GLOBAL PRECIPITATION AT ONE-DEGREE DAILY RESOLUTION
FROM MULTI-SATELLITE OBSERVATIONS
George J. Huffman 1,2, Robert F. Adler 1, Mark M. Morrissey 3, Scott Curtisl, 4, Robert Joyce 5,
Brad McGavock 3. Joel Susskind 1
1:. NASA/GSFC Laboratory for Atmospheres, Greenbelt, MD
2: Science Systems and Applications, Inc., Lanham, MD
3: Environmental Verification and Analysis Center, Norman, OK
4: Joint Center for Earth-System Technology, Baltimore, MD
5: Climate Prediction Center, NCEP/NWS/NOAA, Washington, DC;
Research and Data Corp., Greenbelt, MD
Submitted to Journal of Hydrometeorology
1 April 2000
.Corresponding attthor Address: Dr. George J. Huffman/SSAI, NASA/GSFC Code 912,
Greenbelt, MD 20771.
Global Precipitation at One-Degree Daily Resolution From Multi-Satellite Observations
Abstract
The One-Degree Daily (1DD) technique is described for producing globally complete daily
estimates of precipitation on a l°xl ° lat/long grid from currently available observational data.
Where possible (40°N-40°S), the Threshold-Matched Precipitation Index (TMPI) provides
precipitation estimates in which the 3-hourly infrared brightness temperatures (IR Tb) are
thresholded and all "cold" pixels are given a single precipitation rate. This approach is an
adaptation of the Geostationary Operational Environmental Satellite (GOES) Precipitation Index
(GPI), but for the TMPI the IR Tb threshold and conditional rain rate are set locally by month
from Special Sensor Microwave/Imager (SSM/I)-based precipitation frequency and the Global
Precipitation Climatology' Project (GPCP) satellite-gauge (SG) combined monthly precipitation
estimate, respectively. At higher latitudes the 1DD features a rescaled daily Television Infrared
Observation Satellite (TIROS) Operational Vertical Sounder (TOVS) precipitation. The
frequency of rain days in the TOVS is scaled down to match that in the TMPI at the data
boundaries, and the resulting non-zero TOVS values are scaled locally to sum to the SG (which
is a globally complete monthly product). The time series of the daily 1DD global images shows
good continuity in time and across the data boundaries. Various examples are shown to illustrate
xz'_;d_tie" for ;nd ;'';a'''_l grid.box values shows a vep/high .oot-m.,n-oqt-.,re e=er, but it
improves quickly when users perform time/space averaging according to their own requirements.
1. Introduction
A long time series of fine-scale observation-based global precipitation is needed to support a
variety of studies, including global change, surface hydrology, and numerical weather and
climate model initialization and validation. However, data record, sampling, and algorithmic
considerations limit the range of scales that could be reported with reasonable accuracy. For
example, the World Climate Research Program (WCRP) established the Global Precipitation
Climatology Project (GPCP) with the initial goal of producing precipitation estimates on a
monthly 2.5°x2.5 ° lat/long grid for a number of years (WCRP, 1986). The GPCP is succeeding
in this goal, with over a decade of data available (1986-1999 at present), another seven years in
preparation (1979-1985), and routine production continuing a few months after real time.
Due to the lack of finer scale precipitation data numerous applications remain stymied.
Researchers wish to validate hydrologic stream flow models by forcing them with observed data
that resolve individual storms and catch basins. Even monthly-scale events are hard to study
with calendar-month averages. Extratropical blocking events are a typical example. They
initiate and decay within the span of a few days, but can persist for weeks (Blackmon et al.
1984). Calendar-month averages typically mix blocking and non-blocking periods, whereas
daily data allow the researcher to composite the data much more cleanly.
The fundamental barrier to finer-scale estimates is the lack of accurate, dense global data,
either from in-situ or remote sensors. Some regions do have the possibility of detailed
precipitation estimates thanks to local networks of sensors, such as the WSR-88D radar system
in the United States (Crum et al. t998). However, at most locations around the world it is
necessary to depend on satellite-based passive sensors. Microwave radiometers on polar-orbit
satellites produce fairly accurate instantaneous estimates, but their sparse temporal sampling
constrainsthetime/spacegriddingneededtoachievereasonablerandomerrorsin the GPCPdata
set.
Infrared(IR) radiometersongeosynchronoussatellites(geo-IR)provideexcellenttime and
spacesampling,but thequantitybeingsensed(mostlycloud-toptemperature)is indirectly
connectedto precipitation,particularlyontheshortesttime andspacescales,andin extratropical
latitudes.A problemthatgeo-IRdatasharewith mostsatellitedatasetsis thattheviewing
geometrybecomesunfavorablenearthelimb of eachsatellite'sview. This problemcanbe
solvedin partat tropicalandsubtropicallatitudesby mergingall availablegeo-IRdata,but the
issueremains,mostacutelyathigherlatitudesandin caseswheretheclosestgeosynchronous
satelliteis not reportingdata. An additionalbarrieris theneedto work with severalinternational
partnerstoobtainadministrativepermissions,maintainroutinedatadeliveries,anddedicate
sufficientcomputingresourcesbeforetheuseof global,full-resolutiongeo-IRdatacanbea
reality.
