outline of talk goal: follow deep convective raining systems in a lagrangian framework to see their...
TRANSCRIPT
ISCCP cloud clusters, TRMM rain clusters, and tropical water and
energy budgets
David DuncanChris Kummerow
ISCCP at 30
Outline of talk
Goal:
Follow deep convective raining systems in a Lagrangian framework to see their environmental impacts and subsequent feedbacks on rainfall
This is similar in scope to Stephens et al. (2004) that centered on the MJO, but generalized to all basins
and deep cloud systems
Scene classification System tracking Environmental effects of tracked systems
Why system speed matters Clouds, SST, water vapor, radiative fluxes
Feedbacks Conclusions
Data Sources
TRMM Microwave Imager (TMI) using GPROF retrieval
TRMM Precipitation Radar (PR)
ISCCP cloud regimes
CMORPH rainfall at 3-hrly, 0.25° resolution
Ancillary fields from SeaFlux, ERA-Interim, and SRB
Global Tropical Moored Buoy Array (GTMBA) for in-situ SST
Scene Classification
Classifying by precipitation (Elsaesser et al. 2010)
Clustering by precipitating cloud top height, rain rate, stratiform/convective fraction of rain
Uses TRMM PR 2A25 data exclusively
Similarity across all basins
Classifying by clouds (Rossow et al. 2005) Clustering by cloud top pressure and optical
thickness into 6 regimes
Uses visible and IR data from ISCCP D1 product
Only available during daytime
Precipitation characteristics of these weather states are explored in Lee et al. (2013)
Scene Classification Using TMI data and a K-means clustering
algorithm, 1°x1° oceanic patches are clustered into three regimes
Scenes need to be 100% ocean, lie within 30°N-30°S, and have at least one pixel with a rain rate of >0.5mm/hr
Rainwater path (RWP), surface rain rate, convective fraction of rain rate are the clustering parameters
Advantages to using passive microwave data for classification purposes
Classification Results
The organized convective class exhibits the greatest fraction of stratiform rainfall, largest surface rain rate, and highest rainwater path
Will focus on organized convective class since it should have the strongest effects on water and energy budgets, despite producing the least total rain
18% of scenes produce 64% of all rainfall; in line with 70-80% of rainfall coming from 10-20% of cloud clusters (Mohr et al. 1999)
How do these match up with ISCCP classes?
ShallowUnorg.
Convective
Org. Convectiv
eFraction of
scenes82% 17% 1%
Mean scene RR
0.18 mm/hr
1.6 mm/hr 7.0 mm/hr
Fraction of all rainfall
36% 48% 16%
Classification Results
Matching TMI-derived precipitation classes with ISCCP weather states:
Organized convective class matches with WS1 in most cases
Unorganized convective class is primarily WS1 and the remainder is mostly the other ‘convectively active’ regimes– WS2 and WS3
Shallow class is closer to the background distribution of weather states, shown in dashed lines on plot
Shallow Unorg. Conv
Org. Conv
WS1 Vigorous deep convection 13.7 56.6 83.6
WS2 Thick cirrus, less vigorous convection
12.5 9.1 3.4
WS3 Isolated, smaller-scale convection
26.4 23.6 9.7
WS4 Thin cirrus 15.4 4.2 0.8
WS5 Scattered cumulus 23.8 3.7 1.8
WS6 Marine stratus 8.0 2.8 0.6
Indian
West Pacific
Central Pacific
East Pacific
Atlantic
CFADs of reflectivity from PR,
shown for the Organized
Convective class
Separated by ocean basin (shown
below)
Left column is stratiform pixels
Right column is convective pixels
Demonstrates the similarity and
consistency of vertical structure for
TMI-derived precipitation clusters,
regardless of location Ind W. Pac C. Pac E. Pac Atl
Tracking Method
Other studies have used outgoing longwave radiation (OLR) or brightness temperatures to track deep convection in the Tropics– why not use rain itself?
Catalogue groups of raining pixels that are contiguous in time and space and exceed a rain rate threshold
Latitudinal range of 15°N-15°S
7mm/hr is the rain rate threshold used, the same as the mean scene rain rate from the organized convective regime
Rain rates Tracked groups OLR
Tracking Results
Great similarities are found in the propagation characteristics of systems in each basin
West-moving systems are more likely in every basin
Atlantic basin in a slight outlier, with many more west-movers and more systems that move quickly
In the remaining analysis, systems are separated by speed into fast (>6m/s) and slow (<2m/s)
All tracked systems, 2003-2009, separated by basin
IndWPa
cCpacEPac
Atl
IndWPa
cCpacEPac
Atl
IndWPa
cCpacEPac
Atl
Tracking Results
Putting together the classification and tracking methods—what are we tracking?
Co-locate tracked systems with TMI-derived precipitation classes
55% of tracked systems are co-located with the TMI-derived organized convective class—why not higher?
