Using CloudSat Cloud Retrievals to Differentiate Satellite-Derived Rainfall Productsover West Africa

ADRIAN M. TOMPKINS AND ADEYEMI A. ADEBIYI*

The Abdus Salam International Centre for Theoretical Physics, Trieste, Italy

(Manuscript received 19 March 2012, in final form 25 June 2012)

ABSTRACT

Daily precipitation retrievals from three algorithms [the Tropical Rainfall Measuring Mission 3B42 rain

product (TRMM-3B42), the Climate Prediction Center morphing technique (CMORPH), and the second

version (RFEv2) of the Famine Early Warning System (FEWS)] and CloudSat retrievals of cloud liquid

water, ice amount, and cloud fraction are used to document the cloud structures associated with rainfall

location and intensity in the West African monsoon. The different rainfall retrieval approaches lead to

contrasting cloud sensitivities between all three algorithmsmost apparent in the onset period of June and July.

During the monsoon preonset phase, CMORPH produces a precipitation peak at around 128N associated

with upper-level cirrus clouds, while FEWS and TRMM both produce rainfall maxima collocated with the

tropospheric–deep convective cloud structures at 48–68N. In July similar relative displacements of the rainfall

maxima are observed. Conditional sampling of several hundred convection systems proves that, while upper-

level cirrus is advected northward relative to themotion of the convective system cores, the reduced cover and

water content of lower-tropospheric clouds in the northern zone could be due to signal attenuation as the

systems there appear to be more intense, producing higher ice water contents. Thus, while CMORPH may

overestimate rainfall in the northern zone due to its reliance on cloud ice, TRMM and FEWS are likely

underestimating precipitation in this zone, potentially due to the use of infrared based products in TRMMand

FEWS when microwave is not available. Mapping the CloudSat retrievals as a function of rain rate confirms

the greater sensitivity of CMORPH to ice cloud and indicates that high-intensity rainfall events are associated

with systems that are deeper and of a greater spatial scale.

1. Introduction

Knowledge of precipitation is crucial for sectoral plan-

ning needs in Africa, but is hindered by the dearth of

global telecommunications system (GTS) station obser-

vations. This implies a reliance on satellite rainfall re-

trievals, of which the community is offered a wide choice

that incorporate many sources of conventional and re-

motely sensed data into diverse algorithms (Ali et al. 2005).

While the overall approaches of these retrieval algo-

rithms are well documented, their complex nature im-

plies that a grasp of their intricate detail is elusive to the

general user.Assessmentmostly relies on intercomparison

and evaluation with independent rain gauge measure-

ments where available. The three daily rainfall products

examined in this study [the Climate Prediction Center

(CPC) morphing technique (CMORPH), the second

version (RFEv2) of the Famine Early Warning System

(FEWS), and the Tropical Rainfall Measuring Mission

3B42 rain product (TRMM-3B42)] have, alongwith other

products, undergone such assessment in Adeyewa and

Nakamura (2003), Bowman (2005), Dinku et al. (2007),

Sapiano and Arkin (2009), Stisen and Sandholt (2010),

Jobard et al. (2011), Cohen et al. (2011), and Romilly and

Gebremichael (2011). Several studies highlight the

performance of TRMM, CMORPH, or FEWS prod-

ucts over Africa, but the conclusions depend on the

region examined. This is expected as retrieval algo-

rithms differ in their sensitivities to mesoscale systems

and their associated cloud structures, which change

with season and region. Examining how the rain rate of

each retrieval changeswith the associated cloud structure

would enhance our present understanding of the dif-

ferences between these algorithms.

* Current affiliation: Rosenstiel School of Marine and Atmo-

spheric Science, University of Miami, Miami, Florida.

Corresponding author address:AdrianM. Tompkins, TheAbdus

Salam International Centre for Theoretical Physics, Strada Costiera

11, 34014 Trieste, Italy.

