TOGA COARE International Project Office
University Corporation for Atmospheric Research
Boulder, Colorado USA
The combination proved extremely fruitful and led to the:
Furthermore, researchers used the occasion to construct a possible framework for future coordination of COARE science group efforts, assess explicit progress towards COARE goals, and alert funding agencies to the level of effort required to capitalize on the substantial investment in the COARE field program. Experience at the workshop shows that COARE investigators are accepting the challenges posed by the interdisciplinary aspects of the COARE data and that it is reasonable to expect many new insights into climate processes to emerge from cross-disciplinary studies of the COARE dataset over the coming years.
Preliminary descriptions of the "state-of-the-art" in COARE research are given below, as seen by leaders of the five major science groups at the workshop: Large-Scale Atmosphere, Mesoscale Convection, Air-Sea Fluxes, Oceans, and Coupled Phenomena. Assessments of progress toward COARE goals, future research and product development, and cross-cutting science issues are also given.
However, basic work on the data (e.g., quality control, conversion to common data file formats) is still proceeding, and the various steps toward integrating datasets within and across COARE science working groups will continue for some years to come. The COARE science groups each prepared schedules (see Section 6) for achieving COARE science goals and concluded that it will take a number of years to complete the scientific analysis required to meet the program objectives, perhaps through the year 2000. Over these years, scientific guidance and oversight of investigators' efforts and of the supporting infrastructure will be required to derive the maximum value from the data collected during the field phase.
Mindful of these considerations, the COARE Science Group (CSG), which was responsible for the scientific organization and oversight of the workshop, recommends:
An ad hoc group consisting of the CSG and members of the international TOGA COARE Panel present at the workshop met several times in Toulouse to consider (among other matters) the promotion and management of research using the COARE dataset over the next few years. The group agreed that research on the COARE dataset will have major impacts on our understanding of a number of vital climate issues, and that intensive research on the dataset is likely to continue for at least five more years. However, the dataset is complex, and interpreting each part of it has its own technical problems that require special skills. The computer network and software tools provided by TCIPO at the workshop contributed significantly to enabling scientists to explore the huge variety of datasets available (albeit often in provisional form). This in turn has facilitated cross-disciplinary linkages: TCIPO staff deserve credit for their rapid progress in support of the assembly of high-quality datasets across the spectrum of COARE science, and we will be strongly dependent on their expertise in the continuing scientific effort.
At the same time, this very speed and effectiveness of data handling introduces the danger that data errors (several were detected during the workshop) may propagate through the literature in subtle and hard-to-detect forms. If final COARE research products are to maintain the high standard achieved during the field phase, particularly in the crucial cross-disciplinary areas where the possibilities for misinterpretation are greatest, the close links developed at this workshop among the COARE scientific community, the COARE Science Group, TCIPO, and the COARE Panel must be preserved and fostered. A mechanism to maintain these links requires that formal COARE advisory groups continue.
The original terms of reference for the TOGA COARE Panel were:
The COARE Science Group is willing to continue to keep track of TOGA COARE and to assist TCIPO over the next few years in its task of preparing a "final" COARE dataset that is quality controlled in an integrated way. However, it is desirable that this group (or some subset of it) be formally reconstituted as the COARE Panel, with responsibility to a parent body such as CLIVAR. The responsibilities of the COARE Panel should be as follows:
Since this document may be an introduction to TOGA COARE for some readers, it is worth outlining the program's history. In 1982, an international group of scientists called Tropical Ocean Global Atmosphere (TOGA) was formed under the joint sponsorship of the World Meteorological Organization (WMO) and International Council of Scientific Unions (ICSU). The aim of TOGA was to test the hypothesis that the warm tropical ocean is coupled to the global atmosphere, and that this coupling is a major influence on interannual climate variability. During the TOGA decade (1985-1994), the first systematic observations of the tropical oceans were undertaken, at first through deployment of expendable bathythermographs (XBTs) from merchant ships; later the Tropical Atmosphere Ocean (TAO) mooring array was added. Modeling analysis activity took place at prediction centers around the world.
While this activity led to considerable improvements in our understanding of climate variability (especially of ENSO phenomena), the TOGA SSG saw that progress in modeling climate variability was limited by lack of detailed knowledge of the atmosphere-ocean dynamics in and above the regions of warmest sea surface temperature (SST). In particular, the equatorial westerly winds that are believed to lead to ENSO events are not steady, but occur in bursts, and much too little was known about what caused them. They appeared to be very sensitive to small changes in SST, but little was known of what caused these small SST changes. Therefore, Professor Peter Webster (Chairman of the TOGA SSG) and Professor Roger Lukas took the initiative to develop plans within the United States for a major process study in the western equatorial Pacific, involving both oceanographers and meteorologists. Following the very successful workshop on TOGA COARE in Noumea, New Caledonia in May 1989 (Proceedings of the Western Pacific International Meeting and Workshop on TOGA COARE, eds., J. Picaut, R. Lukas, T. Delcroix, ORSTOM, Noumea, New Caledonia), COARE was made into an international effort. Detailed science plans were formulated (World Climate Research Programme, Scientific Plan for the TOGA Coupled Ocean-Atmosphere Response Experiment, WCRP Publications Series No. 3 Addendum, January 1990). The goals of TOGA COARE are to describe and understand:
Readers interested in pursuing research with COARE data are advised to examine the COARE Experiment Design (TOGA COARE Experiment Design, TOGA COARE International Project Office, October 1991) and the COARE Operations Plan (TOGA COARE Operations Plan, Working Version, TOGA COARE International Project Office, September 1992). These important documents contain details of all 13 ships and 7 research aircraft that participated in COARE, together with descriptions of proposed flight plans in various weather conditions, deployment of ships (including the two ships with Doppler radars of 300-km radius), mooring deployments, plans for enhanced atmospheric monitoring over the COARE Sounding Array, and many other details. A map of the COARE geographical domains (Figure 1) is included below.
In 1990, the TCIPO was formed and Dr. David Carlson was appointed director. The Project Office took responsibility for coordinating assembly of all the measuring platforms in the field, and (with advice from the COARE International Science Oversight Team) determining their deployment on a day-to-day basis. For logistic reasons, and to keep the aircraft in the summer (rainy) hemisphere, it was decided that COARE operations should be directed from Townsville, Australia, though most of the planes operated from Honiara, in the Solomon Islands. The Intensive Observing Period (IOP) was carried out from November 1992 through February 1993 and was embedded within a period of Enhanced Monitoring from September 1991 through May 1994 for physical oceanography and from July 1992 through June 1993 for meteorology.
Thanks to a combination of great dedication on the part of Dr. Carlson and his international team and enthusiastic teamwork on the part of the scientists, the field phase was carried through virtually as planned. Participants considered the data collection effort to be highly successful.
Since the end of the IOP, the COARE scientific community has been working on data processing, quality control, analyses, comparison, exchange, and discussion. As this work progressed, the community recognized that a COARE-wide checkpoint was needed to:
The plan for addressing these goals was realized in the TOGA COARE International Data Workshop.
Prior to the Workshop, preliminary meetings were held to examine data from participants within the four individual Large-Scale Atmosphere, Mesoscale Atmosphere, Air-Sea Fluxes and Oceans Science Groups. However, the Workshop was the first opportunity since the field phase for all the scientists involved to examine their data in toto.
TCIPO prepared a unique working environment for the scientists at Toulouse by setting up a large network of computers, designed to meet the varied requirements of the 276 participants. Many aspects of the data management techniques shown at this workshop are highly innovative, and will make it much easier for scientists around the world to work on these data than in the past. While it may be somewhat early to predict the final "payoff" from COARE, there is no doubt that it was highly successful in collecting data over a vast range of topics. Our impression, as organizers, is that after this workshop, many scientists worldwide will be tackling the interdisciplinary challenges posed by the comprehensive COARE dataset.
We regret the transition of Dr. Carlson from TCIPO to his new position as director of the NCAR Atmospheric Technology Division, but his replacement, Dr. Richard Chinman, is taking on the job of data management with considerable energy and skill.
It is clear from outcomes of the Workshop that considerable headway has already been made toward achieving the goals of TOGA COARE. This report highlights these early achievements, most of which have yet to appear in scientific journals. However, to ensure that the momentum continues for data analysis, intercomparisons, scientific collaborations, and further research, the COARE Science Group prepared a series of recommendations included in Section 2 of this report. The continued success of COARE will depend on very vigorous pursuit of the interdisciplinary aspects of the COARE dataset.
TOGA COARE was an internationally-endorsed addendum (World Climate Research Programme, Scientific Plan for the TOGA Coupled Ocean-Atmosphere Response Experiment, WCRP Publications Series No. 3 Addendum, January 1990) to the TOGA Implementation Plan, and the substantial resources required for its implementation were contributed by 19 nations. Of significant note was the important cooperation of meteorologically- and oceanographically-oriented agency programs and researchers. As a result, the oceanographic community has a landmark ocean-forcing dataset. University-based and government laboratory researchers successfully combined efforts to conduct COARE.
The observational strategy for achieving the objectives of COARE was to combine an intensive field experiment of limited duration with a longer enhancement of the relatively sparse TOGA monitoring network in the western equatorial Pacific and to combine these observations with model-based data assimilation to achieve the best possible description of the warm pool system. Observational efforts were also nested in space to be able to relate processes occurring on widely different scales.