TMI-derived regime
Tracked system frequency
Shallow 2%
Unorg. Convective 43%
Org. Convective 55%
Contoured CMORPH rain rate of a sample tracked system
Environment Analysis Mean latitude and longitude are
computed at every time step, then co-located with closest grid box from each dataset
A time series is extracted for each tracked point for analysis of ancillary fields before and after passage of the system
Time series are composited together to show the mean evolution of each field
All of the following analysis is separated into fast and slow systems, and by ocean basin
t = 1
t = 2
t = 3
SST
Rain rate
Environment– Clouds
OLR provides a good check on the method of co-location
Inter-basin variability is significant, with slow systems in the W. Pac and Indian ocean basins definite outliers
Dashed = fast Solid = slow
IndWPa
cCpa
cEPa
cAtl
Environment– Water Vapor
Total precipitable water (TPW) peaks at system passage (lag=0), with fairly symmetrical increase and decrease before and after system passage
Essentially no net effect found in TPW, so deep convection neither dries nor moistens the atmospheric column at a 72hr lag, common for all speeds and basins
Magnitude of change largely agrees with results from a separate method (Masunaga 2012)
Most moistening, ~4mm of the 6mm total, occurs between 500-850mb
IndWPacCpacEPac
Atl
Dashed = fast
Solid = slow
Environment– Water Vapor Water vapor convergence is calculated from
ERA-Int wind vectors and specific humidity
A sharp decrease in water vapor convergence coincides with sharply decreasing rain rates after system passage
Due to a conservation of water, every grid box should obey the equation below
Reanalysis water vapor convergence is not strong enough to balance the water budget, though rain rates in reanalysis are underestimated
IndWPa
cCpacEPac
Atl
Environment– Evaporation
Evaporation is directly proportional to surface latent heat flux (LHF), which is taken from SeaFlux
LHF is a function of SST, near surface humidity, and surface winds
Speed and basin both play significant roles
The generally larger size of slow systems aids in creating a larger circulation, causing stronger surface winds and higher LHF
IndWPa
cCpacEPac
Atl
Environment– SST
Diurnal nature of heavy rainfall in the Tropics, peaking in early morning, is visible and similar for all basins
Decreases of 0.1-0.3°C witnessed, dependent upon basin and system speed
Total difference (-72hr to +72hr SST) shows that slow systems have a bigger impact on SST in every basin
SST recovery rates are quite basin dependent, likely due to differences in ocean mixing and mixed layer depth Ind
WPacCpacEPacAtl
Dashed = fast
Solid = slow
Environment- Radiation
What is the main driver of the observed drop in SST?
Deep clouds cause ~200W/m2 drop in net surface radiative flux
Changes in SW Down accounts for almost all variation in the net surface flux
Diurnal signal is again quite noticeable
Are radiative fluxes the most important driver of observed SST variability?
IndWPa
cCpacEPac
Atl
IndWPa
cCpacEPac
AtlDashed = fast
Solid = slow
Radiation and SST
Gridded SST products use interpolation for raining grid cells. Does SeaFlux SST match the environmental evolution seen by in-situ buoy measurements?
To ascertain the degree to which radiative changes cause the observed drop in SST, we need to use buoys because SeaFlux shows a fundamentally different evolution
Radiation and SST Shown below, in-situ measurements of SST match very well with radiative
changes
In a simple scale analysis, changes in surface radiative flux are integrated to give an equivalent ocean mixed layer depth (MLD):
Integrating from -18hrs to +6hrs lag, the period of greatest SST decrease, yields a calculated MLD of 15m; integrating from -18hrs to +36hrs gives a MLD of 45m, very close to the climatological mean MLD of ~40m
Feedbacks
Following a slow-moving system, the environment is suppressed—nearly 0.1°C lower SSTs for multiple days. Does this have a noticeable impact on subsequent development of deep convection?
Use rain rate information from CMORPH to see if there’s a signal present
Feedbacks Need to look at rain rates a few days after the system’s passage to
escape the ‘persistence’ of slow-moving systems’ high rain rates in the PDF
The signal is quite noisy due to the low frequency of the highest rain rates
Environments affected by fast systems are more likely to exhibit high rain rates 4 or 5 days after system passage
PDF difference = (PDF[slow] – PDF[fast])/PDF[slow]
Areas on the PDF dominated by slow systems are warm colors and fast systems are cold colors
Conclusions
TPW WV Conv. SST Evap Net Sfc. Rad.
Fast
Slow
Conclusions
Deep convective systems in the Tropics exhibit great similarity and consistency in vertical structure and propagation characteristics in all ocean basins
Differences in cloud fields affect various elements of local water and energy budgets, with system propagation speed of key importance
For tracked systems, SST drops 0.1-0.3°C and TPW increases symmetrically ~5-7kg/m2 in the mean evolution
Depressed SSTs persist in environments affected by slow-moving systems, impacting the likelihood of heavy rainfall days later