E-mail: tompkins@ictp.it

1810 JOURNAL OF HYDROMETEOROLOGY VOLUME 13

DOI: 10.1175/JHM-D-12-039.1

� 2012 American Meteorological Society

Previously, Zuidema and Mapes (2008) compared

cloud structures with rain rates used ground-based cloud

observations, and Liu et al. (2008) collocated TRMM

precipitation with infrared and visible observations. The

CloudSat spaceborne cloud radar provides an opportu-

nity to extend these studies using cloud liquid water

and ice retrievals available at a high horizontal and

vertical resolution (Stephens et al. 2002). An initial view

of cloud–precipitation relationships was given inHaynes

and Stephens (2007), Kubar and Hartmann (2008) col-

located CloudSat information with Advanced Micro-

wave Scanning Radiometer (AMSR) precipitation in

the Pacific region, and Yuan et al. (2011) examined anvil

cloud structures divided into three rain-rate categories.

Here we focus on how the relationship between clouds

and precipitation changes according to the retrieval al-

gorithm employed in three commonly used precipitation

products. The study is conducted over the West Africa

region since the highly zonal nature of the monsoon

system may identify systematic differences between

rainfall algorithms potentially obscured in other tropical

zones when averaging over long periods. The climate of

the region is also well documented (Janicot et al. 2008).

2. Data

a. Rainfall data

This study uses three satellite rainfall products that

are available at relatively high horizontal resolutions in

near real time. These will be referred to as TRMM,

FEWS, and CMORPH. The intercomparison papers

cited in the introduction and Tian et al. (2009) discuss

some of the strengths and weaknesses of these products.

TheNationalOceanic andAtmosphericAdministration’s

(NOAA’s) Climate Prediction Center (CPC) morphing

technique (CMORPH) estimates are derived using

30-min slot rainfall retrievals from the passive microwave

(PMW) sensors of the TRMM Microwave Imager

(TMI), the Advanced Microwave Sounding Unit B

(AMSU-B), the Special Sensor Microwave Imager

(SSM/I), and, until the 4 October 2011 malfunction, the

AdvancedMicrowave ScanningRadiometer for theEarth

Observing System (AMSR-E) (Joyce et al. 2004)withTMI

retrievals taking precedence. Spatial propagation vectors

are derived from the IRdata tomorph rainfall features and

the 3-h 0.258 3 0.258 resolution product is used. The PMW

algorithms are not sensitive to liquid water and precipi-

tation over land where they rely on ice crystal scattering

extinction at the higher frequencies to derive surface rain

rates (McCollum and Ferraro 2003; Huffman et al. 2007).

The TRMM 3B42 rainfall also has a 3-hourly tempo-

ral and 0.258 3 0.258 spatial resolution and is derived

using a combination of the same microwave sensors

(TMI, AMSU-B, SSM/I, and AMSR-E) and the TRMM

2A12 precipitation radar (PR) (Iguchi et al. 2000) re-

trieval. However, in contrast to CMORPH, when

PMW/PR data are lacking, an IR-based retrieval is sub-

stituted. Finally, precipitation amounts are scaled to

match monthly satellite/rain gauge analyses of TRMM

3B43 (Huffman et al. 2007).

The second version (RFEv2) of the Famine Early

Warning System (FEWS) rainfall retrieval from NOAA

is derived combining AMSU- and SSM/I-based precipi-

tation estimates with an IR algorithm based on Meteosat

cold (235 K) cloud duration (CCD), which are then cali-

brated with (or, in a 15-km vicinity, replaced by) station

data available on the GTS system, resulting in a 0.18 30.18 daily precipitation estimate (Love 2002).

b. CloudSat

Launched in 2006, the 94-GHz Cloud Profiling Radar

(CPR) of CloudSat flies in the ‘‘A-Train’’ constellation

of satellites with the 0130/1330 (local time) equatorial

crossing times and a return period of 16 days (Stephens

et al. 2008). Level 2 retrieval products are used: the

Level 2B Cloud Geometries Profile (2B-GEOPROF)

(Mace 2007; Kahn et al. 2008) and Level 2B Cloud Water

Content Radar-Visible Optical Depth (2B-CWC-RVOD)

(Wood 2008) available at a 250-m vertical and 1.1-km

horizontal resolution. The 2B-GEOPROF product pro-

vides a 0–40 range ‘‘cloudmask’’ value and, following Stein

et al. (2011), cloud is assumed present for values exceeding

20. It is noted that precipitation-sized hydrometeors are

also detected and initial evaluations of CloudSat cloud

water and geometry retrievals (e.g., Heymsfield et al.

