The laborious task of identifying artificial change-points (abrupt changes in time series) has been completed. This procedure entailed visual examination and consideration of all data and a variety of ancillary information on a station by station, level by level basis. The tools employed were a unique compilation of station history meta-data (created at ARL), diurnal differences in the temperatures, vertical temperature structure, and non-temperature measures of climate variability, among others.
A complex algorithm for adjusting the data has been developed and coded, along with related graphical tools. The scheme employed attempts to make adjustments in such a way as to preserve the natural vertical coherence of the temperature data. This is accomplished by making adjustments using "reference levels" from the same station (i.e., vertical levels which do not require adjustment and, excluding the effects of any artificial discontinuities, correlate highly with the level to be adjusted). After a level has been adjusted it can potentially serve later as a reference level. This scheme is an attempt to remove the artificial component of change while retaining the natural component. In a minority of cases when suitable reference levels are not available, adjustment is made by a method which assumes that the difference in mean temperature across the change-point is entirely artificial.
After completing the identification of artificial change-points, the newly developed adjustment method was applied to the radiosonde station data. This process was repeated more than once in order to create different "scenarios" of adjusted data based on either various levels of confidence in the identification of artificial change-points or other procedural differences. Such an approach allows for testing the sensitivity of low frequency variability (especially trends) to some of the details of the identification and adjustment processes. Having completed these initial steps, various analyses involving internal sensitivity and consistency checks, comparisons with other observed datasets, and comparison with model (GCM) data were started.
6.2 ANALYSIS OF DATASETS BASED ON SATELLITE OBSERVATIONS
Despite its obvious importance, there have been relatively few observational studies of the long-term variations or trends in water vapor. Moreover, there have been few, if any, attempts to compare observed trends against model-predicted trends, or to inter-compare observed trends from different observing systems. To address this issue, multiple satellite records of tropical-mean water vapor were compared with a general circulation model (GCM) simulation to assess our ability to monitor and predict low-frequency changes in W (1741). Particular attention was focused on the drying between 1979-1995 recorded by a TOVS-statistical retrieval which is calibrated to radiosondes. A version of the GFDL GCM integrated with observed time-varying SST forcing, as well as observational estimates based on microwave and TOVS-physical retrievals, show no sustained drying in the last two decades (Fig. 6.1). This discrepancy is consistent with a previous assertion that the TOVS-statistical
algorithm is vulnerable to radiosonde instrumentation changes over this period, which introduce an artificial drying trend into the retrieval. The comparison also demonstrated excellent agreement between the observed variations in tropical water vapor over the 1979-1999 period and those predicted by the GCM integration.
6.2.2 Reconciling Surface and Satellite Temperature Records
Figure 6.2 compares the globally averaged difference in the lower tropospheric temperature between MSU and an objectively-analyzed radiosonde dataset. The MSU record exhibits a distinct cooling trend over the period 1979-1994 relative to the radiosonde data. Also shown in this figure is the expected spurious trend in the MSU lower tropospheric temperature record based upon known changes in satellite orbit. A notable similarity between the two time series is discernible, suggesting that there still remains a spurious downward (cooling) drift in the MSU temperature record.
Having completed the initial testing of this new coupled model, several ensembles of experiments were conducted using different spatial configurations of the ocean model. Since the main goal of this project is to assess the impact of El Niño-Southern Oscillation (ENSO) on global sea surface temperature (SST) variations, all of the experiments entail prescription of SST anomalies in the deep tropical eastern Pacific (DTEP, defined as the region between ~173°E and the South American coast, and within ~15°N-15°S) as observed in the 1950-1999 period. These experiments will henceforth be referred to as the Deep Tropical Eastern Pacific Ocean-Global Atmosphere (DTEPOGA) runs.
