1. National Aeronautics and Space Administration
Jet Propulsion Laboratory
California Institute of Technology
Pasadena, California
Advancing global precipitation measurement in light of A-Train observations
Ali Behrangi
NASA Jet Propulsion Laboratory, California Institute of Technology
Atmospheric Physics and Weather Group
M. Richardson, G. Huffman, R. Adler, G. Stephens , M. Lebsock
Also thanks to :
Matthew Christensen, Norm Wood and Tristan L'Ecuyer, John Haynes
Bjorn Lambrigtsen, Eric Fetzer
A-Train Symposium 2017, April 2017
Pasadena, CA

2. Estimated relative bias error
Precipitation measurement – status
GPCC total number of stations per 1 deg grid
(April 2009)
Estimated relative bias error
Estimated relative bias error calculated based on spread of climatology products over their mean, which
exceeds 50% at higher latitudes, especially over ocean. The figures are from Adler et al. (2012). The actual errors
are likely higher than that shown here because the climatology products used in the calculation are not independent
1mm/d
29 w/m2
Adler et al. (2012).

3. Zonal precipitation phase frequency from CloudSat
Zonal precipitation occurrence distribution is not symmetric
(snow+ mixed phase) is the dominant type of precipitation Poleward of ~50 deg latitude
30% of global precip. occurs poleward 40 lat. (23% for 45lat)

4. Regional precipitation frequency from CloudSat
LAND
OCEAN
Snowfall
Mixed phase
Snowfall
Mixed phase
Behrangi et al. Water resources research (WRR; 2014)
Behrangi et al. J. of climate (2014)

5. Scaling and sampling consideration
Behrangi et al. (JGR; 2012)

6. Combined precipitation : CloudSat, TRMM PR & AMSR
Behrangi et al. (2012)
Berg et al.
(2010)

7. Comparison with GPCP and CMAP
TRMM
Mean precipitation rate (mm/day)
MCTA
GPCP
CMAP
CloudSat
Merged CloudSat-TRMM-AMSR estimate
AMSR-E
Difference (mm/d)
Relative diff
Behrangi et al. (J. of Climate, 2014)
Latitude

8. Missing precipitation over ocean (TRMM era)
Fraction
mm/day
Behrangi et al. (JGR; 2012)

9. -- TRMM Zonal distribution of precipitation ocean
(pre-GPM)
mm/day
/data/home/abehrang/A/codes/PAPER_GPM_CS/agu_ocean_plots.m
GPCP V2.2 showed ~28% more precipitation than AMSR-E
(~24 W/km2) !

10. CloudSat contribution
ocean
MCTA: Merged CloudSat TRMM(PR) Aqua (AMSR)
Behrangi et al.
(2014; J of climate)
mm/day
It was found that GPCP may underestimate global precipitation by about 5%
The missed 5% precip. matched the water budget closure study by Rodell et al. (2015) performed a year later.

11. CloudSat contribution (update)
ocean
MCTA-V2= MCTA + mixed phase precipitation
mm/day

12. CloudSat contribution (update)
ocean
GPCP V2.3
mm/day
/data/home/abehrang/A/codes/PAPER_GPM_CS/agu_ocean_plots.m

13. CloudSat contribution (update)
ocean
AMSR-2 GPM GPROF 2014 V4
mm/day
/data/home/abehrang/A/codes/PAPER_GPM_CS/agu_ocean_plots.m
MCTA V2 shows about 5.5% more precipitation than GPCP V2.3 over ocean
MCTA V2 shows ~23% more precipitation than new retrievals from AMSR2 (it was 33% AMSR-E).

14. GPM assessment
Gail Skofronick-Jackson et al. BAMS 2017

15. GPM assessment
Land
mm/day
AMSR-2 GPM
GPROF 2014 V4
GPCP V2.3
…, although we expect additional improvement by GPROF V5

16. Analysis based on environmental conditions

17. Missing precipitation over land (TRMM era)
Comparison of precipitation detectability as a function of cloud type
Fraction of precipitation detected by CloudSat
(TRMM era)
LAND
Behrangi et al. 2014
Over land : MW precipitation are robust for:
Convective system
Rainfall over non-frozen land

18. High latitude precipitation map NH
Given that the products have been in state of flux and in preparation to more effectively assess the impact of GPM we started to analyze various precipitrion pocuts and identify zones of disagreement.
Gradient of precipitation amount
Higher precip is seen in reanalysis and cloudsat comapred to GPCP
Reanalysis and cloudsat show similar patterns
CMAP significantly underestiamtes and has missing data
( )
Behrangi et al (JGR)

19. GRACE contribution ! T2m <-1 C Behrangi et al. (GRL, 2017)
T2m <-1 C
Behrangi et al. (GRL, 2017)
Behrangi et al. (JGR, 2016)

20. ? High latitude precipitation maps SH Behrangi et al. 2016 (JGR)
CMAP very low
GPCP around 60 degree shows little sensitivity to changes in longitude
CloudSat and reanalyses have similar patterns
NCEP tends to over estimate
Behrangi et al (JGR)

21. Summary of statistics for high latitude (poleward 50deg S)
Land+ Ocean Ocean Land (Antarctica)
Behrangi et al JGR

22. CMIP5 Model ranks based on CloudSat observation: Rank 1 is the best match.
SH ocean BIAS
SH ocean RMSE
SH land BIAS
SH land BIAS

23. Seasonal future changes in precipitation
The subset of CMIP5 models that best match current CloudSat estimate shows larger increase in future precipitation (i.e., the end of 21st century) than all-model ensemble average. This rate depends on region and season.
Also see Palerme et al for assessing CMIP5 models precip. based on CloudSat snowfall

24. Summary
CloudSat has brought great insights on precipitation occurrence, intensity, and phase, globally and especially in high latitudes.
Through Merging CloudSat, TRMM, and AMSR (MCTA) we found that about 5% of global precipitation is likely missed by GPCP => ~4 W/m2
In fact, this estimate was later confirmed by water closure study by Rodell et al. (2015)
MCTA suggests that GPCP is relatively robust in capturing the global numbers, but there are inconsistencies in zonal and regional distribution of precipitation amount (also confirmed by GRACE). Efforts are underway to improve GPCP in high latitudes.
CloudSat is now helping GPM for both assessment and retrieval purposes, especially in high latitudes. In fact, it is being used to guide GPROF.
A Subset of CMIP5 models that best match current CloudSat estimate shows larger increase in future precipitation (i.e., the end of 21st century) than all-model ensemble average. This rate depends on region and season.

25. Thanks!