The timing of the IOP was set for the northern winter to maximize the probability of strong westerly wind events, and thus the period of most intense air-sea interaction. Furthermore, the November-February period is a time of active transition during ENSO while also being a period of the maximum strength of the east Asia Winter Monsoon.
It was considered especially important to make observations during the well-organized convection associated with the dominant synoptic systems (intraseasonal oscillation and westerly wind burst), because the character of the heat, momentum, and freshwater fluxes associated with these strong signals was not known. The time and space relationships among these fluxes and the oceanic response were needed in order to assess their possible importance during the onset of ENSO events.
During COARE, there was strong emphasis on measuring fluxes between the ocean and the atmosphere. The Intensive Flux Array (IFA) was designed to obtain these detailed flux measurements. A synergistic mix of oceanographic and atmospheric observing systems was assembled to measure and estimate the air-sea fluxes of heat, moisture, and momentum with a variety of complementary methods. Direct measurements of fluxes using eddy correlation techniques were made from ships and from aircraft at limited locations and times, providing the means to calibrate other methods that provided more extensive spatial and temporal coverage. Indirect estimates were made with calibrated bulk formulae using data collected from ships and buoys, and from satellites with algorithms validated with the in situ measurements. Ground, ship, and aircraft-based remote sensing of the atmosphere in the IFA provided additional estimates of the fluxes. Here, the data from the Doppler weather radars are critical, as they permitted estimates of rainfall over broad areas with relatively high resolution. Dual Doppler radar measurements from ships and aircraft gave detailed information on the spatial variability of the momentum fluxes on small scales not otherwise resolved; this information is crucial to the understanding of the relationship of mesoscale convective variability and variations in surface fluxes.
Atmospheric and oceanic profiles of temperature, moisture/salinity, and horizontal velocity combined with data assimilation schemes are beginning to provide integral constraints on interfacial fluxes. In order to achieve adequate coverage of atmospheric profiles, additional ships with sounding systems were utilized. For the ocean, additional moorings and ships were used.
In particular, strong intraseasonal atmospheric oscillations were observed within the IFA during the IOP, along with some episodes of westerly winds, including a strong westerly wind burst (WWB) over the IFA between 20 December and 3 January. Atmospheric budget results show that the maximum precipitation over the IFA occurred one to two weeks prior to the peak of the westerly winds. The WWB generated an eastward-moving oceanic equatorial Kelvin wave which caused the thermocline to deepen (and the sea surface to warm) all the way across the Pacific Ocean, as well as directly forcing the warm surface waters to move eastward in a jet-like feature. The WWB also cooled the upper ocean in the IFA through air-sea heat fluxes and upper ocean mixing. The net effect was to displace the warm pool eastward. The large range of atmospheric and oceanic conditions which were observed during COARE confirms previous hypotheses about the role of air-sea interaction in the warm pool system, and they also provide an excellent database for future model development.
The following COARE working group sections provide a general description of the scientific issues and findings discussed at the workshop. Readers needing further details are referred to the workshop abstract list (Appendix C).
Sea surface temperature (SST) plays a central role in this description. There was considerable discussion of coupled processes during the IOP, particularly with regard to the strong periods of surface westerly winds that occurred in conjunction with the passage of large-scale intraseasonal convective episodes (contrasted with the "break" periods between convective episodes).
In common with the mesoscale group, the large-scale atmospheric circulation group addressed two critical questions:
These questions have not been fully answered, but initial results from the COARE data provide some insights and have helped to frame COARE research efforts for the next several years.
Large-scale forcing for deep convection over the warm pool is provided by intraseasonal oscillations, which were very pronounced before and during the IOP. Several prominent Madden-Julian Oscillations (MJOs) occurred during the IOP; these were accompanied by westerly wind bursts and periods of intense convection across the Large-Scale Sounding Array (LSA) and IFA. One such convective episode seemed to "skip" over the IFA, however. Interactions with mesoscale convective systems provide one possible explanation for this "skip." An intriguing alternative theory involves the interaction of eastward-propagating 30-60 day Madden-Julian Oscillations with higher-frequency large-scale disturbances on 4-5 day or 15-day time scales. An important subject of intensive ongoing COARE research is to describe and understand the effects of various large-scale (i.e., resolvable by general circulation models) transient disturbances on the large-scale flow and to separate these effects from those of sub-grid scale convective systems.
It was clear from multiple vigorous discussions that no clear consensus exists regarding the origin and morphology of strong WWBs. Among several interpretive paradigms offered at the workshop were:
The COARE data represent a unique source of detailed information on the three-dimensional structure and temporal evolution of several WWBs (and the oceanic response to them), and we anticipate intensive efforts on WWB case studies during the IOP.
The integrated effects of coupled processes in the COARE domain are observed in time-averaged statistics of the large-scale circulation. The COARE data have revealed features of the climatological flow that deserve continued study and provide an excellent baseline for assessment of operational large-scale analysis products. Sounding data from the warm pool reveal a frequent occurrence of moisture inversions and stable layers near the 0 deg C level. These features appear to be in some way related to local and/or remote effects of melting in deep convection. They may account for the ship-based radar observations of a bimodal population of cell top heights with one peak in the middle and another in the upper troposphere.
Large-scale budgets of heat, moisture and momentum over the LSA have been calculated by several investigators based on real-time sounding data. Budget-derived precipitation estimates for the IFA reported at the workshop ranged from 7 to 11 mm day ^-1, depending on analysis methods and treatment of surface fluxes. Budget results show that maximum precipitation over the IFA occurred one to two weeks prior to the peak of the WWBs. This pattern was particularly true for the strongest WWB in December, with a maximum average rain rate over the IFA of 18 mm day ^-1 around 12 December, over two weeks prior to the strongest low-level westerlies at the end of the month. Rainfall was a minimum when the westerlies peaked, followed by a slight increase reaching a maximum of about 7 mm day ^-1 on 7 January (far less than the mid-December maximum) when the winds once again became light.
Precipitation can be determined from both heat and moisture budgets; however, there is considerable uncertainty in the heat budget results due to incomplete knowledge of radiative fluxes. Some uncertainty in the low-level moisture field at ISS sites still exists, which affects the accuracy of moisture budgets, highlighting the need for completion of the Level II quality-controlled soundings archive.
Mesoscale convective systems with expansive (~100 km) cold pools or wakes (with cooling up to 4 deg C) were common over the warm pool, contributing to the enhancement of surface sensible and latent heat fluxes of up to 5 times or more.
The Mesoscale Convection working group tackled the second goal of TOGA COARE (see Section 3) directly and contributed to discussion of the other three. In particular the working group focused on the following COARE atmospheric and modeling component goals elucidated by Peter Webster:
As elsewhere, MCSs in COARE are often organized into quasi-two-dimensional bands, sometimes as rapidly moving squall lines approximately perpendicular to the vertical wind shear at low levels (e.g., 10 and 22 February), but more often as slower-moving bands, approximately parallel to the deep tropospheric shear (9, 17 and 18 February). In COARE, it was often very difficult to determine from which side the low level air was feeding into the convective systems. This is a fundamental issue in any analysis, because the very essence of a mesoscale convective system is that its lifetime is much longer (several h) than that of its constituent cumulonimbus clouds (~1 h). If low level, potentially buoyant air is not supplied to the MCS at a sufficient rate, say at a relative inflow speed of ~5 m s^-1, new clouds cannot be generated and the system must decay.
The building blocks of MCSs are the cumulonimbus clouds, and a number of sessions included observations of the characteristics of the convective scale vertical motions in COARE cloud systems. Preliminary aircraft in situ data, Doppler radar data, and lightning data indicate that the convective scale up- and downdrafts in the COARE systems are similar to those described elsewhere over tropical oceans. That is, they are surprisingly weak, considering the Convective Available Potential Energy (CAPE) for deep convection. As elsewhere over the tropical oceans, radar reflectivity statistics show a rapid decrease in reflectivity with height within convective cores, a strong indicator of weak vertical velocities. As elsewhere in the tropics, the lightning flash rates over water are one order of magnitude less than over land, another strong indicator of weaker updrafts.
As elsewhere over tropical oceans, the convective downdrafts contribute to "cold pools" of air which, in turn, result in large surface energy fluxes. The role of unsaturated mesoscale downdrafts is thought to be indirect but important, through mixing drier (and potentially warmer) air into the cold pool, reducing sensible heat flux while increasing latent heat flux. Several studies estimated the magnitude of enhanced fluxes in MCSs. One generalization that will probably stand the test of time is that the squall (shear-perpendicular) systems have very large flux enhancements, compared with all other MCSs. In fact, a surprising result is that some MCSs have rather little flux enhancement, for the reason that in those cases the wind speed within the cold pool is not much different than that of the "environment." There is some controversy about the recovery rate of boundary layer air, some saying it is 6-18 hours, as in GATE; others believing that the fluxes from the warm ocean permit new MCS growth more quickly.
Statistics and stratification of the dataset
Faced with this complicated variety of precipitating convection, a sensible first step in analyzing the large TOGA COARE datasets has been to partition convection into categories. Without spectral gaps, no such partitioning will be completely satisfactory for every event. Moreover, partitioning schemes based on different datasets will not (because they cannot) always produce exact agreement when the individual events are compared.
