Documentation of the BYU-MERS "SIR" image format

The BYU-MERS "sir" image format was developed by the Brigham Young University (BYU) Microwave Earth Remote Sensing (MERS) research group to store images of the earth along with the information required to easily earth-locate the image pixels.

A SIR file consists of one or more 512 byte headers containing all the information required read the remainder of the file and the map projection information required to map pixels to lat/lon on the Earth surface. Pixel values are generally stored as 2 byte (high order byte first) integers though can be stored as bytes or IEEE floating point. Unfortunately, the latter is not portable to all machines and so is not recommended. Scale factors to convert the integer or byte pixel values to native floating point units are stored in the file header. Utilities to read SIR format files are available in C, Fortran, Matlab, and IDL/PV-WAVE from the BYU MERS web site or the NASA Scatterometer Climate Record Pathfinder web site or their corresponding ftp sites: BYU MERS ftp site and the NASA Scatterometer Climate Record Pathfinder ftp site Details of the header byte format are documented in the the various file readers.

The origin of the images are in the displayed lower left corner. The earth location of a pixel is identified with its lower-left corner.

The new version of the standard sir format supports a variety of image projections including: (though typically images are produced in only one projection)

In general, *.sir image data files have been generated using the scatterometer image reconstruction (SIR) resolution enhancement algorithm or one of its variants for radiometer processing. The multivariate SIR algorithm is a non-linear resolution enhancement algorithm based on modified algebraic reconstruction and maximum entropy techniques (Long, Hardin, and Whiting, 1993). The singlevariate SIR algorithm was developed originally for radiometers (Long and Daum, 1997) but also used for SeaWinds (Early and Long, 2001). The SIR w/filtering (SIRF) algorithm has been successfully applied to SASS and NSCAT measurements to study tropical vegetation and glacial ice (e.g. Long and Drinkwater, 1999). Variants of SIR has been successfully applied to ERS-1/2 scatterometer and various radiometers (SSM/I and SMMR). SIRF is used for SASS, NSCAT, and SeaWinds slice data processing. SIR is used for most other sensors and does not include median filtering.

References:

For scatterometers, the multivariate form of the SIR algorithm models the dependence of sigma-0 on incidence angle as sigma-0 (in dB) = A + B * (Inc Ang - 40 deg) over the incidence angle range of 15 to 60 deg. The output of the SIR algorithm is images of the A and B coefficients.

A represents the "incidence angle normalized sigma-0" (effectively the sigma-0 value at 40 deg incidence angle). The units of A are dB. Typically, +2 < A < -45 dB. However, in the SIR images A is typically clipped to a minimum -32 dB with values of A < -32 used to indicate 'no data'.

The B coefficient describes the incidence angle dependence of sigma-0 an has the units of dB/deg. At Ku-band global average of B is approximately -0.13 dB/deg. Typically, -0.2 < B < -0.1. B is clipped to a minimum value of -3 dB/deg. This value is used to denote 'no data' as well.

Single variable SIR or SIRF algorithms are used for radiometers and produce only an A (in this case, the brightness temperature) image. Typically, this can range from 165 to 320. Single variable SIR and SIRF algorithms are used for SeaWinds egg and slice images, respectively. In both cases the A images are at the nominal measurement incidence angle for the sensor and in the sensor measurement units.

SIR images arestored in row-scanned (left to right) order from the lower left corner (the origin of the image) up through the upper right corner. By default, the location of a pixel is identified with its lower-left corner. The origin of pixel (1,1) is the lower left corner of the image. The array index n of the (i,j)th pixel where i is horizontal and j is vertical is given by n=(j-1)*Nx+i where Nx is the horizontal dimension of the image.

Be sure to use binary ftp to transfer *.sir files!