2008; Protat et al. 2009; Lee et al. 2010) show that in-

dividual ice water content (IWC) retrievals can diverge

from in situ measurements by up to 50% while, below

9 km in deep convective conditions, signal attenuation

can affect liquid and mixed phase cloud retrievals.

The low return rate and nonscanning swath of the

CloudSat instrument demands long averaging time/space

windows. Monthly data for 2006–10 are compared using

a region from 108W to 108E over West Africa. One

concern is the bidiurnal sampling frequency ofCloudSat

offering two cloud ‘‘snapshots’’ each day for comparison

to a daily total precipitation, potentially missing events

in regions with a pronounced diurnal cycle. Janowiak

et al. (2005) indicate that, apart from the coast, the peak

in convection over land inWest Africa generally occurs in

the evening. A zonal average of the CMORPH 3-hourly

rainfall diurnal cycle (similar to that of TRMM; Dai

et al. 2007) for the months from July to September for

the years from 2006 to 2010 (Fig. 1) confirms a rainfall

peak at 108N occurring in the early evening. Neverthe-

less, the 0000–0300 UTC window containing the CloudSat

DECEMBER 2012 TOMPK IN S AND ADEB IY I 1811

overpass still samples a relative maximum, while the

afternoon overpass misses the main rainfall peak but

captures the coastal precipitation. It would appear that

the CloudSat sampling captures many events in this

region but remains a caveat.

3. Results

The cloud structures during the months from June to

September (JAS) show the onset and main rainy season

(Fig. 2). In June (Figs. 2a,e) the CloudSat data show

a high incidence of cloud cover in two distinct peaks

centered at around 58 and 128N. The CloudSat cloud

cover shows that the peak at 58N is linked with a deep

cloud structure representative of the mean latitudinal

location of convection in the preonset phase (Sultan and

Janicot 2003). The zone at 128N is dominated by upper-

tropospheric cirrus cloud at 10–13-km height. The West

African monsoon onset often occurs in July (Marteau

et al. 2009); thus, convective cloud structures are in ev-

idence both in the zone of 48–68N and also farther north

between 108 and 128N (Figs. 2b,f). The 28–88N low-level

monsoon flow is marked by (predominately nocturnal;

Schrage et al. 2007; Knippertz et al. 2011) stratus cloud

that deepens as the flow progresses northward. By

August the monsoon is fully established with a clearer

distinction in cloud cover between the two latitudinal

zones with deep convection at 128N, while at 58N shal-

low monsoon stratus and cirrus outflow from the north

are present (Figs. 2c,g). The picture in September (Figs.

2d,h) is similar with the deep convection zone displaced

southward as monsoon cessation commences, while be-

tween 158 and 308N it shows enhanced cloud at the

mixing layer top at 6 km (Thorncroft et al. 2003; Stein

et al. 2011) when midtropospheric humidity is peaking

at these latitudes (Zhang et al. 2006).

Comparing the precipitation for June, FEWS and

TRMM rainfall amounts are similar in structure, with

a clear maximum in rainfall centered over the deep

FIG. 1. Diurnal variation of CMORPH JAS 2006–10 zonally averaged

(108W–108E) precipitation (mm day21) as a function of latitude.

FIG. 2. Zonal averages of cloud and precipitation averaged over 108W–108E, with each column giving a separate month during the

monsoon (June–September) averaged over the years 2006–10: (a)–(d)CloudSat-derived cloud fraction and (e)–(h) comparison of height-

integrated mean cloud fraction (%) from CloudSat [i.e., the vertical integral of (a)–(d)] to the zonal mean of the three precipitation

products of TRMM, CMORPH, and FEWS (all with units of mm day21).