Startingin October1996theGPCPsetthestagefor higherresolutionestimatesby working
with thegeosynchronous-satelliteoperatorsaroundtheworld to collecthistogramsof geo-IR
brightnesstemperature(Tb) on a 1°xl ° grid covering 40°N-40°S at 3-hourly intervals. The
availability of this data set prompted the authors to develop the Threshold-Matched Precipitation
"_ ] .... 1..
Index (TMPl) to cstimate precipitation from zl_c o-,:u,_,,: gco-IR histograms; described in sc,_tion
2. To complete the global coverage, a technique for estimating precipitation outside of the geo-
IR coverage using sounding data from polar-orbit satellites was developed, as described in
section 3. Together, these form the One-Degree Daily (1DD) dataset, which is a first approach to
estimating global daily precipitation at the l°xl ° scale strictly from observational data. Figure 1
illustrates how the 1DD is computed. The philosophy is to use statistical parameters from trusted
estimates to constrain the overall behavior of the 1DD estimates, and then use the geo-IR and
sounder data to determine the day-to-day behavior. In the same vein, the adjustments are
computed and applied a month at a time to ensure simplicity and stability. All of the
computations in this paper are carried out separately for each grid box (perhaps with some
smoothing) unless otherwise stated. Section 4 provides some examples of 1DD-based analysis,
while section 5 summarizes validation results. Throughout the paper the reader should keep in
mind that the geo-IR and sounder data are responding to clouds, rather than hydrometeors, so
there is a substantial algorithmic uncertainty at the finest scales.
2. TMPI
The GOES Precipitation Index (GPI; Arkin and Meisner 1987) is one popular IR technique
that simply labels all pixels with IR Tb below a threshold as "rain" and assigns a single rainrate to
all such pixels. The GPI threshold Tb is Tb(rain)=235K and the conditional rain rate is R_=3 mm
h-1. These constants were chosen to maximize correlations with half-monthly rainfall on a
2.5°x2.5 ° lat/long grid over the Global Atmosphere Research Program (GARP) Atlantic Tropical
Experiment (GATE) ship array in summer 1974. Adler et al. (1993) held Tb(rain) constant and
allowed I_ to vary according to calibration between (approximately) time/space-matched GPI
and SSM/I-based rain estimates accumulated for a month, creating the Adjusted GPI (AGPI). In
this paper we introduce the TMPI, which allows both Tb(rain) and Rc to vary. For simplicity and
stability, the (spatially varying) Tb(rain) and t_ are computed on a monthly basis in the TMPI.
The (observational) datasets used for the TMPI (Fig. 1) at present are as follows:
1. GPCPhistogramsofgeo-IR Tb are compiled on a I°X1 ° lat/long grid over 40°N-40°S at 3-hr
intervals. The histograms have 24 classes covering 190-270K.
GPCP Advanced Very High Resolution Radiometer (AVHRR) low-earth-orbit IR (leo-IR)
GPI estimates are compiled on a l°xl ° lat/long grid over 40°N-40°S in 3-hr averages•
The Special Sensor Microwave/Imager (SSM/I) occurrence of precipitation according to the
Version 4.0 Goddard Profiling algorithm (GPROF 4.0; Kummerow et al. 1996) is computed
by the Goddard Space Flight Center Laboratory for Atmospheres on a 0.5°x0.5 ° lat/long
global grid for each orbit. GPROF is a physically based retrieval that matches observed
radiances from all 7 SSM/I channels to cloud-model-based radiances, producing pixel-by-
pixel precipitation estimates.
The GPCP Version 2 Satellite-Gauge (SG; Huffman et al. 1997) combination precipitation
estimates (rscj) are produced on a 2.5°x2.5 ° lat/long global grid by month, and box-
interpolated to l°xl ° for this study. The SG applies a sequential combination technique to
SSM/I, geo-IR, TOVS (introduced below), and rain gauge analyses
After the geo-IR histograms are corrected for zenith angle effects (Joyce et al. 2000) and
interpolated to a 1K interval (Joyce and Arkin 1997), they are matched within +1.5 hr to the
microwave-based frequency of precipitation and each is accumulated for the month. To further
• "7 -'=t D "" - "
cnsurc s',abilit;,', the monthly matched accumulations are smoothed with a, x,-_11dbox boxcar.
In each gridbox, the geo-IR histogram is summed starting with the coldest bin until the
cumulative fraction of total pixels matches the microwave-based fractional coverage, and
Tb(rain) is set to the corresponding Tb. This Tb(rain) is applied to the 7x7-box-smoothed sum of
all Tb histograms in the box for the month, yielding the fractional occurrence of rain in the full
geo-IR dataset for the month, fla. The single local conditional rainrate for all raining pixels is
computed as
•
•
.
