FCDR Definition

Fundamental Climate Data Records (FCDRs) are sensor data (e.g. calibrated radiances, brightness temperatures, radar backscatter) that have been improved and quality-controlled over time.

User Communities

FCDR users are primarily algorithm developers who use the FCDR as input to TCDR algorithms. They include two user communities:


Beginning with the SSM/I instrument on board the Defense Meteorological Satellite Program's (DMSP) F08 spacecraft in 1987, passive microwave imager data has provided over two decades of nearly continuous global observations of geophysical parameters that can be used to monitor the climate system. This includes such fields as precipitation over land and ocean, sea ice and snow, total precipitable water, cloud liquid water, and surface wind speed over the ocean. The record of passive microwave measurements is expected to continue well into the future to provide long-term global observations for climate applications.

Combining data from SSM/I, SSMIS and other sensors for the purposes of climate analyses has revealed a number of significant issues leading to biases in the long-term record. Prior efforts have created intercalibrated data sets from the SSM/I sensors, but have not produced a well documented and transparent process that is necessary for data stewardship. An essential element of data stewardship is the ability to explain future improvements to the data processing in terms of specific errors or deficiencies that are corrected/improved. This includes preserving the raw data recorded by the sensor, quality control procedures, exception handling, geolocation, calibration, and documentation.

In creating an FCDR, the process of creating a long-term calibrated data set is in many ways more important than the resulting data product. This is because:

Prior work relevant to this project includes:

  Data Availability

The CSU FCDR data is availabile from the SSM/I instruments on board the DMSP spacecraft F08, F10, F11, F13, F14 and F15, and the SSMIS instruments on board the DMSP spacecraft F16, F17, F18, and F19 for the times shown below. The Gray shaded bars for F15, F16, F17, F18 and F19 indicate the availability of the Interim Climate Data Record (ICDR) data. The ICDR data is created using the same software as the FCDR data, but it has not been checked out to verify its suitability for climate applications.

Data Availability Diagram


A key aspect of this project is a collaborative approach that works to understand and resolve calibration and method differences in previous and current work by different teams. Comparing different approaches provides insight into methods used as well as uncertainty estimates. Intercalibration between sensors is achieved through comparing and understanding differences in approaches developed by research groups at NOAA NESDIS and CSU. Towards this goal, a number of different intercalibration approaches have been developed and/or applied to the SSM/I and SSMIS datasets. These include those listed below.


The following is a short summary of the intercalibration techniques applied to the SSM/I and SSMIS sensors. More detailed information, along with references, is available for both SSM/I and SSMIS.

  • Coincident overpasses with TRMM TMI: A double difference approach that uses the difference between coincident observations from TMI to eliminate diurnal variability along with differences between simulated TBs to account for differences in the channel characteristics. The simulated TBs are computed using both retrieved geophysical parameters based on an optimal estimation algorithm as well as model analyses.
  • Coincident overpasses over polar regions: Direct comparisons of TBs from sensors operating concurrently are made over the polar regions, which is the only region where the sun-synchronous DMSP spacecraft view the same scenes simultaneously.
  • Model double differences: Geophysical parameters from both global model analyses including MERRA and ECMWF Interim are used to compute simulated TBs from a pair of sensors operating concurrently. Differences between the observed and simulated TBs are then used to compute intercalibration differences. This approach relies on the model to account for diurnal variability, which avoids the need for a non-sunsynchronous satellite like TRMM, but relies on the model's diurnal cycle. Using multiple models helps to reduce/quantify this error.
  • Vicarious cold calibration: This approach employs an analysis of the cumulative histogram of observed TBs to compute a vicarious cold reference TB, which represents a statistical lower bound on the observed TB. A comparison of the vicarious cold temperatures from different sensors provides a calibration difference for the coldest TBs. Simulations based on model analysis are subsequently used to account for slight variations in the viewing angle of the sensor.
  • Vicarious warm calibration: This approach is similar to the vicarious cold calibration, but it uses highly vegetated (i.e. warm) scenes over the Amazon and other jungles/forests with high surface emissivities. A retrieval is used to account for diurnal temperature variations, etc.

The goal of using these multiple approaches is to quantify calibration differences beween sensors over both radiometrically cold (i.e. ocean) and warm (i.e. land) scenes as well as the uncertainty in the resulting calibration differences. Given the lack of an absolute calibration target for the microwave frequencies employed on SSM/I and SSMIS, using multiple approaches provides confidence in the results and a reasonable estimate of the uncertainties.