Several classifications of the radar-observed mesoscale precipitation patterns were suggested (Houze, Kingsmill, LeMone, Rickenbach, Uyeda). These classifications refer to:
These suggested classifications of the precipitation are different ways to describe the same phenomena. The relationship of one classification to another should be explored; the various classification schemes need to be reconciled. The potential for discrepancies has not deterred and should not deter the development and use of partitioning schemes in analyzing large datasets. However, the usefulness of the results of such analyses for interpreting budgets and in parameterizing convection will be greatly enhanced to the extent that the partitioning can be identified with physical processes that determine the observed convective structures.
One basic stratification, that of convective and stratiform precipitation, has proven useful (e.g., in the interpretation of larger-scale heat and moisture budgets) because it partitions rainfall associated with different precipitation processes. This stratification is supported by observation of differences between the categories in precipitation drop-size distributions, weather radar signatures, and mesoscale dynamical features.
The shipboard radar provides an Eulerian view of mesoscale precipitation structure--high time resolution data over a fixed area for a long period of time (~20 days)--while the airborne radar data provide a Lagrangian view--low time resolution data over an area that follows a mesoscale precipitation area for a few hours. The views provided by aircraft and ship are complementary, and there is a general impression that a better understanding of the mesoscale precipitation patterns can be achieved by reconciling the two views.
Why should anyone care about stratifications of such details as the magnitude of convective updrafts, type of precipitation mechanism, or the orientation of convective bands? Speakers suggested several reasons. The convective draft profile and precipitation mechanism are closely related to mass and hydrometeor detrainment as a function of height, which in turn are directly related to the convective heat and moisture profiles, and of course the radiative heating profile. These are the diabatic heat sources contributing directly to the ascending branches of the Hadley and Walker circulations, modulated by disturbances on scales from days to intraseasonal. The scientific objectives of COARE recognize that the MCSs are strongly interactive with the surface fluxes of sensible heat, latent heat, and momentum. What has not been so obvious is that these fluxes may be directly related to the way that the convective bands are oriented (cf. Lemone et al., Appendix C).
Comparison of COARE to other datasets
The large scale environment in the COARE area is quite different from that of GATE, and the relationship between the MCSs and their environment may govern the evolution of MCSs. On the small scale, for example, those COARE MCSs which take the form of multiple short-lived bands parallel to the shear (e.g., 9, 17, 18 February) have little low-level relative inflow. Speculation is that they must either die quickly or propagate, and the latter is observed quite frequently. When this discrete propagation takes the form of "jumps" as large as 50 km every 2 hours or so, with no apparent low-level forcing such as gust fronts, the question of a mechanism (gravity waves?) must be raised.
On larger scales, a wider variety of interactions is available for the propagation and regeneration of COARE MCSs than in GATE. In the eastern Atlantic, the convectively unstable area is tightly constrained between cold water regions just north and just south of the GATE array, and the MCSs can be and are regularly forced by strong easterly waves moving westward from Africa. In retrospect, it is no surprise that the intertropical convergence zone in GATE was tied to the warm water belt between 5-10 deg N, and that there was tight coupling between the MCSs and the upward vertical velocity half of the 4-5 day, 2000-km waves. The warm pool has quasi-uniform sea-surface temperatures >28 deg C over a huge area, and no dominant forcing mode on time scales between diurnal and intraseasonal. The bad news: MCSs and how they are generated are more complex in COARE. The good news: we have the opportunity to study the problem in its full, rich generality.
At this stage of research, it is appropriate that we have generated more questions than answers.
Thus the responses to interactions at the interface are quite different, and the oceanic deformation radius is similar to the size of atmospheric cloud clusters. Major uncertainties in our knowledge derive from lack of a dynamically consistent definition of mesoscale; the widely varying dynamical differences between the ocean and atmosphere; and the lack of detailed knowledge of the scales that interact directly and of the fundamental constraints on the types of interactions that can occur.
Can we isolate the dynamical and thermodynamical regimes in which the ocean and atmosphere interact? For example, there is a fundamental difference between the chaotic structure of clouds and the organized MCSs that clouds can produce (see the discussion leader's impressions for session CM2, Appendix D). Is the heat budget of the warm pool dependent on the type of mesoscale organization that occurs? Is the degree of detail in small-scale systems of importance to the larger scale? Diagnostic budgets from radiosonde arrays indicate lack of sensitivity to such detail and that parameterizations may be developed to successfully incorporate the sub-grid details in prediction models. But uncertainty remains on the role of mesoscale processes in further developing major cloud systems, such as superclusters, which can markedly affect the environment.
The response of the boundary layer following convective "mining" of the high enthalpy air occurs at a diurnal scale, appears to have a two-day cycle, and varies by the type of large-scale environment (i.e., large regions of dry mid-level air). Diurnal variation of the convection is associated with a strong diurnal cycle of oceanic skin temperatures and may have substantial effects on the larger scales by evolutionary changes of the moisture and heat content of the lower atmosphere. Freshening and cooling of the upper ocean by fresh water input from rain also leads to long-lived systems, and to protection of warm, salty subsurface water, which may be brought back to the surface in periods of strong winds.
Predictability depends on the large-scale structure, and substantial increase of forecast skill for the winter season from general circulation models (GCMs) has been achieved for winter, but not for summer and tropical regimes. This raises fundamental questions--does this mean that the mesoscale is not predictable? The major problem seems to be that the combination of Rossby modes and gravity modes leads to a highly non-linear, chaotic system. We need to determine the conditions under which these non-linear processes are deterministic and those in which rapid growth of errors occur. In particular, does reducing the scale eventually lead to a domain in which chaotic noise dominates.
In summary, determining the two-way scale-interaction process in the atmosphere, the means of interaction at the ocean interface, the subsequent response of both the atmosphere and ocean, and the further scale interactions that can occur are of considerable importance for the COARE research program. Ultimately, this will lead to improved parameterizations, an accommodation of the fundamental uncertainties involved, and improved forecasts of both large and small scale systems.
As a conclusion of numerous e-mail discussions before the workshop, the main issues of numerical and idealized modeling were addressed during the modeling discussions. Following is a summary of these discussions together with important points raised in Toulouse.
The main issues of the cloud and mesoscale modeling group were:
Numerical Models -- Limitations and necessary improvements
Two broad classes of numerical models are relevant to the Mesoscale/Convection COARE Group: cloud-resolving models (CRMs) and mesoscale regional models. Two-dimensional cloud models can be run using mesoscale or regional scale domain sizes (< 2000 km), with or without grid nesting techniques. The most important factor distinguishing the two types of models is the horizontal resolution (dx). The CRMs cannot adequately resolve important cloud features with a dx > 3-5 km. An important consequence is that they are limited in domain size (especially three-dimensional models). They use parameterizations of various physical processes (e.g., microphysics, radiation, turbulence). The regional models are generally used with dx > 10-30 km. They use cumulus parameterization schemes to represent cloud processes occurring at subgrid scales. Some ice-phase processes are being resolved at the grid scale of some mesoscale models.
Cloud-Resolving Models (horizontal grid mesh size < 3-5 km): Besides grid nesting techniques, simple initialization procedures are generally used, in starting from a horizontally homogeneous atmosphere (from a sounding) and a perturbation in planetary boundary layer (PBL). Cloud ensemble simulations use an initial random noise in low levels together with periodic lateral boundary conditions. Cloud system simulations (predominantly cloud lines) use initial cool or warm bubbles in the low levels of the atmosphere together with open or mixed lateral boundary conditions. The domain sizes, sophistication of physical parameterizations, initialization procedures, etc., will vary depending upon whether "Class 1" to "Class 4" organized type systems are simulated. For example, variations from shear-perpendicular storm structure require the use of three-dimensional models but with more computational constraints limiting their usefulness.
Regional Scale Models (horizontal grid mesh size > 10-30 km): In contrast with CRMs, three-dimensional or four-dimensional (space and time) initialization procedures are generally used, based on global forecasts and/or analyses from NMC, ECMWF, NOGAPS, etc., or special assimilated datasets (e.g., from GSFC). In the tropics, the data assimilation is known to be a difficult problem, mainly because the flow is strongly divergent. Plans for large scale re-analysis were presented during the workshop. The mesoscale modeling community unfortunately was not represented at Toulouse. However, other international programs oriented towards mesoscale modeling (see below) were presented and should be able to alleviate this problem.
Interactive nesting technique (address the "scale-interaction" problem): One-way and two-way nesting of cloud and mesoscale models are believed to be the most comprehensive means of addressing the "scale-interaction" problem. This problem is addressed from a modeling view--initiation and evolution of mesoscale convective systems are affected by processes, occurring over a broad range of spatial and temporal scales, that exceed the present capabilities of computers (range from small-scale microphysics and turbulence to large-scale dynamics). There are two approaches for two-way nesting of cloud and regional models: (i) to embed the cloud model within a mesoscale model; (ii) to nest the cloud model within itself using a succession of coarser grid resolutions (grid nesting). One example of one-way nesting is the use of sophisticated initialization of mesoscale models to simulate the four-dimensional variability of the environment in which convection is initiated (instead of initial horizontally homogeneous conditions).