1812 JOURNAL OF HYDROMETEOROLOGY VOLUME 13

convective zone at 58N where the relative difference is

approximately 25%, and there is only a hint of a sec-

ondary maximum at 108–128N. Outside the deep con-

vective zone these products agree well. In contrast, the

CMORPH rainfall algorithm appears to be influenced

by the high ice anvil cloud at 108–128N, producing a

strong secondary rainfall maximum. In July FEWS and

TRMM still bear a good resemblance, with a homo-

geneous distribution of rainfall matching CloudSat

column-integrated cloud cover and minor peaks at 58and 128N. The similarity of TRMM and FEWS is likely

due to their common use of gauge data for calibration

purposes, in contrast to the CMORPH approach. The

TRMM retrievals produce a maximum rainfall peak at

58N, despite the lack of high-level cloudiness, indicat-

ing warm rain generating processes. Instead, CMORPH

rainfall is dominated by a major peak at 128N, high-

lighting the cirrus cloud zone. Once the monsoon is es-

tablished in August, all three rainfall products agree

with the positioning of maximum rainfall, although the

CMORPH extends the maximum to 148N where cirrus

dominates and a similar structure is seen in September.

All three algorithms produce rainfall exceeding 15 mm

total for August and September north of 208N, apparently

exceeding estimates from surface stations (Tompkins and

Feudale 2010), although the latter are sparse in this zone

and miss isolated rainfall events.

The distinction between the rainfall products during

the monsoon onset (June/July) highlights the usefulness

of the West African monsoon as a convection ‘‘labora-

tory’’ due to its zonal nature. There are several possible

reasons for this distinction. In these months, while some

of the upper-tropospheric cloud is likely locally gener-

ated by deep convection, the low mean cloud cover

values in the lower troposphere and long sublimation

time scales of anvil ice could indicate a relative north-

ward propagation (i.e., relative to the system) of cirrus

from the deep convective zones farther to the south,

which would likely not be precipitating [referred to in

Tompkins et al. (2005)]. Although the microwave-based

algorithms attempt to discriminate between precipitating

and nonprecipitating ice clouds (Huffman et al. 2007),

it is possible that the sole reliance on microwave ice

retrievals in the CMORPH rainfall algorithm over land

could associate some of this upper-level cirrus cloud as

precipitating since limited information from the liquid

phase of the clouds is used. This would agree with station

observations such as Nicholson et al. (2003).

On other hand, a study of 1998 TRMM data by Mohr

and Thorncroft (2006) shows that systems north of 108Nin these onset months can be particularly intense, im-

plying that higher precipitation amounts could be asso-

ciatedwith a lowermonthlymean cloud cover.Moreover,

such intense systems would be associated with increased

attenuation, reducing lower-level mean cloud cover to

the north. This idea is corroborated by Fontaine et al.

(2008), who show that in 1998 monsoon onset dates di-

verged greatly between station data (very early onset)

and an IR-based definition (very late onset), indicating

that intense systems are producing high rainfall quanti-

ties. However, the study ofMohr and Thorncroft (2006)

is only for one year of systems; moreover, the study of

Nicholson et al. (2003) found excellent agreement

of the TRMM 3B42 with an unprecedented dataset of

nearly 1000 rain gauges in West Africa in the year 1998.

To investigate this further, deep convective events are

conditionally sampled during the onset months of June/

July (Fig. 3). A deep convective profile is defined as

having a minimum of 12-km vertical cloud extent, where

each cloudy point must have a cloud water content ex-

ceeding a threshold of 50 mg kg21. The center of the

convective system is located from the maximum cloud

(ice and liquid) water path for all profiles that meet this

criterion. The majority of CloudSat overpasses do not

detect any profiles that meets this criterion. In the cases

where more than one distinct convective system is

present in an overpass, the algorithm only samples the

most intense event, defined in terms of the maximum

cloud water path. For each detected system, the ice and

liquid water content are then averaged for 200 km in

each along-track direction from the center profile, and,

for each CloudSat data point, the nearest (colocated)

precipitation retrieval for the 3-h time slot in which the

CloudSat overpass occurs is averaged. FEWS data is

only available as a daily total and is excluded from this

analysis.