Data Processing

The processing from raw sensor data to the SSM/I FCDR files follows the steps in this diagram, and is described below. The SSMIS FCDR processing steps are very similar, although there are some differences in the modules within the code to create the FCDR files.:

Data Flow Diagram

BASE Files

The original TDR files from SSM/I and SSMIS are organized into orbit granules, have ephemeris information added (for later use to calculate geolocation), are reformatted into netCDF, and saved as BASE Files. No data from the original files is changed or removed.

Stewardship Code

The Stewardship Code is a well-documented software package that ingests the BASE Files, does quality control, applies corrections, computes updated pixel geolocation, applies intercalibration adjustments, and outputs the final FCDR file in netCDF4 for use by the broader community. Expert users can be given access to the BASE Files and Stewardship Code to work with beta versions without complicating use by general users. The following list includes a number of the modules currently in the stewardship code:
  • Quality control: (Sets/flags bad data to missing and flags potential problems)

  • Cross-track bias correction: (Adjusts for unphysical end-of-scan dropoffs)

  • Warm/cold load contamination correction: (Adjusts for intrusions into warm/cold loads)

  • Sun-angle correction: (Correction for emissive reflector and other calibration biases related to heating/cooling of the sensor and/or residual intrusion issues. Only applies to SSMIS sensors)

  • Geolocation: (Computes pixel geolocation based on attitude adjustments and TLE-based spacecraft ephemeris)

  • Antenna temperature to brightness temperature: (Accounts for antenna pattern including sidelobes and cross-pol)

  • Intercalibration: (Adjusts for sensor differences for both warm and cold TBs)

FCDR Files

The output of the Stewardship Code is a netCDF file for each granule, containing brightness temperatures together with quality information about the data. Detailed information on the FCDR data format and contents is available for SSM/I and SSMIS.


The test of a multisatellite/sensor FCDR is ultimately whether it can be used to produce an consistent geophyscial climate data record. To test the utility of this FCDR for various applications, two different geophysical retrieval algorithms were run on the entire FCDR dataset. The first of these is an optimal estimation retrieval for non-precipitating ocean scenes that retrieves estimates of total precipitable water (TPW), cloud liquid water path (CLWP), and ocean surface wind speed. The second retrieval is the latest operational GPROF precipitation retrieval algorithm over ocean only.

  • TPW, CLWP, and Wind Speed: These time series figures are monthly mean estimates based on non-precipitating ocean scenes. This is a physical retrieval meaning that it uses radiative transfer calculations to account for small differences in the view angle between sensors and between pixels across and along scans. The retrieval also accounts for the change in frequency between the SSM/I 85.5 GHz channels and the SSMIS 91.665 GHz channels
  • Precipitation:This time series shows monthly mean oceanic precipitation estimates between 40N and 40N with an 11-month running mean filter applied. The region was limited to between 40N and 40N to be able to compare results with estimates from TRMM TMI, which has a lower inclination orbit limited to this latitude band.

The results from both of the retrieval algorithms shown above do not correct for differnces in the local observing times of the various satellites. All of the DMSP satellites are in sun-synchronous orbits viewing a given location on the Earth at the same local times (one for the ascending pass and one for the descending pass) each day. In addition, the DMSP orbits decay over time resulting in changes in the local observing times. The DMSP Local Observation times are shown here indicating the time of the am pass for each of the nine satellites and how that change over time. Note that a satellite with a 8am morning pass will also pass over again at 8pm.

Diurnal cycle variability in geophysical parameters such as precipitation will lead to differences in the monthly mean estimates between satellites if their local observing times differ and/or change over time. An analysis of precipitation over oceans indicates up to around a 3% difference in monthly mean estimates between sun-synchronous satellites such as DMSP that have different local observing times. This will also be the case for the TPW, CLWP, and wind speed retrievals, although the amplitude of the diurnal cycle variabiltiy may differ. Over land regions the diurnal cycle variability is typically much larger. It is important for algorithm developers and other users of the FCDR data to understand that the FCDR is not corrected for either view angle variations or diurnal cycle effects. As a result, the algorithm must account for view angle differnces between sensors as well as diurnal variability or it can lead to artificial biases in the resulting climate products.