Merged Total Precipitable Water 1-deg Monthly Climate Product

Version-7 Release-1 TPW CDR

This data product is constructed from Remote Sensing Systems (RSS) Version-7 microwave radiometer total columnar water vapor values (also referred to as total precipitable water, TPW). The product is updated monthly and available at One netCDF file contains all the data and is named, where YYYY is the most recent year and MM is the most recent month of the data in the file. We will update this file monthly. The V07 signifies that the data are made from the Version-7 RSS data (this will not change until the RSS input data are a new version). The r01 signifies this is the first reprocessing version of the TPW product (V00 was the original version). We added AMSR2 to the product and changed the post-hoc adjustments to those given in Wentz, 2015.  Due to these changes, we incremented the revision number in Jan 2016. We expect to update the climatology data once per decade and will increment the version number at that time.


We have constructed a merged, 1-degree total precipitable water (TPW) data set using the version-7 (V7) passive microwave geophysical ocean products publicly available from Remote Sensing Systems ( The TPW values come from the following satellite radiometers; SSM/I F08 through F15, SSMIS F16 and F17, AMSR-E, WindSat, and AMSR2. These microwave radiometers have been carefully intercalibrated at the brightness temperature level and the V7 ocean products have been produced using a consistent processing methodology for all sensors. These high quality ocean data are made available to users thanks to funding from the NASA MEaSUREs project. The strong spectral signature of water vapor makes it a robust parameter to retrieve from microwave radiometer measurements. Comparisons with small-island radiosonde measurements and and ground-based GPS water vapor data demonstrate rms errors of ~1.0 mm. We find GNSS values are in better agreement with V7 water vapor values than with those of previous RSS versions. This document highlights the construction of the merged product which contains monthly, 1-degree TPW means, a 12-month climatology made using 1988 to 2007 data, monthly anomaly maps, a trend map with associated global and tropical TPW time series and trends, and a time-latitude plot. The data set was originally released in January 2013 in netCDF format and was reprocessed in January 2016.  It is accessible by FTP (, OpenDAP, and browse imagery. The product is updated monthly and will continue for as long as microwave satellite radiometers are in operation.

Development of the Merged TPW 1-deg Data Set

This TPW product is made using a two-step construction process. First, we make monthly 1-deg maps from the individual satellite water vapor values, keeping track of number of obs, number of ice obs and the mean day of month. We re-grid using cosine latitude weights to reduce from the 0.25 deg grid of the RSS satellite products to the 1-deg grid of this product. In the second stage of processing, we apply quality control measures, apply post-hoc adjustments, and combine TPW values from all instruments using simple averaging. The resulting TPW product was constructed using the following requirements: We only calculate a TPW value for a specific 1-deg grid cell if the cell contains more than 160 observations during the month, if ice is present for <= 30 of the included observations, and if the calculated mean day of the month (derived by averaging the time of the data falling within the cell) is within 6 days of the center day of the month. Three satellite-months did not meet our requirements: F08 in Jan 1988, F08 in Oct 1990 and F10 in Dec 1991. In each case, the month values are needed for consistency of the time series so an exception was made and the data were included. These quality checks were applied in the second stage of processing. In developing our methodology, several aspects of data product construction were explored, including geographic sampling, minimal data number requirements, ice flagging, rain flagging, averaging methods, analysis of offsets and need for small bias corrections. These are further described in the following paragraphs.

The quality of the vapor product produced is dependent on the number of data that are averaged into each grid cell. Land and ice proximity affects this number. Radiometers suffer from side lobe interference that prevents obtaining vapor values near land. Due to variations in instrument resolution, look angle, geographic conditions and spatial footprints, some pixels have more observations than others. This results in varying numbers of observations for a given grid cell and poorer quality averages near coastlines and along ice edges. We tested a variety of minimum observation requirements. The figure below shows the count of water vapor data in each one degree grid cell during the time 1990 to 2005. As you can see the number of data falling in any given cell is usually much more than 300. It is only along the coastlines and ice that this number drops and poor quality data can enter the data product. We experimented to see how different thresholds affected resulting trends. We found little difference once a minimum threshold of 160 counts per cell was met.

Number of TPW observation per grid cell (1990-2005)

Ice is likely to exist more at one end of a month than another (with the exception of floating icebergs) and ice removal is necessary. We developed a mean-day-of-month quality calculation to remove ice edge grid cells where the amount of ice increases or decreases throughout the month. To handle icebergs which move between grid cells, we used the number of ice observations within a grid cell during the month to exclude when too much ice existed (number of ice observations must be <= 30).