The "standard" experiment uses the SST forcing in DTEP, with the MLM employed elsewhere in the World Oceans. A sixteen-member ensemble of this experiment has been completed, with each run being initiated from an independent set of atmospheric conditions. To compliment this model scenario, a set of eight "control" integrations has been completed in which the MLM was not used; instead, seasonally varying climatological SSTs (computed from the "standard" DTEPOGA runs) were specified outside of DTEP. Finally, an eight-member ensemble of the "North Pacific" (NP) experiment has been completed. In this scenario, the MLM was employed only in the North Pacific Ocean, and observed climatological SSTs were specified elsewhere outside of DTEP. The purpose of the NP experiment is to isolate the influences of ENSO on SST variability in the North Pacific through changes in the overlying atmospheric circulation. The data generated by the coupled model has been distributed to external collaborators. In particular, a large subset of the model output was sent to CDC.
The DTEPOGA experiments have been analyzed in conjunction with a previous series of integrations based on coupling of the same R30L14 atmospheric GCM to an oceanic mixed layer model with a constant depth of 50 m. The domain of SST prescription in the model runs using fixed mixed layer depth extends across the entire width of the tropical Pacific between ~25°N and ~25°S. Integrations with the 50-m mixed layer incorporated at all ocean grid points outside of the tropical Pacific are referred to as the TOGA-ML runs; whereas the corresponding control experiment (i.e., with climatological SST conditions prescribed outside of the tropical Pacific) is labeled as the TOGA experiment. Four-member ensembles of the TOGA-ML and TOGA experiments have been completed for the 1950-1999 period.
6.3.2 Atmospheric Bridge Linking ENSO to SST Variability in the North Pacific and North Atlantic
Detailed diagnoses of the monthly evolution of the surface heat budget during ENSO episodes at selected sites have been performed (nk). The simulated SST response exhibits a 1-2 month delay relative to temperature changes in the overlying atmosphere. This lag relationship results in a polarity change of sea-to-air gradients in temperature and water vapor mixing ratio in late winter. From late autumn through mid-winter, the action of the climatological wind on these gradients results in enhancement of the developing SST anomalies. In the months thereafter, the reversed gradients are accompanied by attenuation of the SST signal. Shortwave radiative fluxes associated with variations in cloud cover play an important role in SST variability at some of the subtropical sites.
The nature of sea-air feedbacks in the extratropics has been studied by contrasting the output from the TOGA-ML and TOGA experiments (6.3.1). Incorporation of sea-air coupling in TOGA-ML is seen to enhance the persistence of the ENSO-related atmospheric anomalies in the extratropics through late winter and early spring. Comparison with results from previous studies on midlatitude sea-air interactions suggests that part of the atmospheric signal in TOGA-ML may be attributed to forcing from extratropical SST anomalies produced by the atmospheric bridge mechanism.
Preliminary diagnosis of the output from the DTEPOGA experiments (6.3.1) confirms the principal findings based on the TOGA-ML and TOGA runs. Comparison between the results from the "standard" and "NP" scenarios (6.3.1) reveals that the atmospheric signals in the latter scenario are stronger, thus suggesting the presence of negative interferences from atmospheric responses to SST anomalies generated in the Indian Ocean or tropical western Pacific through tropical atmospheric bridges.
6.3.3 Impact of ENSO on Monsoon Systems in East Asia and Australia
Diagnosis of both the TOGA and "control" DTEPOGA experiments (6.3.1) illustrates that the GCM simulates the above-mentioned ENSO influences on the East Asian winter monsoon with a high degree of fidelity. In accord with the observations, the composite model anomaly near the surface in the control DTEPOGA integration is characterized by an anticyclone over the Philippines during warm events (Fig. 6.3). The southerly flow to the west of the anomalous high pressure center leads to warmer and wetter conditions than normal. The reduced wind speeds and cloud cover over the South China Sea result in considerable warming of the oceanic mixed layer, whereas the enhanced northeasterly circulation to the east of the anomalous anticyclone is conducive to oceanic cooling. Both the atmospheric and oceanic anomalies over the subtropical western Pacific exhibit a tendency to migrate eastward during the spring season.
