the present decade, obvious changes in characteristic of ENSO development have been observed (Ji et al. 1996; Trenberth and Hoar 1996; Goddard and Graham 1997). The change in ENSO characteristics can be linked to the weakening of the dominant quasi-oscillatory interannual mode represented by the leading two sea level anomaly EOFs during the 1990s (not shown). Therefore, impact of additional modes of variability in the coupled climate system such as interdecadal mode (Zhang and Wallace 1997; Latif et al. 1997), and their interaction with the ENSO mode may be significant, and need to be taken into account in order to achieve skillful ENSO forecast. What needed is forecast models with complex physics and dynamics coupled with observing system and ocean initialization scheme capable of capturing all relevant scales of variability. References Actual and Potential Use of TAO Data at Meteo-France (S. Planton, Meteo France) The TAO data are currently or have recently been used by the Meteo-France operational and research services. From the operational side, the buoy wind and sea surface temperature are both accounted for in the data assimilation system for weather forecasting but with various sampling according to the network time. The TAO and meteorological data have also been used in conjunction with drifting buoy measurements and satellite data collected during the TOGA-COARE experiment. The objective was to evaluate the surface heat budget along the drifting buoy trajectory through the assimilation of the oceanic data in a one-dimensional ocean model (Roquet et al., 1993). In the near future, the PIRATA data completing the network for the tropical Atlantic are also expected to be incorporated in the operational data processing for weather forecasting. They will also be used to validate the operational oceanic simulation performed in the tropical Atlantic (OPERA) with a version of the LODYC model. The SSTs will also be used to validate the products of the "Satellite Application Facility" of ocean, initiated by EUMETSAT and piloted by Meteo-France. Another application will be the validation of oceanic temperatures and currents of oceanic models integrated for climate studies or seasonal forecasting. An important component of the moored array network will also be the evaluation of the ocean-air heat and momentum exchanges. After validation through a dedicated ocean field experiments (Equalant 1999), this data set will be incomparable to validate the fluxes at different scales. At the larger scale coupled or uncoupled atmospheric model surface fluxes will be evaluated, as well as products from the ECMWF reanalysis. At the mesoscale, the variability of the fluxes in relation to boundary layer variability both in the ocean and atmosphere will be investigated. As an illustration, the Figure 6 reproduces the sensible heat flux, latent heat flux and friction velocity along the trajectory of the research ship Le Suroit, during the Intensive Observing Phase of the SEMAPHORE ocean field experiment (Eymard et al, 1996). It compares the evaluation of the fluxes through the inertial dissipation method (star symbol) and through adjusted bulk formulae applied to the ship observations (Dupuis et al, 1997). Fluxes from a reanalysis performed with a stretched version of the ARPEGE operational model centred on the experiment domain are also shown. References: Toward Determining the Sensitivity of the ENSO Coupled System to Perturbations (R. Kleeman, BMRC/IRI) A mathematical formalism (currently being widely applied to intermediate ENSO models) to determine the sensitivity of the tropical Pacific coupled system to perturbations was outlined (singular vectors). It was shown that results are currently highly model dependent and it was suggested that this was mainly due to the differences in the atmospheric components used. Using a particular model it was shown that perturbations resembling aspects of the Madden Julian Oscillation were highly efficient in perturbing the system. The observed irregularity of ENSO was very well reproduced by such a stochastic model. It was argued using operational predictions from a coupled model that the amplitude (but not phasing) of the current very large warm event could be accounted for by a stochastic input from two very strong westerly wind bursts in the March April 1997 time period. Many coupled models were able to predict that a warming would occur in 1997 at lead times of 9-12 months but were unable to predict its extreme magnitude. This analysis provides a possible explanation for this. Use of the TAO Buoy Data by the U.S. Navy (P. Phoebus, FNOC) Marine Meteorology Division Naval Research Laboratory Monterey California Observations from the TAO buoy array in the equatorial Pacific are used by the U.S. Navy for both operational meteorology and oceanography and for research purposes. The Navy has a high interest in accurate analyses and forecasts of marine surface winds around the globe, since many Navy operations take place near the air-sea interface. Furthermore, many areas of tactical or operational exercises take place in the tropics or sub-tropics, and the Navy centers in Guam and Hawaii