Coupled ocean-atmosphere mesoscale models: During the workshop, it was made clear that coupled ocean-atmosphere models at mesoscale are necessary to achieve the goals of COARE (also discussed during a session dedicated to coupled ocean-atmosphere modeling). Such two-dimensional experiments are underway in some institutions.
Parameterizations: All models use parameterization of physical processes. It is important to know their limitations and possible improvement based on TOGA COARE observations.
Two special intercomparison measurement days had been scheduled for 27-28 November 1992 and 3-4 February 1993, when several ships operated together in the vicinity of the WHOI buoy. At times they were overflown by the aircraft equipped with boundary-layer meteorological and turbulence instrumentation. To date, the thrust of Flux Group activity has been the analysis of data recorded on these days, and on others when certain platforms were in fairly close proximity. This has led to the discovery and resolution of instrumental problems and to the adoption of common routines for data analysis, aimed at refining our flux estimates, and at providing a measure of confidence in their accuracy.
The Flux Group has met twice already to work on these issues, initially in Boulder (September 1993) and again at Scripps Institution of Oceanography (March 1994) when, for the first time, we discussed complementary analyses of surface flux products from satellite and radar scientists. In preparation for the Toulouse workshop, we also identified a number of scientific topics on which attention should be focused. Our main goal for Toulouse was to form links with other groups, merge datasets, and pursue these scientific foci. Detailed proceedings of the Scripps workshop are available on the TCIPO computer (ftp://tcdm.coare.ucar.edu/pub/SCIENCE_GROUP_INFO/air_sea_fluxes/summary_9404.as).
Some notes on the measurement accuracy of various components of the air-sea energy exchange follow. Our intercomparison analyses had generally indicated that certain specific instruments or measuring methods were more reliable than others, and they have been selected as the standard against which other instrument biases were identified and corrected.
Shortwave radiation
This is the largest term in the ocean's surface heat budget, and most platforms (and island sites) were equipped with pyranometers to measure incoming shortwave irradiance. These instruments are not highly accurate, degrade with time, and without considerable care cannot be relied upon to achieve the 10 W m^ accuracy goal. Following the field intercomparisons, pyranometers from several ships and moorings were intercompared at Woods Hole and at CSIRO in Canberra. The Wecoma instrument was then calibrated absolutely at the Bureau of Meteorology; since this agreed exactly with the pre-COARE factory calibration we have adopted this instrument as our reference. One important application of the extensive coverage of the shortwave radiation field will be the development of improved algorithms for its estimation at the surface from remotely-sensed data.
Longwave radiation
The field intercomparisons had shown that pyrgeometers (which measure global downwelling longwave radiation) can incur errors up to 50 W m^-2 during daytime due to solar heating of the dome. This can be avoided if body and dome temperature signals are recorded separately as was done for the Moana Wave instrument. For those sensors which were subject to the error, a correction algorithm requiring solar input has been found and validated against Moana Wave data and radiative transfer algorithms (using atmospheric sounding profiles). Errors in longwave measurement are not as serious as in shortwave, because net longwave in the COARE region does not deviate greatly from around 55 W m^2 outgoing.
Air temperature and humidity
The side-by-side comparisons of air-temperature sensors during the intercomparison days revealed that many were susceptible to solar radiation error. Commonly-used louvered screens depend upon natural wind ventilation to reduce this error to an acceptable level, but radiation exchange between the temperature sensor and the interior of a screen irradiated with 1000 W m^-2 is severe. Under the light wind conditions of the western Pacific, stationary platforms (moorings and ships on station) are most seriously affected. Comparisons with the Franklin's double-shielded, forced-ventilated psychrometers enabled errors in other instruments to be detected and correction schemes applied.
Sea temperature
Intercomparisons of sea temperature sensors from ships and buoys show good agreement at night, indicating reliable laboratory calibration of the various sensors. During the day, however, variability of SST values between sensors at different depths and platform separation led to the realization of just how large near-surface vertical and horizontal non-uniformity could be in this environment of strong solar warming and light winds. This in turn led to the decision to restrict use of the acronym SST to refer to the actual skin temperature, and to incorporate into the developing bulk flux algorithm a model to extrapolate temperature, measured at some depth, to the surface.
Turbulent fluxes
In general, direct measurement of turbulent fluxes is not practicable on a continuous basis. In the context of experiments such as COARE the most valuable purpose of direct flux measurement is to validate bulk flux algorithms and to explore the form of the exchange coefficients on which the bulk flux estimate depends. Work prior to COARE had established that the exchange coefficients in bulk formulae increase markedly at low wind speeds, but their exact form was uncertain during daytime largely due to the considerable increases in SST that occur in the top few meters of the ocean under those conditions.
Direct measurement of the turbulent heat and momentum fluxes involves recording and correlating the rapid fluctuations of wind, temperature, and humidity using sonic anemometers on the ships and pressure probes on the aircraft. Platform motion in each case must be recorded and removed from the apparent wind signal, involving a considerable volume of data processing. At the time of the Toulouse workshop, comparison of the turbulent fluxes measured during the intercomparison days had not been attempted. This will be a major activity during the third fluxes workshop scheduled for August 1995 in Honolulu.
At this stage only the Moana Wave turbulent flux analysis has been completed. The three cruise legs cover a total of 65 days during the IOP and include a wide variety of environmental conditions (including the December WWB). These data have enabled us to determine the light wind interfacial parameters (Roughness Reynolds numbers and their scalar equivalents) and the exchange coefficient functions which are used in a new bulk algorithm.
The new algorithm has been applied to data obtained simultaneously by Wecoma, Moana Wave, and the WHOI central mooring (IMET). The primary meteorological observations needed as input to the algorithm have been carefully quality controlled on the basis of the ship/aircraft/mooring intercomparison days undertaken during the IOP. The resulting fluxes averaged independently over the three ship legs are given in Table 1.
Moana Wave operated in the vicinity of the WHOI buoy most of the time, but Wecoma's 100-km butterfly pattern ranged up to 60 km away from the other two. The remarkably close agreement in each and every parameter supports our notion that flux datasets will achieve the COARE goal of 10 W m^-2 accuracy over reasonable space and time averages. The net atmosphere-ocean heat exchange may be calculated from the data in Table 1, taking a shortwave albedo constant at 0.055 and calculating upwelling longwave from the given SST. The three platforms respectively indicate 51, 54, and 58 W m^-2 for leg 1 and -20, -28, and -24 W m-2 for leg 2. Only in leg 3 is the difference significant with -34, -8, and -37 W m^-2.
The average precipitation rate for the IOP based on point rain-gauge measurements on ships and buoys was about 11 mm day^-1. Precipitation estimates (IFA averages) presented at the workshop based on heat and moisture budgets ranged from 7 to 11 mm day^-1. (Post-Toulouse studies using three different analysis techniques yield an IOP-average rainfall rate over the IFA of 6 mm day^-1, a value quite close to estimates from the Vickers MIT radar taking into account attenuation by rainfall. These results must still be regarded as preliminary and clearly indicate that further intercomparisons are needed.) In comparing these results, it is extremely important to make the distinction between point measurements (e.g., ORGs) and areal measurements (based on radars, budgets). Recent analyses show that the IFA is at a local minimum of precipitation with higher values to the north and south. Point measurements on the fringes or near the edge of the IFA may not agree well with averages over the IFA.
Warm rain occurring in small cells removes water vapor from the atmosphere, just as rain does from larger convective systems. If the consequent change in the water vapor field is transmitted to the scale of the soundings on a 6-to-12-hour time scale (the local change of water vapor is computed using a 12-hour difference), then the effects of the warm rain cells should be reasonably well sampled by the sounding network. However, one could envision a situation where a local recycling occurs, with small-scale cells feeding off enhanced evaporation through local downdrafts, that is not sampled well by the large-scale sounding network. This process could lead to an underestimation of rainfall; however, over a long period this effect should be minimized.
The lower end of the budget estimates (6 to 7 mm day^-1) compares favorably with the IFA-averaged rainfall rate determined from SSM/I data. On the other hand, GOES Precipitation Index (GPI) estimates appear excessive, with a pronounced overestimation of rainfall at times when there is considerable cirrus but no rain; e.g., at the end of December.
Good estimates of the surface precipitation fields are crucial for much of COARE science--ocean mixing, moisture budgets, radar and satellite ground-truthing, convective processes, etc. Rainfall is a topic that cuts across all groups and scales--results of rainfall observations reported to the workshop contain good and bad news regarding the accuracy with which it can be determined. It appears that 0.25-km and 10-min resolution rainfall fields may be feasible from the radar data, which is good news for ocean-mixing people. Also, comparisons showed that GPI, satellite, and sounding moisture budget estimates of rainfall tracked in situ measurements from optical rain gauges on the TAO array well over a 10-month period, but with the mooring about a factor of two higher than the others. At the same time, comparisons on the ships and land calibrations indicate an ambiguity up to a factor of two in the calibrations of the optical rain gauges installed for ground truth during the COARE experiment. This uncertainty impinges seriously on all the scientific effort related to rainfall. Because all concerned were able to interact and exchange information at the workshop it has been possible to initiate an intensive effort, involving the flux and mesoscale/convection groups, to resolve this difficulty quickly.