Also noteworthy in the patterns of Fig. 6.3 is the appearance of the anticyclone off the eastern Australian coast, which exerts a considerable influence on the summertime circulation in that sector. The salient symmetry about the equator of the anomalous features over the western Pacific is suggestive of a Rossby-wave response to reduced near-equatorial convective heating in that sector, when the ascending branch of the Walker Cell is displaced eastward during warm ENSO events.
6.3.4 Modulation of Tropical Transient Activity by ENSO
growth/decay characteristics of these disturbances have been identified
using lagged regression statistics, variance analysis, and complex empirical
orthogonal functions. It has been demonstrated that the GCM is capable
of reproducing the essential regional and seasonal characteristics of the
observed transient activity in the tropics. The results also indicate that
the occurrence of ENSO has noticeable effects on the propagation speed,
as well as locations of growth and attenuation, of the these disturbances.
Further analysis suggests that such influences are attributable to spatial
shifts in convective zones accompanying the development of ENSO events,
and to changes in the atmospheric steering flow when such events occur.
The relative contributions of enhanced persistence and positive dynamical air-sea feedbacks to the springtime atmospheric signals associated with ENSO will be further investigated using the suite of DTEPOGA integrations. The nature of the atmospheric responses in the North Pacific to SST forcing in the Indian Ocean and western tropical Pacific will be studied. The role of tropical Atlantic SST anomalies in atmospheric variability in the North Atlantic sector will also be analyzed.
The relationships between ENSO-related processes and the Asian-Australian monsoon systems will be further delineated using the available observational and model datasets. Particular emphasis will be placed on the origin of the variability of the air-sea coupled system on biennial time scales, and on extratropical influences of monsoon fluctuations during the summer season.
The characteristics of air-sea interactions accompanying the passage of atmospheric disturbances on synoptic and intraseasonal time scales will be documented using output from the TOGA-ML and DTEPOGA experiments. The teleconnections between extratropical flow features and tropical Madden-Julian Oscillations will be examined.
6.4 DEVELOPMENT OF WEB-BASED TOOLS FOR VISUALIZING AND EVALUATING MODEL OUTPUT
This project has been substantially reorganized to include applications to oceanic datasets, and to incorporate the advanced features in the data display and management toolkit maintained by the Thermal Modeling and Analysis Project at the Pacific Marine Environmental Laboratory (PMEL). A long-term partnership between GFDL and PMEL has been established to facilitate this collaboration. The latest version of the PMEL Live Access Server (LAS), which is a web-based interface to data sharing, visualization and analysis, has been successfully installed and tested at GFDL. Plans to link existing model evaluation tools to this framework have been formulated. Enhancements and modifications to LAS aimed at improving its functionality in displaying both meteorological and oceanographic fields have been made. It is envisioned that the LAS facility, in conjunction with the generalized Internet-wide approach to data sharing through the Distributed Oceanographic Data Systems (DODS) network, will constitute a powerful and convenient tool for accessing and visualizing datasets from research centers throughout the world, and for evaluating the performance of various FMS experiments conducted at GFDL.
6.5 GFDL/UNIVERSITIES COLLABORATIVE
PROJECT FOR
MODEL DIAGNOSIS
Significant achievements by GFDL investigators associated with this collaborative project during the past year include the completion of a comprehensive suite of 50-year (1950-1999) long experiments in which a variable-depth ocean mixed layer model is coupled to the atmosphere under different scenarios (6.3.1), improved understanding of the role of air-sea coupling in enhancing the persistence of extratropical atmospheric anomalies associated with ENSO (6.3.2), and identification of the impact of ENSO on tropical monsoon systems and transient activity (6.3.3, 6.3.4). Progress has also been made in summarizing the principal findings of this collaborative effort in a set of review articles to be published in a special issue of a journal.
Efforts will be made to sustain this collaboration through identification of appropriate scientific problems and interested scientists, and consultation with funding agencies.
*Portions of this document contain material that has not yet been formally published and may not be quoted or referenced without explicit permission of the author(s).