Home   |   About us

Measurement Uncertainty and Error Propagation
of Satellite-based Precipitation Sensors


Project Description Data Visualization and Download
Philosophy, Methodology and Data Publications and Presentations
Progresses and New Results References

Philosophy, Methodology and Data


1. Uncertainty is ubiquitous

Observing and measuring Nature generate data; data contain information; information forebears knowledge. No observations or measurements are absolutely definite. Uncertainty, small or large, is always inherent in any observations or measurements. It determines how much information can be extracted from the data, how much knowledge can be distilled from the information, and ultimately, how much clarify is in our knowledge of Nature.

2. How to quantify uncertainty

In practice, uncertainty is the spread, or disagreement, or inconsistency, between independent measurements of the same physical quantity. Thus uncertainty can be be quantified by gauging the spread between the measurements. Such spread is caused by each measurement's specific, varying deviation from the true value.

3. Error modeling

A measurement's deviation from the true value is the error. It is often conveniently modeled as two parts, systematic error and random error. With such a decomposition the uncertainty is essentially produced collectively by the random errors and the difference in the systematic errors in each measurement

The two error components, systematic and random, do not necessarily reflect the nature of the error sources. For Earth science data records (ESDR) of satellite-based precipitation measurements, various efforts have been devoted to study both components of the error, as summarized in Table 1 below.

4. Error sources in passive-microwave (PMW) remote-sensing of precipitation

Besides instrumental errors, ultimately the errors in PMW-based retrieval of precipitation depend on two sources: the precipitation event itself, and the microwave (MW) radiometric signature of the land surface. The former includes factors such as ice particle shapes and size distributions, liquid water content, water vapor profile, etc. Their variation from event to event, and their differences from what are assumed in the retrieval algorithms, will manifest as (largely random) errors. These micro-physical characteristics can roughly be described by the macro-scale classification of rain types (stratiform, convective, MCS, etc.), and may in turn, be roughly surrogated by some seasonal and climatological classifications.

The land surface MW radiometric signatures are produced by complex processes, involving factors such as surface temperature, surface type, soil composition, soil moisture, snow/ice cover, soil freeze/thaw state, topography, and a host of vegetation parameters. Ongoing efforts, such as those made by the GPM land surface working group (LSWG), have been searching for compact and efficient representation of the land surface MW radiometric signatures.

Table 1. Studies of errors in precipitation ESDRs

 
5. Data records to be investigated

Table 2. Level-2 Precipitation Data Sources

Sensor

Start Date

End Date

Source

TRMM PR

8 Dec 1997

Ongoing

GES DISC TRMM_2A25

 

TRMM TMI

 

8 Dec 1997

Ongoing

1. GES DISC TRMM_2A12

2. CSU GPROF2010 V1a

SSMI

DMSP F13

3 May 1995

19 Nov 2009

CSU GPROF2010 V1a

SSMI

DMSP F14

7 May 1997

23 Aug 2008

CSU GPROF2010 V1a

SSMI

DMSP F15

23 Feb 2000

13 Aug 2006

CSU GPROF2010 V1a

SSMIS

DMSP F16

20 Nov 2005

Ongoing

CLASS TDR, in GPROF2004V

SSMIS

DMSP F17

19 Mar 2008

Ongoing

CLASS TDR, in GPROF2004V

SSMIS

DMSP F18

8 Mar 2010

Ongoing

CLASS TDR, in GPROF2004V

AMSU-B

NOAA-15

1 Jan 2000

14 Sep 2010

CICS before 1 June 2007; CLASS

AMSU-B

NOAA-16

4 Oct 2000

30 Apr 2010

CICS before 1 June 2007; CLASS

AMSU-B

NOAA-17

28 Jun 2002

17 Dec2009

CICS before 1 June 2007; CLASS

MHS

NOAA-18

25 May 2005

Ongoing

CICS before 1 June 2007; CLASS

MHS

NOAA-19

25 Feb 2009

Ongoing

CICS before 1 June 2007; CLASS

MHS

MetOp-2/A

5 Dec 2006

Ongoing

CICS before 1 June 2007; CLASS

AMSR-E

1 June 2002

3 Oct 2011

1. NSIDC AE_Rain.2 V10 GPROF

2. CSU GPROF2010 V1a

(Table data courtesy of Eric Nelkin, David Bolvin and George Huffman at NASA/GSFC)
Table 3. Our existing collection of level-3 precipitation data records. These records include gauge- and radar-based products to be used as ground reference (CPC-UNI, CPC, STIV, STII, and PRISM), as well as satellite-based products.



Earth System Science Interdisciplinary Center
5825 University Research Ct., Ste. 4001
College Park, MD 20740