Major progress was made during the workshop in understanding the effects of rainfall on the behavior of the ocean mixed layer. In the warm pool, freshwater has a very significant influence on surface water mixing and on SST. The combined effects of strong daytime heating and frequent rainfall produce a stable density gradient at the surface that can be overturned only by considerable wind forcing. Several complementary results were presented to shed light on the physical processes involved. Rainfall data for the period 4-16 January from the WHOI buoy were used as input to the Pacific Warm Pool mixed-layer model. Without precipitation, the model predicted downward mixing every night to the top of the thermocline at 80 m. With precipitation for this period added to the model, heat was trapped in a depth range less than 40 m, which has important implications for the maintenance of the warm pool and its heat budget.
The COARE bulk flux algorithm contains a calculation of the heat flux due to rainfall. This calculation is based on the assumption that rain reaches the surface close to wet bulb temperature, and some of the first data confirming this were presented. Thus there is a sensible heat loss due to rain which averaged 2.5 W m^-2 over the three Moana Wave legs. During very high rainfall events (e.g., 100 mm h^-1 lasting several minutes) this was as high as 200 W m^-2, of the same order as the latent heat flux.
Working groups identified other matters that must be resolved to carry forward issues such as flux variability and parameterization. For example, the heat budget of the mixed layer and daytime surface warming depend on the variation of solar energy with depth. The near-red is absorbed near the surface and the blue/green at lower depths, depending on water clarity. The COARE bulk algorithm uses extinction coefficients based on measurements made in other oceans, but light penetration data for the COARE water are available and will be processed with urgency.
Shortwave radiation was observed to penetrate below the mixed layer at levels sufficient to lead to a significant contribution to heat content changes which could later influence the sea surface temperature either locally or remotely. This may explain the previous observations of frequent temperature inversions below the mixed layer of the warm pool. During the strong WWB, entrainment of nutrients into the mixed layer apparently led to a phytoplankton bloom which altered the absorption of penetrating solar radiation enough to enhance heating rates by about 0.1 deg C month^-1, in the top 20 m.
The wealth of COARE data brought to the workshop encouraged on-the-spot re-examination of the values of certain constants and parameters in common usage for air-sea exchange calculations. Some, such as the exchange coefficients used in the bulk algorithm, are the subject of continuous discussion and refinement. Others, notably the long-wave emissivity and short-wave albedo of the ocean surface, are based on determinations using early techniques and are not necessarily appropriate to the COARE region. Reliable values for both of these parameters emerged from the workshop activities.
Perhaps the most important aspect of the workshop for the flux group was interaction with other groups and the establishment of collaborative links. Freshwater science issues require input from the mesoscale, surface, and ocean mixed-layer datasets. These will be organized initially with the goal of resolving the serious discrepancies in rainfall measurement referred to above; we expect this to lead to a joint effort with radar scientists to produce rainfall and surface wind maps for the IOP. A parallel effort will be made jointly with the Oceans Group to integrate surface currents into our flux calculations. Scientists from both areas are concerned about the role of fluxes and their spatial variability on SST and the onset of convection within the ocean. As the timeline in Section 6 indicates, joint projects with modelers have been initiated in all areas. Atmospheric GCMs have been found to respond quite sensitively to the introduction of higher exchange coefficients at low wind speeds. The response of these models to the new COARE algorithm remains to be seen. Ocean mixing, mesoscale, weather prediction, and coupled models all have need for air-sea interface fluxes, and we in turn require their gridded products.
One feature documented in the early analysis of the IOP data was that data from the ship surveys were aliased by the strong tidal signals. These signals have been quantified by the moored array. Part of the tidal signal should be removable in the ship survey data by using a deterministic model and simple mapping techniques. All participants in a session devoted to the study of the ocean heat budget expressed optimism that the final closure will be to an accuracy on the order of 10 W m^-2. It also seems probable that the internal tides are at least partially predictable. The presence of strong, non-linear solitons has been documented, occurring at spring tides (cf. Pinkel et al., Appendix C). Their importance needs to be studied.
Substantial progress has been made using 1-D models to quantify the heat budget in the oceanic mixed layer. Results suggest that advective terms averaged over many days are small but are clearly important at particular times. Early estimates of the advective terms based on time scales of days are of the right magnitude to explain the residual in the 1-D estimates. Data and analyses to date suggest the Oceans Group has the means to quantify the oceanic heat budget in the COARE IFA domain. Similar studies on the salt budget have begun. The work is not as far along, but early analyses are encouraging. Collaborations began at the workshop, and excellent progress should be made in the coming year on both the heat and salt budgets.
A rich dataset has been obtained to describe the velocity field over the COARE domain. The moored array has provided information on the importance of diurnal and semidiurnal tides as well as other high frequency features. At lower frequencies, the velocity field is dominated by near-inertial variations, presumably forced by the local wind fields. Since this period is about 12-18 days in the ocean IFA domain, this is the major signal velocity in the individual survey ship data. Lower frequency signals are also present in the moored records associated with larger scale phenomena in the near-equatorial region (e.g., the Yoshida jet). Results presented at the workshop suggest that the moored and survey ship data will provide information on the detailed evolution of the velocity field on a wide variety of scales. The Oceans Group has begun using the observations to provide estimates of the various terms in the momentum balance and suggests the importance of nonlinear terms in the region. A combination of the available velocity, density, and forcing fields will enable scientists to address the momentum balance on a variety of scales.
A rich dataset of near-surface temperature and salinity observations in the 0-5-m layer of the ocean has been obtained from both the WHOI mooring and survey ships. The temporal evolution of the diurnal cycle of numerous rainfall events has been documented at the mooring. Survey ships have substantial information on the spatial structure of these near-surface features. The data will allow for quantification of numerous individual events, as well as statistical analysis of the time and space series.
During the strong WWB of December 1992, the mixed layer deepened and cooled; when the WWB ceased, heavy rains caused the mixed layer to become very thin, allowing it to warm very rapidly to temperatures near those prior to the WWB. However, the net reduction of heat content caused by the WWB remained for a substantial period after the burst, leading to enhanced sensitivity of the SST to subsequent wind events. Strong eastward surface flow forced by the WWB was observed in the upper ocean, which led to enhanced vertical shear (gradient of current). This strong shear sustained vertical mixing below the surface mixed layer for several days after the WWB had ended, resulting in an even greater thermodynamic impact of the WWB in the warm pool region. Vertical shear in the atmosphere during the WWB was intense (60 m s^-1 between 850 mbar and 200 mbar), helping to suppress atmospheric convection over the IFA.
Analyses to date of the ocean data indicated the need for atmospheric forcing field information of high space and time resolution. A forcing product with hourly resolution (to resolve the diurnal cycle) and about 25-km spatial resolution (to resolve wind field curl and divergence) are desired. The highest resolution rainfall data (from the radars) is essential to study the freshwater budget near the surface layer.
The ocean community has been asked to prepare a spatial dataset and a variety of products in the near-surface layer of the ocean, 0-20 m. Products needed by the flux and atmospheric community are SST, SSS, and ocean velocity. Collaborations were established to produce this dataset and address the scientific issues. Work will be carried out over the next two years.
Spatial variability of SST during COARE was, at times, observed to be unexpectedly large. Infrared radiometer measurements from aircraft indicate robust variability with a range greater than 1 deg C over distances of tens of kilometers during light-wind conditions. Under light-wind conditions, this small-scale SST variability is at least sometimes due to an overlay of a thin (1-2-m) layer of freshened water near rain events, which cools rapidly due to the cloudy and cold conditions above it. Similar variability was directly measured from near-surface temperature sensors mounted on the bow of R/V Moana Wave.
The apparently strong interaction of the warm pool and the overlying atmosphere on the diurnal time scale and convective-cell space scale was observed well enough during the COARE IOP to eventually determine whether the diurnal cycle must be explicitly resolved in coupled models, or whether its influence on longer time-scale variability can be simply parameterized. The diurnal cycle of precipitation exhibited a complex behavior, varying from an afternoon maximum during light-wind regimes to a late-night/early-morning mx /ximum during strong-wind regime (see Figure 3a, Figure b and Appendix C: Ando et al.; Soloviev and Lukas).
Hydrological cycle and ocean thermodynamics
Ocean observations from COARE show that heavy rainfall events are responsible for salinity and temperature changes in the ocean which substantially modify upper ocean mixing, and thus the heat and momentum fluxes (cf. Chen, D. et al.; Paulson and Lagerloef, Appendix C).
