Table of Contents

The following topics are discussed in this section:

Handling Overlapping Basin Boundaries

The QGIS Zonal Statistics tool produces correct statistics even if the features of the Watershed Basin Shapefile Input are overlapping.

Below is an example of overlapping features in a shapefile. The dark grey lines represent basin boundaries and the red circles pinpoint the areas of overlap.

Overlapping Features

Often watershed basin shapefiles have overlapping polygon features due to the combining of datasets, overlaying of reservoir boundaries, etc. With the ArcGIS zonal statistics tool, this overlapping phenomenon creates inaccurate results. This is because the ArcGIS zonal statistics tool converts the vector input zonal dataset into a raster grid before processing the statistics. Raster grid cells can only be classified by one feature of the input zonal dataset. Therefore, if a SNODAS cell is overlapped by two or more basins of the input zonal dataset, only one of the basins will include the cell's SNODAS value.

Fortunately, the QGIS Zonal Statistics tool iterates over each feature independently to calculate the statistics creating accurate statistics for both independent and overlapping polygons. The code for the QGIS Zonal Statistics tool can be viewed here.

The Colorado watershed basin shapefile input used for the Colorado Water Conservation Board project has many overlapping basin boundaries. This is because it was created by merging basin from many different sources. Refer to SNODAS Tools User Documentation for more information on the creation of the watershed basin shapefile input for the CWCB project.

Handling SNODAS cells Overlapped by Multiple Features

The QGIS Zonal Statistic tool calculates statistics on the daily SNODAS raster cells inside of each basin feature. There are scenarios, however, where a SNODAS cell is split by a basin boundary causing the cell to be in two different polygon features at once. However, each raster cell can only be assigned to one polygon. QGIS zonal statistics use the location of the cell's center to determine which polygon the cell "belongs".

  • Note: The only scenario where this is not true is if the cell resolution is larger than the polygon area. In this scenario, the snowpack statistics are calculated with the weighted proportion method. For more infomation about weighted proportions, reference line 329 of the QGIS Zonal Statistics code.

An example of the daily SNODAS cells overlapped by multiple features is shown below. The red line is a basin boundary. The green dots represent the center point of each SNODAS cell. Using the cell center technique, cell 1 is used to calculate the zonal statistics of the upper-left basin whereas cell 2 is used to calculate the zonal statistics of the lower-right basin.

Overlapped Cells

Handling Large Bodies of Water

Snowfall on large bodies of water, like a lake or reservoir, will react considerably differently than snowfall on the ground.

Calculating SNODAS SWE Statistics

To mitigate data errors due to the large bodies of water phenomenon the SNODAS model applies an open water mask to the landscape assigning open water cells a null value. The QGIS Zonal Statistics tool disregards cells with no-data values when calculating all output statisiics - count, mean, etc. The SWE statistics, therefore, are only representative of non-water areas.

Note that the open-water mask is dynamic. A cell that is assigned an open-water null value one day could be assigned a snow statistic the next day. For this reason, the effective area of each basin is calculated every day. The effective area is the approximate area of land for each basin (null cells are not represented in the areal calculation). If the open-water mask changes, the effective area will also be changed.

Below are two images displaying the open-water mask phenomenon. The aerial image on the top is of the Eleven Mile Reservoir in Colorado (the basin is outlined in red). The image on the bottom is the Eleven Mile Reservoir boundary (also in red) atop a daily SNODAS raster grid. The grid is set to color null values, cells representing open bodies of water, as white. As shown, the open water body is not included in the SNODAS grid. Given the open-water mask, the effective area of the Eleven Mile Reservoir basin would be approximately half of the total area.

Aerial Imagery


Calculating Snow Cover Statistics

For each basin, the areal snow cover statistic (a percent of land convered by some measurement of snow) is calculated by dividing:

the sum of cells covered by snow in the basin
the total count of cells in the basin.

