A Gridded Snowfall Verification
Method using ArcGIS: Zone-Based
Verification & Bias Maps
Joe Villani
Ian Lee
Vasil Koleci
NWS Albany, NY
Motivation
 Utilize Geographic Information Systems
(GIS) technology for a gridded and
more representative snowfall
verification method
 Legacy method (adding reports and
dividing by # of reports) has become
antiquated:
Does not factor in spatial variability
between data points
Methodology
 Use ArcGIS 10.2 software ArcCatalog
and ArcMap
 Create contoured snowfall maps and
zonal statistics based on observations
Methodology
 Careful QC of snowfall observations is
necessary (from trained spotters,
public, social media, etc.)
 Summary of reports issued as Public
Information Statement (PNS)
 A Local Hydrometeorological Data
Message (RRM) file is generated
automatically from PNS
 Contains location name, snowfall
amount, LAT/LON
Methodology
 Important to have snowfall reports
outside forecast area
 Allows for more representative
interpolation
 Gather RRM files from surrounding offices
 Python script collects snowfall reports
from RRM files and compiles them to a
CSV file
 Import CSV file into a spreadsheet
RRM File
Example
Relevant Data
Extracted from
RRM File
Creation of Gridded Snowfall Map
 Generate
GIS shape
file of
snowfall
reports
from
spreadsheet
using
ArcCatalog
 Import into
ArcMap
Creation of Gridded Snowfall
Map
 Run an Inverse Distance Weighting
(IDW) function in ArcMap
 Can try other interpolation schemes such
as Natural Neighbor or Spline
 Creates a gridded (raster) dataset and
contoured snowfall map
 Color scheme can be matched to ER
standardized forecast snowfall ranges
Gridded (IDW) Snowfall Map
IDW takes
into account
distance
between
points &
interpolates
Zonal Statistics
Run Zonal
Statistics
(including
mean) on
the
Gridded
Map
Image source: http://help.arcgis.com
Verification by Forecast Zone
A verification
map is created
using ranges
corresponding
to local office
advisory &
warning
criteria
Verification by Forecast Zone
 Run Zonal Statistics as a table for
statistical values
 Output for Mean Snowfall and other
stats within each forecast zone
Verification
 Table is based on
statistics computed
from the Gridded
(IDW) snowfall map
 Mean snowfall used
for verification by
forecast zone
An Argument for the Gridded
Method…
 Legacy method
would result in
mean snowfall
skewed towards
higher totals in
eastern part of
the zone where
the snowfall
reports are
concentrated
 Gridded method
more
representative
based on
incorporating
lower amounts
west of the zone
Creation of Forecast Bias Maps
 Preceding an
event, export
Forecast
Snowfall from
GFE to a
netCDF file
 Script converts
netCDF file to
a shape file,
 Import into
ArcMap and
convert shape
file to raster
Creation of Forecast Bias Maps
 Create
Gridded
snowfall map
from
observations
after event:
Subtract
Forecast –
Observed
to compute
bias
Creation of Forecast Bias Maps
 Forecast –
Observed
Snowfall =
Bias
 Warm colors
indicate
over-forecast
 Cold colors
indicate
underforecast
Conclusions
 This Gridded GIS-based verification
method yields a more representative
depiction of snowfall spatially across
a region
 Aides in computing legacy zonebased verification in a more
comprehensive manner
Conclusions
 Used for official snow verification at ALY
during the 2013-2014 winter season and
ALY staff has given positive feedback
 Can be used in
Social Media
posts to give
our users a
graphical
representation
of snowfall
after an event
Conclusions
 Bias Maps show how well or poor a
gridded forecast verified spatially
 Bias maps of snowfall events can be
compiled over entire winter seasons
to compute potential positive or
negative biases
 Can be valuable for determining if certain
areas consistently show a bias
More Examples
More Examples
Acknowledgements
 John Quinlan – NWS Albany, providing
conceptual ideas for computing zone
average statistics
 Steve Welch – NWS Buffalo, method for
creating shape file for snowfall based on
reports and LAT/LON
 Luigi Meccariello – NWS Albany, assisting
in creating observed snowfall maps during
the 2013-2014 winter season
Questions/Comments?
Joe.Villani@noaa.gov
Ian.Lee@noaa.gov
Vasil.Koleci@noaa.gov