SE PACIFIC BL CLOUDS OVER THE EQUATORIAL UPWELLING REGION.

SE PACIFIC BL CLOUDS OVER THE EQUATORIAL UPWELLING REGION

There is considerable evidence that the subtropical and tropical climate is strongly influenced by extensive boundary layer clouds off the coast of the South American continent. It is therefore imperative that the processes controlling the development, maintenance and dissipation of these cloud fields is better understood in order that their effects can be accurately modelled in coupled ocean-atmosphere models.
Strong equatorial upwelling in the eastern Pacific causes the ITCZ to be displaced away from the equator. For reasons related to the NW-SE direction of the North and South American coastlines this displacement results in a northwards-displaced ITCZ. There is a resulting cross-equatorial Hadley cell which is thought to be strongly modulated by the presence or absence of boundary layer cloud over the cold SST region. It is postulated that the stratus/stratocumulus cloud can cool the SST in the upwelling region, thereby strengthening the Hadley circulation. The strengthening cell increases the subsidence over the cold region which can further enhance the boundary layer cloud. Potentially, this gives rise to an instability which is, as yet, poorly understood.
Crucial to our being able to accurately model the strength of this feedback is the ability of models to maintain a persistent cloud deck over the cold regions of the south east Pacific. Currently, this appears to be a particular problem for many climate models to achieve. In the north east Pacific boundary layers evolve over generally increasing SSTs as they move westwards and become the easterly trades. In contrast, boundary layers in the south east Pacific stratus region are initially warmed from below as they advect equatorwards, but then experience a cooling as they pass over the east Pacific cold tongue (see Fig. 1 below). A very rapid warming then ensues in the cross-equatorial flow. The dynamical (and possibly microphysical) processes are therefore somewhat different to stratocumulus transition regions observed elsewhere. Detailed observations of these clouds are lacking, a shortfall that the planned 2001 EPIC (East Pacific Investigation of Climate Processes in the Coupled Ocean-Atmosphere System) observational campaign seeks to redress.

The above image shows a snapshot of the cloud optical depth from MODIS to the west of the Galapagos Islands (on equator at 91W) together with the TRMM TMI SST field. The width of the image is 1030 km and the height 1750 km. The sea surface temperature front just north of the equator is clear and is modulated by tropical instability waves. Low level cloud is suppressed as the boundary layer passes across the equatorial cold tongue but then forms again after the air has passed over warmer waters. Notice the strikingly different morphology of the cloud fields to the north and south of the front.

Fig. 1: Climatological trajectories for Sept/Oct 2000 (solid, circles are spaced 1 day apart). The trajectories are generated using the two-monthly mean wind fields taken from the NCEP reanalysis. The trajectories are started along the 10S parallel at 1 degree intervals between 100W and 90W and therefore represent the general advection of the south east pacific stratocumulus-topped boundary layer. Also plotted are the mean SST contours for the same period.

Fig. 2: Sea surface temperature field for July 1984 (from two weeks of satellite data). Warm water is red and cool water blue. The east Pacific cold tongue is clearly visible. (Figure taken from the EPIC Science and Implementation Plan)

Fig. 3: Schematic of the troposphere across the equator in the cool season (July-September). (Figure taken from the EPIC Science and Implementation Plan)