The sum of cells covered by snow is calculated by using the QGIS Zonal Statisitcs sum tool on the binary SNODAS_SnowCover_ClipandPrjYYYYMMDD.tif raster. The binary snow cover raster contains the value of 1 for any cell with a SWE value greater than 0 (there is some presence of snow on the ground). Therefore, the sum of the cells within each basin is indicative of how many cells within each basin are covered by snow. Cells that are not included in the sum are those valued at 0 (SWE values of 0) and those of no-data value.

The total count of cells in each basin is calculated by using the QGIS Zonal Statistics count tool on the SNODAS_SWE_ClipAndReprojYYYMMDD.tif raster. The QGIS Zonal Statisics tool does not count cells of no-data values.

Cells representing large bodies of water are not included in either the sum of cells covered by snow or the total count of cells. The aeral snow cover statistic, therfore, describes the approximate percentage of land in each basin (non-water) covered by some value of snow.

Handling Missing SNODAS Values

The no-data value of SNDOAS products is set to -9999. This is set in the SNODAS Tools' configuration file. If the SNODAS no-data value is changed, it is pertinent that the new value is entered in the configuration file under section SNODAS_FTPSite option null_value. The configured null_value variable is used in the SNODAS_raster_clip function and the assign_SNODAS_projection function of the script to set the no-data value of the clipped SNODAS grid to the original no-data value of the SNODAS national dataset.

Handling the One Week Change in Total Snow Volume Statistic

The One Week Change in Total Snow Volume statistic is calculated by subtracting the current date's total snow volume statistic by the T-7 day's total snow volume statistic.

If the T-7 date has not already been processed by the SNODAS Tools, then the One Week Change in Total Snow Volume statistic for the current date will not be calculated because it will not have the T-7 day's total snow volume value to refer to during the calcaulation process. In turn, the csv column representing the One Week Change in Total Snow Volume will be filled with NULL.

If one is using the SNODAS Tools to populate the repository of all available historical data, then the SNODAS Tools MUST be run from the most historical date to the most recent date. If the SNODAS Tools are run in sections starting with the most recent data, some One Week Change in Total Snow Volume statistics will not be calculated.

Creating the basin boundary GeoJSON file for the Web Application

The web application for the Colorado data uses Leaflet to dispaly a choropleth map of the daily snopwack statistics. In building the web application, a GeoJSON file of the basin boundary shapefile is required. This GeoJSON file provides the geometry for the basin boundaries that are displayed in the web application. The snowpack statistics are imported from the daily csv files (organized by date) and are appended to the GeoJSON file by the foreign key, LOCAL_ID.

If the watershed basin shapefile input is changed, the basin boundary GeoJSON file must be recreated and reloaded into the web application. Below are the step-by-step instructions on how to create a GeoJSON file from an exisiting shapefile.

  1. Open QGIS Desktop.
    QGIS Desktop
  2. Add the basin boundary shapefile layer. Click Layer in the top menu bar. Mouse-over Add Layer > and click Add Vector Layer….
    Add Vector Layer

    Browse to the basin boundary shapefile layer and click Open.
  3. The shapefile should display in the main screen and the layer name should appear in the Layers Panel in the left side-panel.
    Shapefile in Desktop
  4. Right-click on the layer name in the Layers Panel. A pop-up menu will appear. Click Save As…
    Save As
  5. The Save vector layer as … window will appear.
    Save VL As Window

Change the following settings and then click OK

Setting Change to:
Format GeoJSON
Save as Browse to storage location and give the GeoJSON an appropriate name.
CRS Browse to desired output coordinate reference system. Ensure CRS is WGS84 EPSG:4326 if using the GeoJSON in Leaflet application.
Layer Options-Coordinate_Precision Set to 5 or other desired output coordinate precision.
Select fields to export and their export options Optional. Default exports all attribute fields. Deselect attribute fields that are not be included in the final GeoJSON output.