have forecast responsibility for Northern Hemisphere tropical cyclones in the Pacific that often pose a threat to Navy assets. TAO data have been used by the Navy for a number of validation studies. Data denial studies have identified systematic biases in the Navy's global atmospheric prediction system, and the impact of the TAO wind observations compared to the SSM/I wind observations have pointed to likely problems with the SSM/I data in the tropics due to water vapor contamination. TAO data have been used to substantiate the skill of the global atmospheric prediction system in analyzing and forecasting the development of tropical cyclones near the equator, thereby increasing the Navy's confidence in the global model as a tropical forecasting aid. The Navy also routinely produces a number of oceanographic products. Accurate wind forcing is needed for a variety of ocean models, including dynamical and mixed-layer models, ice, and wave models. Buoy measurements of ocean temperature provide a valuable source of data for the Navy's twice-daily, three dimensional global ocean temperature analysis, which in turn provides a timely and accurate lower boundary condition for the Navy's global and regional atmospheric models. This analysis can be critical to the Navy model's tropical cyclone forecasting skill, and the TAO buoys are a valuable additional data source in the Pacific, where tropical cyclones frequently develop within 5 or 10 degrees of the equator. Because of the Navy's high interest in the tropical Pacific, there have been a number of research projects that have used the TAO data. These projects encompass both observational studies, which increase our understanding of tropical air-sea interactions, and modeling studies, where the TAO data are most often used for verification of the modeling results. Several of these research projects will be discussed. Comparison of NCEP/NCAR and ECMWF Reanalyzed Fields with TOGA-TAO Buoy Observation (B. Smull, U. of Washington) This study seeks to assess the quality of reanalyzed surface fields (especially winds) compiled over the TAO array during the period 1991-1993. Our broader objective is to quantify these differences in conjunction with other observed quantities (e.g., temperature and humidity) as a function of season and climatic regime, and moreover to point toward needed improvements in operational global data assimilation and modeling systems as applied over the tropical oceans. This evaluation is performed through comparison of pentad-, monthly- and annual-mean surface observations derived at each of ~65 sites comprised by the TAO array to corresponding mean conditions derived from both the NCEP/NCAR and ECMWF reanalyses, which are available at a frequency of 4 per day and spatial resolution of T62/L28 and T106/L31, respectively. NCEP/NCAR zonal and meridional wind components were obtained from the "Gauss Grid" (effective equatorial resolution ~2° x 2°) obtained from NOAA/CDC, while analogous ECMWF quantities at a resolution of 1.25° x 1.235° were provided courtesy Dr. David Anderson of ECMWF. Figure 7 summarizes
annual-mean winds and TAO-reanalysis wind differences for 1993; largely
similar results were evident in 1991-1992 (not shown) but were restricted
spatially owing to the reduced coverage of the TAO array during those
earlier years. NCEP reanalyzed and TAO observed wind vectors are overplotted
in panel (a), while corresponding vector differences are shown in (b).
In (a) we see that directional differences of 20-30 deg are common, though
these reach nearly 90° near the ITCZ. In general, NCEP winds are
weaker and more zonal (i.e., less divergent) than those observed by TAO.
Vector differences in (b) form a pronounced divergence/convergence pattern
in association with the ITCZ/cold tongue complex in the east Pacific basin.
Magnitudes of these differences were generally 1 2 m s-1 (peaking near
3 m s The Value of Well-Instrumented Surface Moorings (R. Weller, WHOI) Data from the recent deployment (October 1994-October 1995) of a surface mooring in the northwestern Arabian Sea (15.5°N,61.5°E) is used to illustrate the value of deploying a well-equipped surface mooring. This mooring was equipped with both a Vector Averaging Wind-Recorder (VAWR), sampling every 7.5 minutes, and an Improved Meteorological Recorder (IMET), sampling every 1 minute. Wind speed and direction, air temperature (aspirated and non-aspirated), relative humidity, incoming shortwave radiation, incoming longwave radiation, sea surface temperature (floating sensor), barometric pressure, and precipitation were measured. Pre- and post-deployment calibrations and in-situ intercomparisons between shipboard and buoy sensors conducted for a day just after and just before deployment were used to ensure the quality of the meteorological measurements; and the air-sea fluxes were computed using the COARE flux algorithm. These high sampling rate, high quality surface meteorological and air-sea flux time series have allowed us to: 1) examine the validity in the Arabian Sea of the available climatological fields of the surface meteorology and air-sea fluxes, 2) investigate the extent to which numerical weather prediction models including NCEP, ECMWF, and FNOC, fluxes in the Arabian Sea, 3) demonstrate the improvements made by Josey and Taylor in developing the new Southampton Oceanography Center (SOC) climatology, 4) select (in collaboration with Dave Halpern) coincident, collocated data from satellite wind sensors and examine the validity of that data as processed by several algorithms, and 5) develop a new understanding of the physics of the atmosphere-ocean coupling during the monsoons. The buoy observations
