Warning: This writeup was done to help organize
my thoughts prior to the presentation.
The actual presentation may have been quite different!
Acknowledgements: Roger Lukas (UH) and Joel Picaut (ORSTOM)
Paul Freitag (PMEL), Margie McCarty (PMEL), Craig Nosse (UH)
This work I will be presenting today is described in a manuscript which my co-author, Mike McPhaden, and I have submitted to JGR-Oceans. Feedback is very helpful at this stage so please jump in.
Before getting started, I would like to acknowledge a Roger Lukas and Joel Picaut for providing SEACAT data from nearby moorings. This data was used to evaluate the horizontal advection term. As I will show zonal advection is critical in the surface salinity balance. Also, thanks go to Paul Freitag and Margie McCarty who processed the data collected by PMEL, and Craig Nosse who was responsible for the data collected by UH. An immense amount of work has been put into post-processing the salinity data.
As advertised by the title, the main part of this talk will be on the upper ocean salinity balance at 0,156E for the 32 month period from September 1991 through April 1994. Because most of the data are hourly time series, I will discuss the balance in 4 different timescales: a) hourly to daily, b) daily to monthly, c) monthly to the record length (32 months), and d) the record length mean balance.
One reason this is such an exciting piece of work is that open ocean rainfall measurements and longterm, hourly time series of salinity are extremely rare. There are many challenges and potential sources of error with these types of observations. Indeed, the results of the balance at first surprised me. So I will spend a portion of this talk discussing the calibration and validation of the precipitation and salinity data.
To get oriented, here are 3 panels showing the longterm mean SST, SSS and rainfall in the tropics. The study location is here at 0,156E, in the western equatorial Pacific Warm Pool. The Warm Pool is a region, usually in the western Pacific, where SST>28C. From the bottom panels, we see that this is also a region of intense rainfall (accumulations of 2-5 m/yr) and low SSS. One of the major goals of TOGA COARE was to understand how these fields are related.
This is a time-longitude plot of the monthly GOES Precipitation Index along the equator. Notice that there is alot of variability both zonally and temporally. In general, the rainfall is maximum over the eastern edge of the warm pool, and the rainfall migrates zonally on intraseasonal timescales associated with the Madden-Julian Oscillation, and on intra-annual timescales associated with ENSOs. Although we can't see it with monthly data, there is also a 2-day mesoscale variability in the rainfall associated with the diurnal cycle in SST. How does the ocean respond to this variability in precipitation?
This figure shows the COARE enhanced monitoring array which was in place from August 1991 through April 1994. The budget analysis I will be describing uses data from this central mooring at 0,156E as well as salinity data from nearby ATLAS moorings with SEACATS.
This dashed line shows the Intensive Flux Array region. During the Intensive Observational Period from November 1992 through March 1993, there were 7 aircraft and 12 ships making measurements. All ships were heavily instrumented with multiple rain gauges of various kinds, etc... 2 ships had doppler radar to measure rainfall.
Right now there's a raging debate going on about the reliability of
the ORGs. Dave Short and company have recently done a comparison of
the rainrates at 2S,156E during the COARE IOP and found that the
2S,156E buoy ORG measured rain rates a factor of 2.5 higher than those
measured by the radars. The shipboard ORGs also measured high
relative to the radar.
At the central location, the surface data included hourly wind speed and direction, air temperature, relative humidity, incoming shortwave radiation, rain rates from an optical rain gauge, and the 1-M sea surface temperature and sea surface salinity. Using this hourly surface met data with the COARE v2.5b bulk algorithm, I was able to compute the turbulent fluxes of heat, moisture and momentum.
The hourly subsurface data included temperature to 500m, salinity to 200m, and zonal and meridional velocity to 250m from an ADCP.
Because of the Radar intercomparison, I would like to spend a few minutes discussing the ORG measurement.
As can be seen on this diagram, ORGs are a short black box that sends an LED beam of light through an opening. Rain falling through this openning casts a shadow. This scintillation rate is what gets converted into a rainrate.
There are several ways in which the measurement can be contaminated: a background voltage..., splash from the instrument housing, rainfall slant during driving rain, and variations in the raindrop size distribution.
During COARE we had ORGs at 6 sites and thus could compute spatial correlations. Correlations computed with hourly data are all insignificant, but as the averaging time increases, the correlation increases. Correlations for the 10-day averaged ORG rainrates are significant even at 10-degrees of longitude separation.
This is consistent with the multiscale variability observed by others such as Chen-Houze-Mapes (1996): Individual mesoscale convective cloud clusters often have a westward propagating envelope with near 2-day periodicity, embedded within a larger scale eastward propagating intraseasonal oscillation cloud envelope. Thus the large-scale patterns become more evident with increasing temporal averaging.
To further test the accuracy of the ORG time series, we compared pentad averages of the 0,156E ORG data to three different satellite estimates of rainfall: the GOES precipitation index (GPI), special sensor microwave/imager (SSM/I) precipitation, and microwave sounding unit (MSU) precipitation. GPI data are derived from geostationary infrared measurements of cloud top temperature and polar orbiting satellite measurements of outgoing longwave radiation flux which are regressed against direct measurements of rainfall [Janowiak and Arkin, 1988, data provided by Bob Joyce of NOAA/NCEP]. SSM/I and MSU data are derived from different frequency channels of polar orbiting satellite microwave observations of emmission, scattering, and absorption from raindrop-sized hydrometeors. The SSM/I data were provided by Ralph Ferraro of NOAA/NESDIS, and MSU data were provided by Roy Spencer of NASA/MSFC.