Even though this was a data workshop, it quickly became apparent that a potential gap might exist in the COARE analysis phase without adequate participation from the coupled modeling community, or for that matter the large-scale modeling community in general. Since part of the justification for COARE was predicated on the importance of the warm pool in the coupled atmosphere-ocean system, it was important that links to coupled model studies be identified, if not established. As it turned out, a number of coupled ocean-atmosphere modeling groups were represented at the workshop. In view of the relevance of COARE to coupled modeling, a special session was organized at the workshop to identify:
The meeting included several healthy discussions regarding the use of COARE data in developing and testing the component ocean and atmosphere models used to form a coupled system. Flux fields from coupled model experiments are just now being examined for the COARE domain. Of particular interest are COARE derived latent heat and shortwave radiative fluxes and improved parameterizations thereof. Discussions with the Oceans and Flux Groups considered the spatial and temporal scales that should be important in coupled simulations and the scales on which model output was desired. Turbulence and convection parameterizations were discussed as critical areas in need of additional development in coupled atmosphere-ocean models. The atmospheric modeling community looks to COARE to gain a better understanding of the nature and relation of convection to SST distributions in the western tropical Pacific, and the resulting structure or profile of heating in the atmospheric column. Sensitivity studies of atmospheric GCMs to variations on the Kuo and Arakawa-Schubert schemes result in significant differences in rainfall and surface wind patterns over the western Pacific warm pool. The sensitivity of low-level convergence to observed and modeled SSTs was also discussed in the context of the impact on the prediction of North American temperature and rainfall anomalies. Within the ocean, one-dimensional vertical mixing schemes are being tested against COARE data. In parallel, various turbulent parameterizations are being implemented in ocean GCMs. Subsequently, the OGCMs are being tested for idealized and realistic situations on the equator and off, and in response to weak winds, energetic WWBs, freshwater forcing, diurnal forcing, and penetrating radiation. Highlights of the modeling presentations included:
On larger space and time scales, why does this complex of phenomena tend to organize itself into Madden-Julian Oscillations, which propagate eastward across the warm pool at speeds of order 3-5 m^1 at 30-60 day intervals? And do the details of the smaller-scale processes affect the way MJO episodes combine together to create ENSO events?
However, the longer-term trends depend quite strongly upon the freshwater input, and this remains relatively poorly-determined as a function of both space and time. It is also unclear at this stage whether the strong effort put into observations of ocean mixing will yield a well-defined algorithm relating eddy diffusivities to larger-scale parameters (e.g., mean Richardson numbers); or whether other quantities such as high vertical mode internal wave motions need to be explicitly included. As found before COARE, the ocean response to WWBs takes the form of a Yoshida jet along the equator, with a strong shear zone beneath, in which strong mixing occurs. There is general agreement that horizontal advection of heat is usually quite small in the COARE region, because large-scale horizontal temperature gradients are small; however, there are spectacular exceptions to this rule at smaller length scales; e.g., in "solitons" which appear to be tidal in origin (Figure 5).
The ocean itself may in principle be used to provide better quantitative estimates of rainfall, by considering mixed-layer salinity budgets around closed paths, using data from SeaSoar-equipped ships. However, the rainfall field is so patchy and the freshwater layer that results is so shallow that it remains to be seen how effective this technique will be, compared to calibration of rainfall with a few rain gauges (despite their inherent problems).
Other, less-anticipated interactions between large-scale and mesoscale variability were also observed and are now being studied in more detail. During the IOP, we saw several instances of apparent advection of dry air masses from the greater Australian region. More generally, the observation of mid-tropospheric dry layers in the COARE domain most probably involves interaction (as yet not well understood) between off-equatorial air masses and in situ mesoscale convective processes.
Disturbances with 10-20 day time scales were observed to propagate westward into the COARE domain and appear to play a prominent role in the large-scale modulation of organized convection over the warm pool. These waves appear to be of non-convective origin, but they interact with the convective envelope associated with MJOs over warmer waters west of the date line. We will need to understand the morphology and interactions of this rich suite of large-scale waves with smaller-scale mechanisms of convective organization in order to fully describe the variability of precipitation and latent heating over the warm pool.
As regards variability from COARE affecting the far field, relatively less research has been carried out at this early stage, and this topic was touched upon only lightly in Toulouse. We anticipate that scale interactions carrying the effects of warm pool variability to the rest of the world will be a topic of intensive research using the Level III large-scale analysis products to be produced in 1995. The first and most basic question to be addressed is: how different is a globally reanalyzed field using the full COARE dataset from the operational version of the analysis generated in real time?
Research is in progress to study the regeneration of El Niño conditions near the date line during the IOP, which was not predicted well by current coupled models. There are intriguing hints that SST warmed to the east of the COARE domain in response to strong transient wind events within the COARE Large-Scale Domain.
In addition to investigating the horizontal propagation of warm pool variability, another "far-field" effect now being examined involves the vertical propagation of gravity waves (forced by deep convection) into the stratosphere. Preliminary examination of COARE data indicates that gravity waves are well-resolved in the high-resolution soundings; further study will await the completion of soundings post-processing.
In order to pursue these objectives, each science group was charged with constructing timelines which address the future research and product development needed to achieve the goals of COARE. In developing the timelines, the science groups deliberated on progress toward COARE goals to date and were asked to consider the following questions:
Part I of the timeline Figure 6 shows that we plan to complete the large-scale atmospheric component of the COARE Level II dataset, as outlined in the COARE soundings data management plan released in June 1993, during the next 6-12 months. Upper air soundings from the six Integrated Sounding Systems (ISSs) have been reprocessed from data recorded in situ. The 915-MHz lower tropospheric wind profiler data at all ISS sites have been processed, reanalyzed, and released. The ISS data will be transferred to the UCAR Office of Field Project Support (OFPS) for further quality control and ingestion into the Cooperative Distributed Interactive Atmospheric Catalog (CODIAC) browsing and archiving system for public dissemination.
The final component of the soundings archive for COARE large-scale research consists of the Priority Sounding Sites (PSSs), principally operational radiosonde sites whose activities were augmented during the IOP. An intensive post-processing/quality control effort is in progress at OFPS to ensure that these soundings are of Level II research quality (many of the soundings in this set failed operational acceptability tests and were archived in non-digital form). The timeline calls for OFPS to have a complete, quality-controlled archive of COARE soundings data by the end of 1994. Meeting this deadline is essential for timely commencement of the Level III assimilation efforts at the large-scale numerical modeling centers.
Effort 2 of the data products timeline concerns a recently begun effort to provide quality control and accessibility for surface meteorological data. The current timeline calls for assembly of an initial archive in early 1995 and subsequent completion of full quality control efforts in early 1996. We hope to accelerate this timeline in the near future.
Effort 3 describes the satellite data products being developed by COARE large-scale atmospheric investigators. These products include infrared and visible GMS imagery and cloud drift winds. The initial GMS IR and visible products are available now. Efforts continue to fill gaps in the operational versions of these datasets and to complete the processing of the cloud drift winds. The quality-controlled, final versions of these products should be available in 1995.
There was considerable discussion at the workshop about two aspects of the tasks remaining to complete the Level II COARE soundings archive. First, several scientists questioned the integrity of the humidity values at the lowest few levels of the omegasonde ascents. These values are critical for estimation of the convective potential of the atmospheric environment. The ISS data from several of the sites were found to have suspect values of temperature and humidity at the lowest few levels due to the unexpectedly long time constant of the sonde sensors (which require some time to equilibrate after launching from ambient storage temperatures). A physically-based correction has been applied to the lowest few levels at some of the ISS sites. At the workshop, an ad hoc committee of COARE scientists was organized to develop an assessment scheme based on horizontal consistency checks. This committee will attempt to derive maps of temperature and humidity at the surface and low levels of the soundings in and near the IFA in the next few weeks. The availability of surface meteorological data represents a critical uncertainty for the timely completion of this quality control effort.
Second, as noted above, it is important for the Level II archive to be completed without undue delay so that Level III processing can begin. Pilot efforts are about to begin at NMC to prepare data ingestion software to accommodate input data from the COARE archive. It was noted at the workshop that the proposed schedule for completion of the surface meteorological archive does not meet our desired date for shipment of COARE data to Level III assimilation centers. Every reasonable effort should be made to accelerate the development of the initial surface meteorological archive.
The first two efforts on the list (see Figure 6) would use Level II quality-controlled soundings data to carry out the same objective analyses that were implemented in real time during the IOP (on data received over the GTS, which due to operational constraints are necessarily degraded in terms of both quality and quantity). These two schemes share the common characteristic of having fewer dynamical constraints than the schemes based on primitive equation models. Efforts 3 and 4 involve large-scale reanalysis of the COARE IOP by research centers using the Level II data as input. These reanalyses would be the basis for intercomparison and data impact studies at these centers and full surface flux diagnostics would be a product of the reanalyses.
Efforts 5 and 6 would take place at operational weather forecasting centers using high-resolution operational assimilation systems. In addition to intercomparison, data impact studies, and flux diagnostics, these efforts offer the promise of carrying out predictability studies to gauge the effects of enhanced data over the warm pool during convectively active periods.
Precipitation
Rainfall estimation over the COARE domain is one of the more important tasks addressed by the Mesoscale/Convection Group. The standard rainfall map product will have 2-km x 2-km x 10-min resolution for a 150-km radius from the TOGA and MIT radars (see the list of products). This will give rainfall at 2 km above the sea surface. It is expected that useful comparisons with this product and other sources of information on rainfall will result in some iterations of this standard product over the next year or more. For oceanographic purposes, it is desirable to have a product with 0.25-km x 1 degree azimuth x 10-min resolution for a 50-km radius from the radar. This may give rainfall at a height of 500 m above the sea surface. It is recognized this product will be less accurate than the standard product. Apparent overestimation problems with optical rain gauges on some platforms require further investigation.