have an annual mean heat flux of 60.3 W m Comparison of concurrent (1994-1995) data from the SOC climatology and the buoy shows good agreement, substantiates the improvements made to that ship-based climatology, and supports wider use of the SOC climatology in the Arabian Sea as a better choice for forcing fields than either previous climatologies or fields from numerical weather prediction models. This allows us to use the SOC climatology to provide spatial coverage over the Arabian Sea during the ONR/JGOFS projects in 1994-1995. It also allows us to use the SOC data set from 1980-1995 to establish that 1994-1995 was a typical year. Further comparisons of well-instrumented surface moorings (the Subduction Experiment, TOGA COARE, the Pan American Climate Study) with climatological and model fields have been and are being carried out. These studies show that the present generation of modern buoy sensors, which can also be deployed on Volunteer Observing Ships, obtain surface meteorological and air-sea flux data of high quality that clearly identify the shortcomings of the climatologies and model flux fields and provide an accurate record of local air-sea coupling. Thus, it is suggested that the TIP consider deploying a number of well-instrumented moorings for surface fluxes within the TAO array. Interannual Variability of the High-Salinity Tongue South of the Equator at 165°E (W. Kessler, NOAA/PMEL) In 1996, the ENSO forecast group at NCEP noticed an apparent contradiction between the west Pacific sea surface height found in their model assimilating TAO data and that found from the TOPEX altimeter. The difference was as large as 9 dyn-cm, a significant fraction of the total variability. It turns out that the difference was largely due to salinity changes extending over the upper 200 m of the water column, and these were not accounted for by the model assimilation of TAO temperature-only data. A nearly 1 psu increase in salinity between 1995 and 1996 occurred in the upper 50 m, which was not surprising due to the change in rainfall associated with the waning of the El Niño of 1994-95, but there was also a 0.4 psu increase of salinity in the high-salinity tongue at about 150 m depth. The high-salinity tongue is a layer of water moving westward and equatorward from the surface near 10-20°S in the eastern Pacific as part of the southern subtropical gyre circulation. 165°E is one of the few places in the Pacific where time series of subsurface salinity can be reasonably constructed back to about 1984, based primarily on CTDs taken during the France-U.S. and U.S.-PRC hydrographic programs, and TAO deployment cruises. Time series at the level of the salinity maximum south of the equator showed low-frequency variability with fluctuations of 0.2 to 0.4 psu, most strongly near 4°S - 8°S. This variability was not well correlated with vertical motion of the thermocline or with the ENSO cycle. Instead, the variations of salinity in the high-salinity tongue at 165°E appear to be controlled by zonal advection in the SEC, and variations of both zonal current and zonal salinity gradient contribute to the signal. The changes described here show that subsurface salinity can be an important element in the variability of dynamic height. While the TOPEX altimeter can be used to correct the model dynamic height through assimilation, without salinity observations it is not straightforward to know how to distribute the differences in the vertical. However, those choices impact the assimilated product current and density fields. Therefore it would be desirable to improve our understanding of the relation between surface and subsurface salinity. Since CTD sampling in this region has fallen off with the termination of the French New Caledonia deep-sea research effort, the present level of cruises (now essentially just the TAO deployment cruises) this can only be accomplished through subsurface salinity sampling on TAO moorings. The variability described here suggest that 5°S, 165°E would be a good location to sample this signal, and that instruments need to be placed down to 200 m and include a sample at 150 m. Effects of Salinity and the Indonesian Throughflow on the Tropical Indo-Pacific Basin (R. Murtugudde, U. Maryland) State of the art
heat flux formulations have reduced the errors in SST simulations by OGCMs
to an extent where the remaining errors that can not be explained by uncertainties
in the surface fluxes must be attributed to model errors. However, the
discrepancies between various precipitation products are too large which
makes it difficult to blame OGCMs for poor simulations of salinity fields.
