550
A.D. Del Genio
been estimated indirectly from satellite using simple infrared or microwave indices, but often
based on detailed observations for only a single region. Over land, precipitation is more well
defined climatologically, but errors are nonetheless significant - precipitation averages over large
areas are notoriously difficult to measure with a network of local rain gages because much of it
occurs in high rain rate events of small spatial scale and short time duration (cf. Dai, 1995).
In a companion paper (Del Genio et aI., 1996) I have suggested that a useful accuracy standard
for data sets used to validate climate models is that the uncertainty in observed monthly mean
values of a climate parameter over spatial scales resolved by the model (hundreds of kilometers)
be significantly less than the differences among various models. For many climate parameters,
this criterion is met by existing data sets, but not so for precipitation. For example, the global
mean boreal winter precipitation simulated by 8 coupled atmosphere-ocean general circulation
models (GCMs) for the most recent IPCC Scientific Assessment (IPCC, 1995) varies from 2.4
to 3.8 mm/d. (Since 1 mm/d::::; 30 W/m 2 if converted to equivalent energy flux units, this
difference is comparable to model errors in simulation of the planetary radiation budget.) But
existing climatologies give estimates as low as 2.7 mm/d (Jaeger, 1976) and as high as 3.5 mm/d
(Legates and Willmott, 1990), almost as large as the range of model disagreements and much
larger than equivalent observational errors in radiation fluxes. Figure 20.1 shows differences in
January rainfall between two data sets, and between one of them and the most recent version of
the GISS GCM (cf. Del Genio et aI., 1996). Although model-data discrepancies are somewhat
larger than differences between data sets, in many places the differences in both cases are 1-2
mm/ d and> 4 mm/ d in certain parts of the tropics. Model disagreements are also particularly
large in the tropics: in an intercomparison of 19 GCMs, Randall et aI. (1992) reported zonal
and monthly mean rainfall differences of as much as 6-7 mm/d between individual GCMs in the
Intertropical Convergence Zone (ITCZ). It is also worth noting that existing precipitation data
exist only in climatological form: Annual cycle information is available, but neither interannual
variations nor diurnal cycle information exists.
To address the absence of suitable precipitation data sets for climate and tropical dynamics
applications, the Tropical Rainfall Measuring Mission (TRMM) satellite will be launched in
August 1997. This paper outlines the TRMM philosophy and approach to obtaining meaningful rainfall estimates, the challenges that must be overcome to achieve the desired accuracy
standard, the instruments and retrieval techniques to be used, and supporting observations
and modeling that will enhance the impact of the data set. Section 20.2 describes the TRMM
mission and instrumentation and Section 20.3 the challenges of measuring rainfall that produced the final mission configuration. Section 4 discusses rainfall retrieval approaches with the
primary TRMM instruments, while Section 20.5 describes some of the complementary science
that can be done with other TRMM instruments and models. Section 20.6 outlines the validation approach to be used by TRMM. We conclude in Section 20.7 with some thoughts about
extending the TRMM results to other data sets to create a longer-term record of precipitation
variations.
20.2 TRMM Mission and Instrumentation
TRMM is a joint mission of the United States and Japan. The Science Team is led by Project
Scientist Dr. Joanne Simpson of the NASA Goddard Space Flight Center. The TRMM spacecraft will be launched in August 1997 on a Japanese H-II rocket into a rapidly precessing 35°
inclination, 350 km altitude circular Earth orbit. The spacecraft will complete 16 orbits per
day, resulting in the ground track pattern indicated in Figure 20.2. TRMM is part of the Earth
Probes series of NASA launches, designed as relatively low-cost missions with targeted science
A.D. Del Genio
been estimated indirectly from satellite using simple infrared or microwave indices, but often
based on detailed observations for only a single region. Over land, precipitation is more well
defined climatologically, but errors are nonetheless significant - precipitation averages over large
areas are notoriously difficult to measure with a network of local rain gages because much of it
occurs in high rain rate events of small spatial scale and short time duration (cf. Dai, 1995).
In a companion paper (Del Genio et aI., 1996) I have suggested that a useful accuracy standard
for data sets used to validate climate models is that the uncertainty in observed monthly mean
values of a climate parameter over spatial scales resolved by the model (hundreds of kilometers)
be significantly less than the differences among various models. For many climate parameters,
this criterion is met by existing data sets, but not so for precipitation. For example, the global
mean boreal winter precipitation simulated by 8 coupled atmosphere-ocean general circulation
models (GCMs) for the most recent IPCC Scientific Assessment (IPCC, 1995) varies from 2.4
to 3.8 mm/d. (Since 1 mm/d::::; 30 W/m 2 if converted to equivalent energy flux units, this
difference is comparable to model errors in simulation of the planetary radiation budget.) But
existing climatologies give estimates as low as 2.7 mm/d (Jaeger, 1976) and as high as 3.5 mm/d
(Legates and Willmott, 1990), almost as large as the range of model disagreements and much
larger than equivalent observational errors in radiation fluxes. Figure 20.1 shows differences in
January rainfall between two data sets, and between one of them and the most recent version of
the GISS GCM (cf. Del Genio et aI., 1996). Although model-data discrepancies are somewhat
larger than differences between data sets, in many places the differences in both cases are 1-2
mm/ d and> 4 mm/ d in certain parts of the tropics. Model disagreements are also particularly
large in the tropics: in an intercomparison of 19 GCMs, Randall et aI. (1992) reported zonal
and monthly mean rainfall differences of as much as 6-7 mm/d between individual GCMs in the
Intertropical Convergence Zone (ITCZ). It is also worth noting that existing precipitation data
exist only in climatological form: Annual cycle information is available, but neither interannual
variations nor diurnal cycle information exists.
To address the absence of suitable precipitation data sets for climate and tropical dynamics
applications, the Tropical Rainfall Measuring Mission (TRMM) satellite will be launched in
August 1997. This paper outlines the TRMM philosophy and approach to obtaining meaningful rainfall estimates, the challenges that must be overcome to achieve the desired accuracy
standard, the instruments and retrieval techniques to be used, and supporting observations
and modeling that will enhance the impact of the data set. Section 20.2 describes the TRMM
mission and instrumentation and Section 20.3 the challenges of measuring rainfall that produced the final mission configuration. Section 4 discusses rainfall retrieval approaches with the
primary TRMM instruments, while Section 20.5 describes some of the complementary science
that can be done with other TRMM instruments and models. Section 20.6 outlines the validation approach to be used by TRMM. We conclude in Section 20.7 with some thoughts about
extending the TRMM results to other data sets to create a longer-term record of precipitation
variations.
20.2 TRMM Mission and Instrumentation
TRMM is a joint mission of the United States and Japan. The Science Team is led by Project
Scientist Dr. Joanne Simpson of the NASA Goddard Space Flight Center. The TRMM spacecraft will be launched in August 1997 on a Japanese H-II rocket into a rapidly precessing 35°
inclination, 350 km altitude circular Earth orbit. The spacecraft will complete 16 orbits per
day, resulting in the ground track pattern indicated in Figure 20.2. TRMM is part of the Earth
Probes series of NASA launches, designed as relatively low-cost missions with targeted science