why does ALOS PALSAR data have pixels with Data Number=0

[xushanthutud]

Hi Guys,
I am using ALOS PALSAR data on Canada for my research, however, I found at some regions (bare soil around the fluxnet tower CA-Obs in Canada) has pixels of which the Data Number is 0. My core download lines are:
opts = {
'platform': asf.PLATFORM.ALOS,
'instrument': asf.INSTRUMENT.PALSAR,
'beamMode': ['FBD', 'PLR'],
'processingLevel': 'L2.2',
'start': '2005-01-01T00:00:00Z',
'end': '2011-12-31T23:59:59Z'
}
results = [asf.geo_search(intersectsWith=wkt_square, **opts)]

My purpose is to repeat the BIOMASAR algorithm, i.e. select the pixels (around CA-Obs flux tower) of ALOS of which the land cover classification is bare soil (based on ESA CCI land cover 300m map), then do the nanmean (or nanmedian) of the ALOS backscatter at those "bare soil" pixels to assume the nanmean (or nanmedian) value is the backscatter of bare soil. However, I found some pixels has data number 0. I am not so sure whether it indicates these pixels shall not be used (i.e. I need to filter out those pixels with DN=0)? or the actual data number is 0 at those pixels...
Here is a map illustrating my problems. The pixels with DN=0 are plotted as red while other pixels are plotted by viridis cmap. For your further reference, I attached the normal ALOS scene I downloaded in the same region.
It seems that in this case, I need to filer out those DN=0 regions...but I am wondering is that a general rule? to filter out the DN=0 pixels...

Thanks!

[ASF - bhauer]

One of the SAR experts at ASF has been investigating and hasn't been able to reproduce the conditions you describe yet. What he would like to have is a more complete description of how you have processed the data. In particular, since he's not familiar with the BIOMASAR algorithm, what does this entail? Providing a reference would be helpful.