Here is a random node’s output for pixcvec. Blue is from the Sac pixel cloud, red is from gdem “pixel cloud”. For some reason, the pixel cloud pixcvec does not include anything in the channel–that’s why the widths and areas are so biased.

Here is what my rdf file says–this is just copied from the swot_pixc2rivertile.py docstring:

width_db_file             (-) = None
use_width_db              (-) = False
reach_db_path             (-) = D:/data/SWOT-prior/PriorDistributionFolder/netcdfv2
class_list                (-) = [2, 3, 4, 22, 23, 24]
use_fractional_inundation (-) = [True, True, False, False, False, False]
use_segmentation          (-) = [False, True, True, False, True, True]
use_heights               (-) = [False, False, True, False, False, False]
min_points                (-) = 100
clip_buffer               (-) = 20.0
ds                        (-) = None
refine_centerline         (-) = False
smooth                    (-) = 0.01
alpha                     (-) = 1
max_iter                  (-) = 1
scalar_max_width          (-) = 600.0
minobs                    (-) = 30
trim_ends                 (-) = False
min_fit_points            (-) = 3
do_improved_geolocation   (-) = False
geolocation_method        (-) = taylor
height_agg_method         (-) = weight
area_agg_method           (-) = composite

Any ideas what might be causing this?