The analysis is divided between the zones 08–108N and

108–208N, and in order to increase the sample size the

zone is widened to 208W–208E inwhich a total of 426 such

events were captured. In both zones, the mean sampled

system highlights the asymmetry with a relative northerly

advection of the ice cirrus with respect to the system

centroid. The northern zone has greater ice amounts,

which reduces the liquid water content retrievals through

signal attenuation. This results in relatively low liquid wa-

ter amounts with respect to ice; indeed, reducing the cloud

water threshold in the search criterion below 50 mg kg21

in the analysis algorithm actually increases the mean liquid

water in the conditionally sampled system (not shown).

The collocated rainfall retrievals confirm the greater pre-

cipitation amounts retrieved from CMORPH but, in-

terestingly, the region of significant TRMM rainfall is not

confined to the systemcore. In fact, the relative distribution

of rainfall in CMORPH and TRMM is very similar, in-

dicating that their respective assessment of whether cold

cloud is precipitating does not differ greatly. The behavior

DECEMBER 2012 TOMPK IN S AND ADEB IY I 1813

of the precipitation products is strikingly different between

the two zones with the CMORPH producing far more

rainfall in the northern zone corresponding to the more

intense systems with greater ice amounts aloft. Instead,

TRMMactually produces less rainfall in these regions than

the zone to the south.

To document the contrasting microphysical sensi-

tivities, the cloud ice and liquid amounts are binned

according to the collocated TRMM/CMORPH 3-hourly

precipitation rates (Fig. 4). The expected relationship

is revealed, with cloud ice and liquid increasing with

increasing rain rates for both TRMM and CMORPH.

The greater sensitivity to ice cloud of the CMORPH re-

trieval algorithm relative to TRMM is confirmed. Asso-

ciated with this is a contrasting sensitivity to cloud-top

height, visible from the cloud fraction. Both rain rates

have a similar overall sensitivity to cloud liquid water,

which initially increases monotonically as a function of

rain rate. This may be due to direct influence of liquid

water on precipitation rates, or simply due to the corre-

lation between ice and liquid water path in these systems.

Moreover, it is recalled that, under 9 km in intense sys-

tems, the CloudSat retrievals suffer from attenuation and

that the division between ice and liquid is diagnostic; thus,

some of the liquid water could likely be misidentified ice.

For the intense rainfall events cloud cover is higher in

CMORPH, indicating that the system size is on the whole

larger, although it should be highlighted that these ex-

treme events are relatively rare and sampling is an issue.

Nonraining or very light drizzle conditions are associated

with localized maximum in liquid water content at about

2 km. CMORPH detects fewer instances of zero rainfall

(Fig. 4c) and the mean IWC for this rainfall category is

lower relative to TRMM. This confirms that CMORPH

diagnoses precipitation associated to ice clouds that the

TRMM algorithm classifies as non-precipitating.

4. Conclusions

To give another view of the differences between re-

motely sensed rainfall retrievals we classify them accord-

ing to their association with CloudSat-derived cloud

structures. The analysis focuses on three commonly

used, daily precipitation products over West Africa—

CMORPH, TRMM-3B42, and FEWS-RFEv2—but could

be extended to other products and regions. The study

performs regional averages of the datasets, in addition

to conditional sampling of deep convective systems with

FIG. 3. (a)Mean ice/liquid water content (mg m23) of 274 conditionally sampled deep convective events during the

period of June and July with core occurring between 08 and 108N with contours shaded for values exceeding

100 mg m23. In the lower panel, collocated rainfall is given. The positive direction is always to the north. (b)Mean of

152 events in zone 108–208N.

1814 JOURNAL OF HYDROMETEOROLOGY VOLUME 13

the collocated rainfall and cloudmicrophysical properties

as a function of rain rate.