The preliminary intercomparison of satellite-observed clouds and precipitation and surface-measured rainfall (Chen and Houze) showed that the temporal and spatial variabilities of the clouds and precipitation over the COARE domain are important factors for intercomparison of the rainfall measurements from various platforms. We need to compare the locations and time periods of the measurements made from the ships and buoys against the mean distribution of the clouds and the ISOs during the IOP. The correlation between satellite-derived precipitation and rainfall measured from gauges on ships and TAO moorings is much better on the longer time scales. More than 90% of the hourly accumulated rainfall measurements of 2 mm or higher from the gauges are correlated with the cold cloudiness (< 235 deg K) seen by IR data.
The ship-based radar provided an excellent dataset to examine both horizontal and vertical structure of the precipitating systems (DeMott et al., Rickenbach et al., and Rutledge et al.). Horizontal morphology of precipitating systems is examined to associate the degree of mesoscale organization and rainfall to changes in the environmental wind and wind shear. It is found that isolated unorganized convection is the most common mode, occurring about 50% of the time and contributing 10-20% of radar measured rainfall. Organized linear convective structure was most common in sheared environments. It is also found that, on the shorter temporal and smaller spatial scales, the correlation between cold cloud area seen by the satellite IR data and radar observed rain area is not very good.
The 915 MHz wind profiler data from Manus and Kapingamarangi showed a very promising potential for establishing a climatology of the vertical structure, both vertical velocity and reflectivity, of the precipitating cloud systems (Williams, Ecklund and Gage). There is a nearly year-long data record available from the two locations.
A few intercomparisons of satellite-observed clouds and precipitation by IR and SSM/I data and radar data, on case-study bases, were made using data obtained from Manus Island (Katsumata et al.).
The most important question is still wide open: how to intercompare the fine precipitation datasets from the COARE IOP with the existing satellite data, such as SSM/I, MSU, and GPI, to evaluate the global rainfall estimate over the tropical ocean. Whether the fine structures or the statistical ensemble property of the cloud systems observed by radar data is more important to compare with the low spatial and temporal resolution satellite measurements remains uncertain. Some preliminary ideas were discussed. Much work needs to be done in this direction.
The list of main data sources necessary to cloud and mesoscale modeling in COARE includes:
Output expected from cloud and mesoscale models
Convective organization and interactions of physical processes at different time and space scales. One area where cloud and mesoscale models are expected to bring valuable help is in the explanation of the convective organization. Cloud models can look at this organization vs. vertical wind shear or other thermodynamic parameters (e.g., shear parallel, shear perpendicular, or more complicated shear profiles). Different two-dimensional and three-dimensional simulations of three cases observed during COARE were presented during the workshop. Numerical results show good agreement with airborne Doppler radar observations.
Two major related issues are (1) to classify convective events observed during the COARE IOP and (2) to look at the convection predictability. Modeling strategies will need to be developed given the hierarchical structure and modulation of convection over the broad range of time and scales from ~100 km to synoptic scales.
Several institutions are currently running two-dimensional and three-dimensional cloud-scale models with domain sizes from 100 to 2000 km in order to study cloud-radiation forcing under different large-scale conditions (disturbed or undisturbed) for a wide range of time scales (12 h to 100 days). The main question addressed is the relationship between cloud and mesoscale processes and the large-scale flow (e.g., are they being driven by the larger scale flow or are they in equilibrium?) These studies also can explain the diurnal variation of convection (vs. different time and wind regimes) as well the relationship between large-scale forcing (ascent) simulated from mesoscale models and imposed in cloud models to different Madden-Julian Oscillation manifestations.
Water, Q1, Q2, momentum budgets, and mass fluxes in the troposphere.
Interfacial and BL fluxes of heat, moisture and momentum. Cloud models can provide simulated spatial and temporal variability of fluxes in convective systems, together with quantitative explanations of this variability. Using this information, one problem currently addressed is the parameterization of cloud effects on surface fluxes in mesoscale and larger-scale models.
COMPARE, sponsored by WMO/WGNE, will propose and perform collaborative numerical experimentation to further mesoscale understanding and predictive capability. One objective is to establish over a period of years a testbed of a broad range of mesoscale cases using high-quality raw datasets, assimilation systems, and analysis. A selection of datasets from TOGA COARE needs to be realized in: (1) identifying scientific objectives that COARE participants believe could be addressed by meso-alpha/meso-beta scale models; (2) identifying a well-documented, specific case with good data coverage that is well suited for these objectives; (3) sketching out a possible experimental strategy for critical review and refinement; and (4) seeking collaborative effort for evaluating the results and a workshop for their discussion.
Preliminary discussion on the GCSS/COMPARE/COARE initiative took place during an informal meeting in Toulouse, at which representatives of the three groups were present. Three main common issues to these programs were identified: (1) to improve understanding and parameterization of the effects of convection in regional and general circulation models; (2) to have initial fields for initializing cloud and mesoscale models; and (3) to broaden the perspectives for the use of COARE datasets.
As a first step, collaborative research was proposed to simulate convective systems associated with large-scale WWBs (before and/or during the mature stage of wind bursts), with the goal of improving our understanding of the interaction of organized precipitating convection with the large-scale tropical atmosphere.
Work to intercompare COARE surface meteorological and SST measurements is nearly complete. Substantial questions about rain gauge calibration remain and will be followed up through 1994 and into 1995 by members of the Flux Group collaborating with radar and mesoscale investigators.
See Figure 8.It is likely that further calibration work will be required to resolve this issue. The radiometers on the NOAA P-3s must be calibrated, then the aircraft radiation datasets must be carefully intercompared. Support, in terms of funds and people, must be identified for the aircraft radiation work.
Version 2.0 of the COARE bulk flux algorithm was made available at the workshop. Further changes will likely arise as the group receives feedback from other COARE users and completes intercomparisons of fluxes from various platforms. The refinements will be made public by issuing an update (Version 3.0) in late 1995 or early 1996. In the timeline, the bulk algorithm notation "LW [[arrowdown]][[arrowup]], SW [[arrowdown]][[arrowup]] extinction" means that a formula to estimate net longwave radiative flux from only SST, air temperature, and shortwave radiation will be developed. For shortwave radiation, the bulk flux algorithm for carrying shortwave radiation down into the water column will be refined by improving the parameterization of the extinction of shortwave radiation as a function of depth. TCIPO was asked to develop a means of registering users of the algorithm and flux datasets so that they may be advised of improvements and corrections.
The UK C-130 investigators were unable to attend the workshop. Their participation was missed, and it is hoped that they will be involved in the ongoing analyses and future workshops. Considerable effort is focused on completing intercomparisons of turbulent flux measurements made on various platforms, aircraft, and ships. A matrix was developed (Table 2) to assess the status of the work and identify tasks that remain. Work assignments were made. These intercomparisons will be the initial topic at a workshop to be held 2-4 August 1995 in Honolulu. That workshop also will involve aircraft investigators from the Mesoscale Convection Group, and together the participants hope to move on from intercomparison studies to case studies. These case studies would focus on events (e.g., squalls, convective systems) as well as on times when the boundary layer was undisturbed. The context (mesoscale analyses, winds, budgets, cloud information) that comes from the work of the Mesoscale Group is essential to the further work of the Flux Group during these case studies.
The Flux Group also needs products from the ship and aircraft radars. Rain and surface wind data are the most important products to assist in determining the freshwater flux and spatial variability of all the fluxes. Joint work on calibrating rainfall is anticipated, and quick-look radar products will be used in the near term. At least by mid-1996, it is hoped that high-resolution (0.25 km by 1 degree azimuth, and 2 km by 2 km, both every 10 min) radar products will be publicly available. Satellite flux estimates are also needed. In particular, it appears that shortwave and longwave radiation fields will be best achieved by satellite data analyses. These fields would be best derived by working with satellite investigators. Wind fields and indications of surface roughness (sea state) will also be sought. Additional information may be available from a scatterometer that flew on one aircraft, but the data have not yet been examined.
The Flux Group can now provide good surface humidities and aircraft soundings to assist in analyzing other soundings and in large-scale analyses. It is hoped that area average surface heat and moisture fluxes resulting from budget studies using the soundings will be passed on to the Flux Group as they are developed and refined.
Joint work with modelers was initiated at the workshop, with surface flux datasets provided to numerical weather prediction center modelers and a dialogue started with mesoscale and coupled ocean-atmosphere modelers. It will be some time, however, perhaps late 1996, before gridded fluxes are produced. This can only result if the Flux Group is helped by others interested in data assimilation and modeling.
For the calculation of fluxes during light wind conditions in the IFA, it is desirable to have values of the surface current. This information has been requested from the Oceans Group. Near-surface optical and temperature data as well as all "sea surface" (near-surface) temperature data have also been requested from the Oceans Group in order to help refine profiles of the extinction of solar radiation in the upper ocean. Ocean budgets, asthey become available, will provide a useful check on the area-averaged fluxes developed by the Flux Group and by models.
Flux workshops are planned in July and August 1995, and a number of papers should be prepared for the 1995 IUGG meeting.
Aliasing by semidiurnal tides remains a noise source for ship surveys of current and density. Prediction and removal of tides from these records remains a hopeful prospect because moored measurements indicate that substantial (but not all) variance at tidal frequencies is coherent with the equilibrium tide, thus is deterministic and can in principle be predicted. The technique will be to compute admittance functions for current and vertical displacement and convolve them with time series of the equilibrium tide to predict semidiurnal tidal fluctuations at desired locations and times of survey ships. A prediction model should be ready by the end of 1994.