Our analysis determined that light rain has little effect on water vapor values measured by microwave radiometers, and the removal of values measured in rainy conditions can have adverse effects on this product. Comparison of SSM/I F13 vapor values to GPS vapor in rain-free and rainy conditions convinced us that little difference can be attributed to rain. For wind, we find that we need to remove data in grid points adjacent to rain to ensure the highest quality data. This is not the case for water vapor. We also found that exclusion of data adjacent to rain creates a geographic sampling problem by removing data from primarily rainy, high vapor tropical areas which results in lower mean vapor values and higher global trends. For this reason, we only excluded data from grid points with very heavy rain, so high that our microwave instrument processing does not derive an accurate vapor value. Data in regions with light to moderate rain are not excluded.

We use only RSS Version-7 water vapor values in making this product. The data are from all SSM/I instruments, the F16 and F17 SSMIS, AMSR-E, WindSat, and AMSR2. Analysis of individual instrument TPW values showed that some small adjustments remained necessary before merging the data.  The analysis compared all instruments listed to the TMI TPW data (TMI is not used in this product).  We believe that some of the differences may be due to the later sampling time of AMSR-E or AMSR2 (1am/pm as opposed to 6-9 am/pm of the other radiometers) or are due to slight differences in calibration - use of only rain-free data as opposed to using all data (rain-free and rainy) for the intercalibration. This minor calibration difference may account for why the AMSR instruments and WindSat are slightly different from SSM/I and SSMIS. 

Data Issues / Known Problems

3 satellite/month exceptions are needed in order to maintain time series consistency and avoid large data gaps.



Jan 1988



Oct 1990



Dec 1991

In each case, these are the only (or the best) satellite/month data available. In the case of F08, no other radiometer was in operation at that time.

This TPW product is intended for climate study as the input data have been carefully intercalibrated and consistently processed. Strict Q/C requirements are used to omit data that are not suitable for climate study. For more information, see the description of the development of the product.

We update the product monthly, adding the data for the last month to the time series. Trends are recalculated. The 12-month climatology supplied in the file and used for anomaly calculations is calculated by averaging all observations that pass our Q/C requirements from the years 1988 to 2007 (20 years). We will recalculate the climatology every 10 years.

Some users have had trouble reading the netCDF4 format file. Therefore, we provide both a netCDF3 file and a netCDF4 file. Both files contain identical data and information.

Data Product Format Description

This product is constructed by merging all valid TPW (water vapor) data from SSM/I, SSMIS, AMSR-E, WindSat, and AMSR2 instruments. The netCDF file contains mean total precipitable water (TPW) values on a 1-degree grid, a 12-month TPW climatology, monthly anomalies, a TPW trend map, and a time-latitude (hovmoeller) plot. The product contains data for the global oceans at 1-deg spatial resolution and represents monthly means for January 1988 to the date of most recent processing. The product is updated monthly. There are no data values in regions of land, persistent ice, and coastal areas.

Total precipitable water (TPW) is reported in kg per square meter (kg/m-2). 1 kg/m-2 is equivalent to 1 millimeter (mm). The total columnar water vapor or total precipitable water from the microwave radiometers is the most robust of all the geophysical parameters derived from radiometer brightness temperatures. TPW values range from 0 to 75 kg/m-2.

File Format and Contents:

We provide 2 netCDF files, both contain the same information, one is netCDF3 format and the other is netCDF4 format.

Temporal coverage for this data set is monthly maps from 1988 through to the present. 
The product file is re-created every month by the 15th adding the previous month's value. 
The number of months in the data file will continue to increase until there are no more satellite radiometer instruments in operation.

Content of the netCDF File:

1) monthly means of TPW, grid size 360 x 180 x number of all months since Jan 1988 (increases over time)

2) a 12-month set of climatology TPW, grid size 360 x 180, the climatology is an average calculated over the 20-year period 1988-2007

3) monthly anomalies of TPW derived by subtracting the above climatology maps from the monthly means, grid size 360 x 180 x number of months since Jan 1988 (increases over time)

4) a TPW trend map, grid size 360 x 180, the trend is calculated from Jan 1988 through Dec of the latest complete year of data

5) a time-latitude plot (a minimum of 10% of latitude cells is required for valid data), grid size 180 x number of months since Jan 1988 (increases over time)

In addition to the above variables, we include:

1) satellites_used:

information about which satellites were included that month in calculating the mean 
the array is sized #months x 10 satellites, where 1= used, 0=not used 
The satellite instruments are in the following order: 1=SSMI F08,2=SSMI F10,3=SSMI F11,4=SSMI F13,5=SSMI F14,6=SSMI F15,7=SSMIS F16,8=SSMIS F17,9=AMSRE on AQUA,10=WindSat on Coriolis,11=AMSR2 on GCOM-W1

2) global_mean_precipitable_water_anomaly:

the monthly Average Near-Global Mean Precipitable Water Anomaly over Ice-Free Oceans, 60S to 60N, array size = #months

3) tropical_mean_precipitable_water_anomaly:

the Monthly Average Tropical Mean Precipitable Water Anomaly over Ice-Free Oceans, 20S to 20N, array size = #months

The file size is approximately 151 MB, and grows as each new month of data is added to the file.The remaining variables are time, latitude, longitude, climatology_time, and bounds for each of these.