Over the record length of the 0,156E ORGs, the means were: ... rxy = ...
SUMMARY OF RAIN I/C:
Now lets talk about salinity. PMEL, UH and ORSTOM had several deployments of Seabird SEACAT temperature and conductivity sensors at sites along the 156E longitude and at sites along the equator.
At 0,156E, 2s,156E and 0,165E, the SEACATs were placed all the way down below the high salinity core. At other sites, the sensors were placed in the upper 30-100m.
Most deployments were for 6-9 months, although a few deployments were longer.
While Seacats which are in the field for less than 6 months generally have an accuracy of 0.02 psu, many of the COARE seacats were in the water for longer. Sources of errors include: Electronic drift, scouring, biofouling, mismatch of T&C response time.
We have 3 different ways to calibrate them: Pre/Post calibrations in lab, in situ calibrations with CTD, and stability analysis when two or more sensors were placed on line within the mixed layer.
Freitag et al. (1998)
All PMEL data have been completed. The UH data is still in the works. When this hasn't been done, I've assigned an error value of .1 psu.
This slide shows the timeseries precipitation minus evaporation and of the salinity stratification as measured by the top 3 sensors (generally at 1m, 5m, and 10m, although during the first deployments it was 1m, 10m, and 30m).
While there is variability on a full range of time scales, as I will show in the next slide it is clear that there is correlation between these time series.
These 3 panels show the cross correlation between dSdz and EMP as function of depth for the 3 timescales: hourly-daily lowpassed filtered, daily lowpass - 29-day triangular monthly lowpass filtered, and monthly-mean (29-day triangular lowpass filtered).
As shown in the left panel, for hourly-daily timescales, the maximum stratification comes about 1 hour after the peak rainfall. At that point, mixing kicks in and brings salty water up from below. Consequently we can expect a near steady state balance between the rainfall and mixing. The slight offset in timing gives rise to fresh spikes associated with the rain.
Notice that this correlation in fact increases for increasing timescale.
To look at these rain events in a little more detail, I will show you the salinity profile time series as simulated by the PWP 1-d mixed layer model. This case study event was chosen because 1) it was raining and 2) the currents were weak. In fact it was very difficult to find such an event because normally the currents on the equator are very strong.
The model physics are 1-dimensional: surface forcing and mixing determined by 3 criteria (gravitational instability, bulk Ri < 0.65, Ri < 0.25). In order that the momentum balance doesn't become unbounded, a 7-day damping term is included. The model is initialized with the observed T/S/U&V profiles on October 26? and then forced for 24 hours with the observed surface heat, moisture, and momentum fluxes.
Notice how the freshwater stratification builds up and then as mixing occurs, it gets injected into the deeper waters. This is also observed as can be seen in the lower panel. The steppyness is due to the vertical resolution of the mooring. When the model is subsampled, we see a similar pattern.
There are differences however: the PWP freshwater is more surface intensified. This is because the model's mixing is not strong enough.
When I integrate down to 40m, in fact, the cummulative change in the salinity in both the model and observations, matches what is expected based on the amount of rainfall. Put another way, this case study shows yet again, that the magnitude of the rainfall is consistent with the response of the ocean.
If the ocean was like a bathtub, then we would expect that the rainfall would be highly correlated with the salinity tendency rate. In fact it's not.
terms, MLD, Sa, how I measure each term, residual
High salinity core, surface layer, MLD, thermocline...
Sa captures low frequency variability, but filters out the spikes in the sea surface salinity associated with rainfall.
balance is... dSadt not well correlated with E-P term. Zonal Advection is big and fairly well correlated with dSadt.
What's going on: big largescale fronts!
big fronts! but highly variable v.
Record length mean balance is near steady state balance between E-P term and residual "mixing". Zonal advection is on average a positive term due to the correlation of u and the zonal gradient: when it is strong westward current (SEC), you tend to have salty water to the east. When the u is eastward as during WWB, you tend to have freshwater to the east. This is consistent with the concept of a fresh pool which migrates eastward in association with ENSOs.
Optical Rain Gauge rain rates are ~20% higher than other rain estimates. The "Ocean Rain Gauge" was consistent with the ORG during a moderate rain event.
SEACAT salinity required extensive post-processing for deployments longer than 3-6 months.
Freshening due to rain was largely balanced by mixing. Changes in SSS were predominately due to zonal advection of large-scale salinity gradients in the western equatorial Pacific.
The zonal SSS gradient is sharp and positive on the eastern edge of the warm pool, and weak and negative on the western edge of the warm pool. This "fresh pool" shifts zonally associated with ENSO.
The "fresh pool" structure is probably formed by gradients in rain associated with the Walker Cell, and sharpened by convergences in the zonal currents.
Meghan F. Cronin
Pacific Marine Environmental Laboratory
7600 Sand Point Way NE
Seattle, WA 98115 USA
Meghan Cronin's Home Page
DOC | NOAA | OAR | PMEL