The WFC3/UVIS CCDs and WFC3/IR array contain pixels that vary in their area on the sky given the geometric distortion in the images. However, the flatfielding process in the CALWF3 pipeline flattens the sky intensity, thereby suppressing counts in larger pixels relative to smaller pixels. Hence, photometry of point sources on _flt images will vary depending on the position of the star and the areas of the pixels at that location. For the UVIS channel, this represents a 7% effect across a diagonal of the mosaiced image. For the IR channel, the area of the pixels varies by 8% from the bottom to the top of the image.
The drizzling process removes the geometric distortion and leaves the sky flat, so photometry of any sources in _drz images is uniform across the image. This is not true of the _flt images, and therefore a field-dependent correction factor is needed to 1.) achieve uniformity in the measured counts of an object across the field, 2.) match the output drizzled counts. This correction is called the "pixel area map" and simply reflects the area of the pixels at the location of the source. By multiplying the _flt images by the pixel area map, users will recover the same counts on _flt images and _drz images.
For WFC3, the normalization of the pixel area maps has been set to unity at a reference pixel near the center of each image. The exact position of this pixel is (x,y) = (2072, 2046) on the UVIS 2 CCD of the UVIS camera, and at (x,y) = (557, 557) on the IR camera. Therefore a star centered at this pixel (i.e., near the center of the detector) will produce the same counts on an _flt and _drz image, even in the absence of a pixel area map correction. The correction simply acts as a relative scaling to translate the counts at any off-center position to what would be measured at the center. Applying the correction simply involves multiplying the _flt images by the pixel area map (see below), which then yields the same flux as would be measured on a _drz image. As the _drz images have units of electrons/s, the counts on any UVIS _flt image should also be normalized to 1 second (the IR _flt images already have units of electrons/s). The WFC3 photometric zero points reflect the response of the instrument at the reference pixel on the _flt images (and therefore also on the _drz images).
For WFC3, the SCALE specified in a column of the IDCTAB has been set to the square root of the area of the reference pixel above (0.039621" for UVIS and 0.128254" for IR). This scale currently serves as an input to drizzle and tells the drizzling process the size of the input pixels at the reference point. The value of SCALE also serves as the default output pixel dimension in drizzled images, however users may redefine this to whatever they wish (e.g., drizzle will appropriately scale the counts to preserve the flux in the input images).
_drz flux = (_flt flux)*(pixel area map), where the _flt image reflects the counts per second of exposure time.