Over land, microwave retrievals used in all three

precipitation products rely on relating precipitation to

detecting cloud ice. CMORPH rainfall relies solely on

this source of information without use of infrared re-

trievals or rain gauge calibration and is shown to have

a much higher sensitivity to cloud ice relative to TRMM

and FEWS. The distinction is particularly acute during

the onset months of June and July where the zonal na-

ture of the monsoon and the abrupt rainfall shift high-

lights the contrast between the algorithms. The study

highlights the advection of cirrus to the northward side

of systems and reveals a significant contrast between

convective systems in the zone 108–208N and farther

south, with the former producing more ice cloud, while

lower-tropospheric cloud is reduced by signal attenua-

tion. The CMORPH algorithm produces increased rain-

fall in these zones related to the increased ice cloud,

thus also increasing cloud-top height, while FEWS and

TRMM tend to enhance rainfall in the zone associated

with tropospheric deep structures to the south. In

conclusion, while CMORPH may be overestimating

the precipitation in the northern zone, it also seem

likely that FEWS and TRMM are underestimating

precipitation there from these intense convection sys-

tems. This work highlights the use of CloudSat data in

discerning potential reasons for rainfall retrievals dif-

ferences from one region to another, and future work

will extend this technique to other tropical regions

and additional datasets.

Acknowledgments.The authors thankPaquitaZuidema

and Michela Biasutti for comments on the manuscript

and Robert Joyce for answering many questions concern-

ing details of the CMORPH algorithm. Peter Knippertz

and an anonymous reviewer provided detailed com-

ments and suggestions that greatly improved the orig-

inal submission.

REFERENCES

Adeyewa, Z. D., and K. Nakamura, 2003: Validation of TRMM

radar rainfall data over major climatic regions in Africa.

J. Appl. Meteor., 42, 331–347.

Ali, A., A. Amani, A. Diedhiou, and T. Lebel, 2005: Rainfall es-

timation in the Sahel. Part II: Evaluation of rain gauge net-

works in the CILSS countries and objective intercomparison

of rainfall products. J. Appl. Meteor., 44, 1707–1722.

Bowman, K. P., 2005: Comparison of TRMM precipitation re-

trievals with rain gauge data from ocean buoys. J. Climate, 18,

178–190.

Cohen, L. T., J. P. Matos, J. L. Boillat, and A. J. Schleiss, 2011:

Comparison and evaluation of satellite derived precipitation

products for hydrological modeling of the Zambezi River

Basin. Hydrol. Earth Syst. Sci. Discuss., 8, 8173–8201.Dai, A., X. Lin, and K. Hsu, 2007: The frequency, intensity, and

diurnal cycle of precipitation in surface and satellite obser-

vations over low- and mid-latitudes. Climate Dyn., 29 (7–8),

727–744.

Dinku, T., P. Ceccato, E. Grover-Kopec, M. Lemma, S. J. Connor,

and C. F. Ropelewski, 2007: Validation of satellite rainfall

products over East Africa’s complex topography. Int. J. Re-

mote Sens., 28, 1503–1526.

Fontaine, B., S. Louvet, and P. Roucou, 2008: Definition and pre-

dictability of anOLR-basedWest Africanmonsoon onset. Int.

J. Climatol., 28, 1787–1798, doi:10.1002/joc.1674.Haynes, J. M., and G. L. Stephens, 2007: Tropical oceanic cloudiness

and the incidence of precipitation: Early results from CloudSat.

Geophys. Res. Lett., 34, L09811, doi:10.1029/2007GL029335.

Heymsfield, A. J., and Coauthors, 2008: Testing IWC retrieval

methods using radar and ancillary measurements with in situ

data. J. Appl. Meteor. Climatol., 47, 135–163.

FIG. 4. Mean CloudSat cloud properties binned as a function of

both height and collocated, cotemporal precipitation amount re-

trieved from (a) CMORPH and (b) TRMM. The shaded contours

show cloud fraction (color scale in legend on right), while the white

solid and black dashed lines show ice and liquid water content,

respectively (mg m23). (c) Number of observations used.