File Naming Convention:, where:



total precipitable water



file revision number



four-digit year, first value is start of data, second is end of data



two-digit month, first value is start of data, second is end of data



netCDF file format

Read Routines

Read routines are available in Idl, Matlab and Python at:

Any third party tool/software capable of reading netCDF files can be used (such as Panoply).




Python and

Browse Images/ Graphic Image Maps

We have developed a browse environment for this data product. The environment allows the user to switch between monthly mean vapor plots, anomaly plots, the 12-month climatology and the TPW trend. More capability will be added to this environment at a later date. The 'Get Data' button connects to the ftp server for data file download.

Product Validation

We have compared the TPW from this product with TLT air temperature values to see if the two parameters are still strongly correlated as reported in previous papers (Mears et al, 2007). The figure below shows this agreement for the deep tropics (-20S to 20N) ocean regions only. The value of the line at each point is the linear trend, starting in January 1988, and ending at the time indicated on the x-axis. Here the TLT trend is shown in black and the TPW (vapor) trend in orange:

TPW and TLT trends in the tropics (20N to 20S)


More product validation results will be available soon.


Wentz, FJ, 2015,  
"A 17-yr Climate Record of Environmental Parameters Derived from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager",  
Journal of Climate, vol. 28, pg. 6882-6902.

Wentz, FJ, L Ricciardulli, KA Hilburn and others, 2007, 
"How much more rain will global warming bring?", 
Science, vol. 317, pg. 233-235.

Wentz, FJ, MC Schabel, 2000, 
"Precise climate monitoring using complementary satellite data sets", 
Nature, vol. 403, pg. 414-416.

Wentz FJ, 1997, 
"A well-calibrated ocean algorithm for SSM/I", 
Journal of Geophysical Research, vol. 102(C4), pg. 8703-8718.

Wentz, FJ and RW Spencer, 1998, 
"SSM/I rain retrievals within a unified all-weather ocean algorithm", 
Journal of the Atmospheric Sciences, vol. 55, pg. 1613-1627.

Wentz, FJ and T Meissner, 2000, 
"AMSR Ocean Algorithm, Version 2", 
report number 121599A-1, Remote Sensing Systems, Santa Rosa, CA, 66 pp.

Wentz, FJ and T Meissner, 2007, 
"Supplement 1 Algorithm Theoretical Basis Document for AMSR-E Ocean Algorithms", 
report number 051707, Remote Sensing Systems, Santa Rosa, CA, 6 pp.

Hilburn, KA, FJ Wentz, CA Mears, T Meissner, DK Smith, 2010, 
"Description of Remote Sensing Systems Version-7 geophysical retrievals", 
presented at 17th Conference on Satellite Meteorology and Oceanography and 17th Conference on Air-Sea Interaction, Annapolis, MD.

Mears, CA, BD Santer, FJ Wentz, KE Taylor, and MF Wehner, 2007, 
"Relationship between temperature and precipitable water changes over tropical oceans", 
Geophysical Research Letters, vol. 34, L24709, doi:10.1029/2007GL031936.



Carl A. Mears, Kyle A. Hilburn, Deborah K. Smith, Lucrezia Ricciardulli 
Remote Sensing Systems 
444 Tenth St, Suite 200 
Santa Rosa, CA 95401 USA

Technical Contact:

Marty Brewer and Michael Densberger
Remote Sensing Systems User Support

Acknowledgement and Citation

This research was supported by National Aeronautics and Space Administration (NASA) Making Earth System data records for Use in Research Environments (MEaSUREs) Grant #NNX09AC19A. Microwave radiometer data are processed by Remote Sensing Systems with funding from the NASA Earth Science MEaSUREs Program and the NASA Earth Science Physical Oceanography Program.

Please acknowledge use of this data set when used in your work.

For example, include this citation:

Remote Sensing Systems, 2016, updated MMMYYYY [this should be the date of the file]. Monthly Mean Total Precipitable Water Data Set on a 1 degree grid made from Remote Sensing Systems Version-7 Microwave Radiometer Data, V07r01, [accessed on DATE ACCESSED]. Santa Rosa, CA, USA. Available at