The Oceans Group has a chance, unprecedented in oceanography, to close COARE heat, salt, and momentum budgets by using knowledge of vertical fluxes, both at the surface and at depth (from microscale mixing observations), horizontal advection, and local rates of change. Advection terms appear estimable from both SeaSoar surveys and moored array measurements. Preliminary calculations suggest approximate budget closure for heat, but truly careful quantitative estimates will require work in the coming year. Closing the salinity budget may be more difficult, particularly due to large uncertainties in fresh water flux at the sea surface. The open link in closing the momentum budget is estimation of horizontal pressure gradients. As with heat and salt, SeaSoar surveys and moored data offer independent measures of these terms. These calculations should be completed by mid-1995.
Microscale mixing estimates are available for one 1-week and two 3-week periods in the IFA during the IOP. The fluxes from these measurements are not only crucial to closing upper ocean budgets, but must also be parameterized in terms of larger scale indices of stratification and shear to be useful for extending budget computations for the IOP models and eventually for prediction. Work will proceed during 1995 to find useful parameterizations for mixing fluxes. See Figure 9.
The Oceans Group plans to compare budget estimates calculated in various ways over a variety of space and time scales during 1995. Budget studies are useful because they rank the importance of various physical processes in the fluid dynamics of forced upper ocean flow. In particular, we are interested in distinguishing wave response from mixing response. The former is characterized by thermocline depth variations linked with pressure gradient forces, while the latter is characterized by coherence with surface fluxes. Because we have both, prospects are good for making this distinction between purely dynamic and ultimately thermodynamic forced response.
The group plans to continue studies of various processes that govern evolution of upper ocean flow structure, emphasizing mechanisms that are active in the surface mixed layer, the layer of weak but significant stratification just beneath, and the pycnocline beneath that. The very definition of a mixed layer depth is a matter of debate, as surface fluxes of heat and fresh water in the COARE region combined with light winds lead to significant stratification at depths that are normally mixed in other regions of the ocean (e.g., a few meters depth). In addition to changes seen in the top few meters, our observations indicate substantial changes in water properties down to and within the pycnocline. Ship surveys suggest the presence of fronts and flows whose relative and planetary vorticities are comparable, indicating nonlinearity. Moorings separated by 50 km indicate episodes of impressive salinity difference, together with currents so strong and long-lasting that they could be produced only by motions on smaller scales or through large vertical excursions. Investigation of the processes that work to produce the various features detected is anticipated through 1996 and possibly beyond.
Models will be crucial in understanding COARE observations. At present, models of the evolution of the entire tropical Pacific exist at horizontal resolution of roughly 100 km and temporal resolution of a day or so. These are valuable to understanding the large-scale evolution of the western Pacific warm pool, but they are inadequate for processes that take place on smaller scales or for comparison with COARE observations on these scales. Nested models are now being developed to resolve small features within the COARE domain while still following constraints of basin-scale tropical flows. The principal impediment to their application in COARE is lack of an adequate description of forcing on smaller scales. Such forcing fields are to be developed by other groups of COARE investigators and their colleagues at numerical weather prediction centers. Forced ocean response predictions can be made by nested models as soon as the forcing fields become available, anticipated by the end of 1995.
Two other classes of models are anticipated to be of great use to COARE analyses: data assimilation and coupled air-sea models. Data assimilation models are somewhat new to oceanography, and their development and successful use are neither widespread nor straightforward. Despite the relative intensity of COARE observations, they are still terribly coarse in space and time compared to the extent of the COARE domain and the IOP. Assimilation must be pursued, as it offers the only promise of effective interpolation of oceanic data in COARE. Coupled air-sea models are the ultimate objective of COARE studies, as the links across the sea surface of atmosphere and ocean are the primary focus of COARE. Present coupled models are somewhat fanciful and demonstrably in disagreement with major trends of circulation in the two fluids (i.e., ENSO events and their opposite states). Development of realistic coupled models that capture the physics of air-sea interaction is a major modeling goal of COARE from about 1995, when forced ocean models will be run with gridded atmospheric products, through 1997.
COARE offers models an oceanic test to compare predictions with the budgets observed in COARE. Prediction merit can be quantified by explaining the observed variance from COARE field studies. One particularly important area is testing model simulations of SST over time scales of months. Accurate SST predictions over these time scales may prove crucial to the success of coupled models for climate forecasting. Comparison of observed SST with predictions, using ocean models and the accurate COARE dataset on fluxes, ocean currents, and subsurface thermal structure, should provide a benchmark on how well such predictions can be made in coupled models. Comparison of models with observations is often limited, but the extensive COARE dataset makes possible much more rigorous model testing than is now conventional. These activities are anticipated in 1996 and 1997.
To be useful in COARE oceanography research, flux fields must be made available in a timely manner. The needs range from flux time series at points or averages over small regions to gridded products computed from numerical weather prediction centers. Ocean circulation, waves, and mixing studies can make use of flux estimates this year. Delivery of all but small-scale gridded products will require recomputation by the various weather prediction centers in 1995.
While flux fields are of crucial importance to understanding COARE oceanic fields, a description of the upper 20 m or so of the ocean is important to estimating fluxes, since in low wind conditions currents can contribute substantially to the relative speed of air over the ocean. Similarly, stratification at shallow depths is important to overall fluxes due to diurnal cycling. The preparation of an upper-20-m dataset from COARE ship surveys and the few moorings with measurements in this range will proceed through 1995.
While most of the discussions on coupled modeling at the workshop pertained to the coupling of general circulation models, the need to develop mesoscale coupled models became obvious. This grew out of the need to obtain a better understanding of the mesoscale atmospheric response to prescribed SST variations. What are the space and time scales of SST variations for which the atmosphere is sensitive? Studies along these lines would range from 1-D coupled atmosphere-ocean models and mesoscale atmosphere models coupled to 1-D mixed layer ocean models to fully coupled mesoscale atmosphere-ocean models.
Another area of research that needs additional emphasis is ocean data assimilation. The discussions at the workshop made it clear that while atmospheric data assimilation would proceed at a number of numerical weather prediction centers, the outlook was less clear for ocean data assimilation studies and products. A number of the working groups expressed an interest in having access to synthesized fields of ocean data resulting from an ocean data assimilation scheme.
In terms of timelines and requirements, the following chronology (Figure 10) was developed at the workshop.
7.CROSS-CUTTING SCIENCE ISSUES
The science and data issues specific to each group are discussed in Section 4 of this document. This section highlights some overreaching issues that cut across the interests of several or all of the science groups. Although this list of issues is not comprehensive, it was discussed in the final plenary session, and there was consensus agreement on the proposed resolutions. Modifications and clarifications that were raised during discussion have been incorporated. 7.1 SEA SURFACE TEMPERATURE
What to do about the problem of sampling with respect to diurnal cycle?
Should the turbulent fluxes be parameterized using bulk or skin temperature?
[2] Section 4.4.2 authored by Holland, G., B. Mapes, and D. Raymond
[3] Section 4.4.3 authored by Ferrier, B. and J.-L. Redelsperger FF
Richard Chinman, TOGA COARE International Project Office, University Corporation for Atmospheric Research, USA
E. Frank Bradley, Commonwealth Scientific and Industrial Research Organization, Australia
Anthony Busalacchi, National Aeronautics and Space Administration, USA
Charles Eriksen, University of Washington, USA
J. Stuart Godfrey, Commonwealth Scientific and Industrial Research Organization, Australia
David Gutzler, National Oceanic and Atmospheric Administration, USA
Peter Hacker, University of Hawaii/Joint Institute for Marine and Atmospheric Research, USA
Dunxin Hu, Academia Sinica, PRC
Richard Johnson, Colorado State University, USA
Roger Lukas, University of Hawaii/Joint Institute for Marine and Atmospheric Research, USA
Frank Marks, National Oceanic and Atmospheric Administration, USA
Tetsuo Nakazawa, Meteorological Research Institute, Japan
Jean-Luc Redelsperger, Centre National de Recherches Meteorologiques, France
Kensuke Takeuchi, Hokkaido University, Japan
Robert Weller, Woods Hole Oceanographic Institution, USA
11. ACKNOWLEDGMENTS
The TOGA COARE International Data Workshop received support from the National Oceanic and Atmospheric Administration, the National Science Foundation and the National Aeronautics and Space Administration (USA). Co-sponsors included Météo France, ORSTOM (Institut Français de Recherche Scientifique pour le Développement en Coopération) and INSU (Institut National des Sciences de l'Univers), within the framework of the French national program on climate dynamics (PNEDC).
Hardware and software components of the computer environment, so indispensable to the workshop's productivity, were contributed by the following companies:
Published by the TOGA COARE International Project Office
University Corporation for Atmospheric Research
Post Office Box 3000
Boulder, Colorado 80307-3000
USA
February 1995
The TOGA COARE International Project Office is supported by the following U.S. government agencies: the National Oceanic and Atmospheric Administration (NOAA), the National Science Foundation (NSF), and the National Aeronautics and Space Administration (NASA). The Data Workshop was sponsored by NSF, NOAA, NASA, Météo France, ORSTOM (Institut Français de Recherche Scientifique pour le Développement en Coopération) and INSU (Institut National des Sciences de l'Univers).