Rc = rs___G_O (1)
fIR
because we require the TMPI to sum to the (monthly) SG product over the month to maintain
consistency between daily and monthly products.
Preliminary work showed that the Tb(rain) and 1% computed in this first round of estimation
contain some unrealistic values. Various combinations of parameters were tested for diagnosing
this problem, of which the plot of 1% as a function of Tb(rain) seemed the most useful. Not that
most points are tightly clustered (Fig. 2), but there is a scatter of high-1% outliers. Outliers
usually occur in coherent patches (i.e., inside the red lines on Fig. 3 for the sample month
January 1998) that tend to be associated with light rain or strong gradients in rain amount, likely
due to sampling problems in the SSM/I data. Such values would tend to cause estimation of a
few unrealistically heavy precipitation events during the month. Accordingly, an audit procedure
was developed that determines new Tb(rain) that are consistent with the SG values, when
required. We lack a definitive theoretical basis for identifying outliers, so we have subjectively
chosen a linear cutoff that parallels the main cluster and excludes the highest 10% of boxes for
the example month (Fig. 2). At this level of approximation we have neglected the hints at non-
linear behavior for low Tb(rain). The outlier 1% values are deleted and the pockets of missing are
smoothly filled from the surrounding values (Fig. 3). The "audited" values are used to generate
auditedfand then Tb(rain) estimates. As shown in Fig. 3, the resulting audited fields are
physically reasonable. Small errors in setting these parameters are likely not crucial, since the
smoothingdueto 1°xl ° gridding anddaily summationwill maskthedetailedpixel-level
occurrenceof rain. TheauditingprocedurealwaysreducesR_andincreasesTb(rain)andf
(exceptfor Tb(rain)--269K).
Eventhis auditingcannotsolveall problemsat Tb(rain)=269KbecausethewarmestTb bin in
the merged geo-IR dataset contains all values >270K. A region with low cloud-tops, such as a
subtropical high, likely should have Tb(rain)>270K, but Tb(rain) must limited to the last
distinguishable temperature bin of 269K (the uniform orange areas in Fig. 3), and will be given
an artificially high R¢ in the first round of estimation. Presently the auditing procedure replaces
R_ outliers by (lower) smooth-filled values, even though Tb(rain) is constrained to 269K. This
prevents unreasonable instantaneous values at the cost of failing to sum to the monthly SG value.
The penalty for this choice is small because nearly all cases are in low-precipitation regions.
Note that the corresponding lower limit of the histogram bins (190K) is not a problem in
practice.
The final audited R_ and Tb(rain) fields (Fig. 2) provide important insights into the systematic
regional variations in geo-IR data. R_ and Tb(rain) are near the GPI values in tropical oceanic
zones with "heavy" precipitation, but the corresponding land areas show much colder Tb(rain)
with higher I_. This implies that heavy convection in tropical land areas is deeper and more
concentrateci than over tropical oceans. In the relatively dry subtropical highs Tb(rain) is warm
because there is little or no penetrative convection. The relatively low Tb(rain) and Rc over the
Sahara Desert help screen out non-precipitating cirrus that passes over this region during the
boreal winter. The gradients along the northern coast of Africa reflect the transition from desert
to the pattern of boreal wintertime storms that typifies the Mediterranean.
Holes occur in individual geo-IR images, most routinely in the Indian Ocean sector, where no
geo-IR data were available until June 1998. To compensate, leo-IR estimates are processed to
fill in holes as necessary. The leo-IR data are only available as GPI estimates in the GPCP
merged IR dataset. Therefore. the instantaneous leo-IR estimate for the TMPI is simply the
instantaneous leo-IR GPI value scaled by the ratio of the SG for the month to the sum of all leo-
IR GPI values in the grid box for the month. The adjustment ratios are limited to the range
[0,2.5] to ensure reasonable behavior, but this limitation only becomes important in areas of light
rain.
The basic output of the TMPI is 3-hrly instantaneous estimates, mostly resulting from geo-
IR. In the current release, the 3-hrly images in each UTC day (00Z ..... 21Z) are summed to
produce the daily value. The daily product is considered more reliable than individual 3-hrly
images for two reasons. First, GPI-type IR estimates show better correlation with precipitation
as the averaging period increases (Arkin and Meisner 1987). Second, the current procedure does
not take into account the time of day (i.e., diurnal cycle biases). As a result, the individual 3-hrly
estimates are not part of the current release.
One interesting result of the TMPI procedure is insight into the frequency of occurrence of
precipitation. If we define " fractional coverage " as the fraction of all satellite pixels
contributing to the gridbox that have non-zero rain, and "rain days" as the fraction of days on
which a gridbox has non-zero rain (i.e., at least one pixel with non-zero rain sometime during the
day), then we expect the fractional coverage to be less than rain days. In the case of the TMPI,
the ratio of fractional coverage to rain days is less than 0.4 almost everywhere (Fig. 4).
Furthermore, TMPI and GPI rain day maps are closer to each other than to the rain days
4