DECEMBER 2012 TOMPK IN S AND ADEB IY I 1815

Huffman, G. J., and Coauthors, 2007: The TRMMMultisatellite

Precipitation Analysis (TMPA): Quasi-global, multiyear,

combined-sensor precipitation estimates at fine scales. J. Hy-

drometeor., 8, 38–55.Iguchi, T., T. Kozu, R.Meneghini, J. Awaka, and K. Okamoto, 2000:

Rain-profiling algorithm for the TRMM Precipitation Radar.

J. Appl. Meteor., 39, 2038–2052.

Janicot, S., and Coauthors, 2008: Large-scale overview of the

summer monsoon over West Africa during the AMMA field

experiment in 2006. Ann. Geophys., 26, 2569–2595.

Janowiak, J. E., V. E. Kousky, and R. J. Joyce, 2005: Diurnal cycle

of precipitation determined from the CMORPH high spatial

and temporal resolution global precipitation analyses. J. Geo-

phys. Res., 110, D23105, doi:10.1029/2005JD006.

Jobard, I., F. Chopin, J. C. Berges, and R. Roca, 2011: An inter-

comparison of 10-day satellite precipitation products during

West African monsoon. Int. J. Remote Sens., 32, 2353–2376.

Joyce, R. J., J. E. Janowiak, P. A. Arkin, and P. Xie, 2004:

CMORPH: A method that produces global precipitation es-

timates from passive microwave and infrared data at high

spatial and temporal resolution. J. Hydrometeor., 5, 487–503.

Kahn, B. H., and Coauthors, 2008: Cloud type comparisons of

AIRS, CloudSat, and CALIPSO cloud height and amount.

Atmos. Chem. Phys., 8, 1231–1248.

Knippertz, P., A.H. Fink, R. Schuster, J. Trentmann, J.M. Schrage,

and C. Yorke, 2011: Ultra-low clouds over the southern West

African monsoon region. Geophys. Res. Lett., 38, L21808,

doi:10.1029/2011GL049.

Kubar, T. L., and D. L. Hartmann, 2008: Vertical structure of

tropical oceanic convective clouds and its relation to precipi-

tation.Geophys. Res. Lett., 35,L03804, doi:10.1029/2007GL032.

Lee, S., B. H. Kahn, and J. Teixeira, 2010: Characterization

of cloud liquid water content distributions from CloudSat.

J. Geophys. Res., 115, D20203, doi:10.1029/2009JD013.

Liu, C., E. J. Zipser, D. J. Cecil, S. W. Nesbitt, and S. Sherwood, 2008:

A cloud and precipitation feature database from nine years of

TRMM observations. J. Appl. Meteor. Climatol., 47, 2712–2728.Love, T., 2002: The Climate Prediction Center rainfall algorithm

version 2. NOAA Climate Prediction Center Tech. Rep.

[Available online at http://www.cpc.ncep.noaa.gov/products/

fews/RFE2.0_tech.pdf.]

Mace, G.G., 2007: Level 2GEOPROFproduct process description

interface control document (V5.3). Colorado StateUniversity,

CIRA Tech. Rep. [Available online at http://www.cloudsat.

cira.colostate.edu//ICD/2B-GEOPROF/2B-GEOPROF_

PDICD_5.3.doc.]

Marteau, R., V. Moron, and N. Philippon, 2009: Spatial coherence

of monsoon onset over western and central Sahel (1950–2000).

J. Climate, 22, 1313–1324.McCollum, J. R., and R. R. Ferraro, 2003: Next generation of

NOAA/NESDIS TMI, SSM/I, and AMSR-E microwave land

rainfall algorithms. J. Geophys. Res., 108, 8382, doi:10.1029/2001JD001512.

Mohr, K. I., and C. D. Thorncroft, 2006: Intense convective systems

in West Africa and their relationship to the African easterly

jet. Quart. J. Roy. Meteor. Soc., 132, 163–176.Nicholson, S. E., and Coauthors, 2003: Validation of TRMM and

other rainfall estimates with a high-density gauge dataset for

West Africa. Part II: Validation of TRMM rainfall products.