With the high accuracy precipitation products on the horizon, need for
salinity data for validation of OGCMs is evident. Salinity contributes
to large scale dynamics and thermodynamics through density and pressure
gradients. Even a small bias in precipitation of 1 m yr Seasonal-to-interannual (1980-1995) Indonesian throughflow (ITF) simulations are performed with a primitive equation, reduced gravity, sigma-coordinate model. An ENSO related reduction in the ITF is clearly evident which is driven by winds over the Pacific. However, winds over the Indian Ocean generate a significant non-ENSO related signal. The ITF cools the Pacific and warms the Indian Ocean. The seasonal cycle of SST in certain regions such as the equatorial Pacific and the Leeuwin Current are shifted whereas in other regions such as coasts of Java and Somalia, the ITF signature is prominent during the upwelling season. The winter rainfall deficit over Australia may be enhanced by the ITF as seen be its influence over the central Indian SST gradient. Difference of sea level across the Indonesian channel shows a high correlation with ITF indicating that an index may be defined based on TOPEX. Short-Term Climate Variability in Taiwan and the Tropical Atmosphere-Ocean (H.-H. Hsu, National Taiwan University) This study calculates EOFs of SST in the Pacific, Indian, and Atlantic Oceans, separately. The first EOF of the three oceans, which exhibit the characteristics of the El-Niño/Southern Oscillation (ENSO), are closely correlated, with the Pacific leading the Indian Ocean and the Atlantic by 4 months and 6 months, respectively. The first EOF in the Indian Ocean appears to have the strongest correlation with the monthly-mean temperature and precipitation in Taiwan. The major features are the above-normal (below-normal) precipitation in March and the above-normal (below-normal) temperature in September following the warm (cold) years. The increase of precipitation in March following the warm years is associated with a deepened trough and strong baroclinic wave activity in East Asia. The case of March 1983 was investigated using a GCM. Results of the simulation suggest that the warm SST may have caused the abnormal circulation and precipitation. The above-normal temperature in September is apparently associated with the anomalous anticyclone in South Asia that expands more eastward than normal to cover the east coast of East Asia. The reason for the displacement of the anticyclone is not clear. However, it is strongly correlated with the warm SST in the Indian Ocean several months before. In addition to the influences of the tropical ocean-atmosphere, the interannual variations of the summer temperature and precipitation in Taiwan are also associated with the variations of the SST in the extra-tropical Pacific. Anomalous circulation, precipitation, and SST of the 1993 and 1994 summers were investigated to identify the relationship. Surface Oceanic Eastward Jet and Its Relation With the Westerly Wind Bursts in the Western Equatorial Pacific: Use of Satellite Scatterometer Data (Y. Kuroda, JAMSTEC and K. Kutsuwada, Tokai University) Current data obtained
by upward-looking moored Acoustic Doppler Current Profilers at two equatorial
stations (142°E and 147°E) since 1994 are used to detect events
in which eastward flow stronger than 50 cm s Data sets of daily
surface wind on 1° x 1° grid in the tropical Pacific region
are constructed using data of satellite scatterometers (NSCAT and ERS-1).
In all the periods in which the surface eastward jet occurred, the westerly
winds stronger than 5 m s Penetration of Visible Light in the Upper Ocean Heat Budget (M. Lewis, Dalhousie University) Solar radiation in the visible portion of the spectrum, which represents approximately half of the global downwelling flux, penetrates to a varying degree to heat the ocean interior by absorption. The degree to which these visible wavelengths penetrate depends on the optical properties of the water itself, and the particulate and dissolved constituents; most of the variability is due to variations in the concentration of phytoplankton or algae in the upper ocean. For the tropical
Pacific, and for the western Warm Pool in particular, a significant energy
flux is represented by the penetration of visible radiation through the
shallow mixed layer which actively interacts with the atmosphere. On a
climatological basis, Lewis et al. (1990) demonstrated that most of the
computed net surface heat flux of order 20-40 Wm Over the last six years, such a situation has been observed twice. Siegel et al. (1995) observed high pigment blooms in the western Pacific during TOGA-COARE, and Lewis et al. observed such high concentrations in early 1997. The coincidence with the El Niño may not be fortuitous. Pigment concentrations also vary strongly in the eastern Pacific and in general are depressed during El Niño, permitting greater penetrative irradiance fluxes. Given this strong variability, and given the magnitude of the heat fluxes, it will be useful to monitor the ocean optical properties more closely, on time and space scales consistent with the other important contributors to the upper ocean heat budget. Recently, a buoy deployment has been carried out by MBARI and PMEL where optical sensors were placed on TAO buoys with independent data transmission. Initial results from these deployments in the eastern Pacific in 1997 show derived pigment concentrations approximately half of those seen in climatology, with a consequent increase in the penetration depth of visible irradiance. Future deployments of such sensors, particularly in the sensitive western Pacific, should be strongly encouraged, in order to provide constraints on the magnitude of the upper ocean heat budget, as well as for critical biogeochemical investigations. References: |
|||||||||||||
| Recommendations >> |
|
|
| Home | Project overview | Data display | Data delivery | El Niño & La Niña | Site map |
|
|
TAO
Project Office |