Fig. 4: Time series of observed and reanalysis parameters along the mean [90W-100W, 10S] climatological trajectory. All data are means (or medians where stated) from all MODIS scenes during September/October 2000. Error bars show +/- 1 standard deviation of results from the 11 different mean trajectories, and therefore represent the variability in the mean values across the initial 100W-90W parallel.
From left to right, top to bottom:
Latitude/longitude plot along trajectory
Mean cloud fraction (from 256x256km MODIS scenes, 16-17 UTC). Also shown are ISCCP low cloud amount climatology for the same time of day and for September/October (1984-1993) and EECRA low cloud amount (types 1,2,4,5,6,7,8) for all daytime observations cold season mean 1954-1992 (see EECRA Low cloud type over the global ocean)
Mean liquid water path (LWP) (from 256x256 km MODIS scenes). Also shown is the ISCCP water path taken from the same subset as cloud amount above.
Surface latent heat flux (NCEP)
Temperature advection (NCEP+Reynolds blended SST)
Lower tropospheric stability
Median ratio of (mean LWP to standard deviation of LWP)^2 for all 256x256 km cloud scenes from MODIS with cloud fraction > 0.1
Median integral scale (characteristic scale of mesoscale convective cells) of all cloud scenes with cloud fraction > 0.5. For method of calculation from MODIS data see .
Subsidence speed at 850 hPa (NCEP)
Sea surface temperature (Reynolds blended product) and cloud top temperature (ISCCP, see under cloud fraction above for details)
Mean cloud particle effective radius (MODIS). Also shown is the AVHRR retrieval data for October 1988 (a negative cold tongue index (CTI) year as is 2000) from Kawamoto et al. (2001)
Mean effective droplet concentration (MODIS)
Questions:
Why do the ISCCP and MODIS cloud amounts diverge north of 4S-5S? Taking only ISCCP data from negative CTI years (e.g. 1988) does not markedly improve the agreement. For the particular analysis of MODIS scenes presented here, it is required that the mean cloud-top temperature in the 256x256km scene is greater than 270K. This is because a known bug in the MODIS calibration/processing routines leads to very noisy cloud top temperature data. Rejected scenes due to cold cloud presence runs at around 20-25% at 5S increasing to 40-50% at 5N. However, cloud top temperature in MODIS scenes analysed has some inaccuracies which result in warm clouds being classified as cold clouds. If the MODIS scenes rejected due to misclassification of cloud top temperature have lower cloud fractions than those correctly classified then this could suggest a positive bias in the MODIS cloud fraction. My thoughts here are that MODIS is more likely to misclassify clouds when the true cloud top temperature is close to 273K and broken. However, it does not appear that, in the region of interest, there is a significant correlation between mean cloud top temperature and cloud fraction. Hopefully, this issue will be resolved when the accurate cloud top temperature data from MODIS becomes available.
In addition, ISCCP high and medium cloud amounts along the trajectories south of 5N are less than less than 1% and 5% respectively (SEE FIGURE), and so it is unlikely that ISCCP low cloud amounts will be strongly affected by the presence of cloud above the boundary layer.
The form of the EECRA low cloud amount-latitude curve is similar to that from MODIS, with initially high values at 10S falling to a minimum at around 3S-5S and rising to a peak at 2N-3N. They are 0.05-0.07 lower than MODIS values between 7S and 5N, but this may reflect the fact that EECRA cloud amounts are averaged over the entire cold season (July-November inclusive) while the peak cloud amount for this region is found in September-October. EECRA cloud amounts are derived from all daytime observations while MODIS are taken at around 10-11am local time. It is interesting that all three measures agree very well at 10S. The diurnal cycle (the amplitude of the first harmonic) of low cloud amount from ISCCP for the same location is some 5-15% with a peak cloud amount at around 2am local time (Rozendaal et al. 1995). A simple sinusoidal model of cloud amount diurnal variation using this phase gives 10-11am cloud amounts that are very similar to the daylight mean cloud amounts (which are assumed here to be 7am-7pm local time). This suggests that the diurnal cycle may not be responsible for the difference between EECRA and MODIS cloud amounts.
Rozendaal et al. show that ISCCP VIS/IR (not used here) retrievals of cloud amount generally exceed those of ISCCP IR (used here) by approximately 0.1-0.2 in the region of interest, which could explain the discrepancy. All three measures of cloud amount agree very well at the start of the trajectories and so the VIS-IR ISCCP difference would need to be small here if this particular effect is to blame. It is unlikely in the region of interest that the presence of fog or very low cloud (such as that under sub 500 m inversions) would be responsible for the VIS-IR difference. The presence of smaller cloud elements around the equator could be responsible. There is some evidence from the MODIS dataset that the clouds become smaller to the north of 5S (see integral scale in figure), although the decrease does not appear to be large (<20%).
Why are MODIS effective radius means around 4-5 microns larger than the corresponding AVHRR climatology values? I don't think that this can be explained using the diurnal cycle of effective radius, and probably is more likely to represent early retrieval problems with MODIS. Given this, it is probably unwise at present to rely upon MODIS effective radius retrievals.
Kawamoto, K., Nakajima, T. and Nakajima, T. Y., 2001: A global determination of cloud microphysics with AVHRR remote sensing, J. Clim., 14, 2054-2068.
Rozendaal, M. A., C. B. Leovy and S. A. Klein, 1995: An Observational Study of Diurnal Variations of Marine Stratiform Cloud. J. Clim., 8, 1795-1809.