J. Appl. Meteor., 42, 1355–1368.

Protat, A., and Coauthors, 2009: Assessment of CloudSat reflec-

tivity measurements and ice cloud properties using ground-

based and airborne cloud radar observations. J. Atmos. Oceanic

Technol., 26, 1717–1741.

Romilly, T. G., and M. Gebremichael, 2011: Evaluation of satellite

rainfall estimates over Ethiopian river basins. Hydrol. Earth

Syst. Sci., 15, 1505–1514.

Sapiano, M. R. P., and P. A. Arkin, 2009: An intercomparison and

validation of high-resolution satellite precipitation estimates

with 3-hourly gauge data. J. Hydrometeor., 10, 149–166.Schrage, J. M., S. Augustyn, and A. H. Fink, 2007: Nocturnal

stratiform cloudiness during the West African monsoon. Me-

teor. Atmos. Phys., 95, 73–86.Stein, T. H. M., D. J. Parker, J. Delanoe, N. S. Dixon, R. J. Hogan,

P. Knippertz, R. I. Maidment, and J. H. Marsham, 2011:

The vertical cloud structure of the West African monsoon:

A 4 year climatology using CloudSat and CALIPSO. J. Geo-

phys. Res., 116, D22205, doi:10.1029/2011JD016029.

Stephens, G. L., and Coauthors, 2002: The CloudSat mission and

the A-Train. Bull. Amer. Meteor. Soc., 83, 1771–1790.

——, and Coauthors, 2008: CloudSat mission: Performance and

early science after the first year of operation. J. Geophys. Res.,

113, D00A18, doi:10.1029/2008JD009982.

Stisen, S., and I. Sandholt, 2010: Evaluation of remote-sensing-

based rainfall products through predictive capability in hy-

drological runoff modelling. Hydrol. Processes, 24, 879–891.

Sultan, B., and S. Janicot, 2003: The West African monsoon dy-

namics. Part II: The ‘‘preonset’’ and ‘‘onset’’ of the summer

monsoon. J. Climate, 16, 3407–3427.

Thorncroft, C. D., and Coauthors, 2003: The JET2000 project:

Aircraft observations of the African easterly jet and African

easterly waves. Bull. Amer. Meteor. Soc., 84, 337–351.Tian, Y., and Coauthors, 2009: Component analysis of errors in

satellite-based precipitation estimates. J. Geophys. Res., 114,

D24101, doi:10.1029/2009JD011949.

Tompkins, A. M., and L. Feudale, 2010: Seasonal ensemble pre-

dictions of West African monsoon precipitation in the ECMWF

System 3 with a focus on the AMMA special observing period in

2006. Wea. Forecasting, 25, 768–788.——,A.Diongue,D. J. Parker, andC.D. Thorncroft, 2005:African

easterly jet in theECMWF integrated forecast system: 4D-Var

analysis. Quart. J. Roy. Meteor. Soc., 131, 2861–2886.

Wood, N., 2008: Level 2B radar-visible optical depth cloud water

content (2B-CWC-RVOD) process description document

(v5.1). Colorado State University, CIRA Tech. Rep., 23 pp.

[Available online at http://www.cloudsat.cira.colostate.edu/.]

Yuan, J., R. A. Houze Jr., and A. J. Heymsfield, 2011: Vertical

structures of anvil clouds of tropical mesoscale convective

systems observed by CloudSat. J. Atmos. Sci., 68, 1653–1674.

Zhang, C., P. Woodworth, and G. Gu, 2006: The seasonal cycle in

the lower troposphere over West Africa from sounding ob-

servations. Quart. J. Roy. Meteor. Soc., 132, 2559–2582.

Zuidema, P., and B. E. Mapes, 2008: Cloud vertical structure

observed from space and ship over the Bay of Bengal

and the eastern tropical Pacific. J. Meteor. Soc. Japan, 86,

205–218.

1816 JOURNAL OF HYDROMETEOROLOGY VOLUME 13