severity module

Functions that estimate and locate burn severity

Function	Description
classification	Returns a dict with the BARC4 classification scheme
mask	Returns a mask of the specified burn severities
estimate	Estimates a BARC4-like burn severity raster

The severity module is used to generate and work with rasters that record BARC4-like burn severity. The BARC4 classification is as follows:

Class	Description
1	Unburned
2	Low burn severity
3	Moderate burn severity
4	High burn severity

This module has two main functions: mask and estimate.

Burn severity rasters are typically used to derive data masks used to implement various parts of a hazard assessment. For example, a burned pixel mask used to delineate a stream network, or the high-moderate burn mask used to implement the M1, M2, and M3 models in the staley2017 module. Use the mask function to generate these masks from a BARC4-like burn severity raster. Note that mask searches for burn-severity levels by name, and you can return the supported names using the classification function.

We recommend using field-verified BARC4-like burn severity data when possible, but these maps are not always available. If this is the case, users can use the estimate function to estimate a BARC4-like burn severity raster from dNBR, BARC256, or other burn severity measure.

---

pfdf.severity.classification()

Returns the BARC4 burn severity classification scheme

classification()

Returns a dict reporting the BARC4 burn severity classification scheme used by the module. Keys are the strings “unburned”, “low”, “moderate”, “high”, and “burned”. Values are the integers associated with each burn severity level.

Outputs:

dict – Maps burn severity levels to their integers in the classification scheme

pfdf.severity.mask(severity, descriptions)

Generates a burn severity mask

mask(severity, descriptions)

Given a burn severity raster, locates pixels that match any of the specified burn severity levels. Returns a Raster holding the mask of matching pixels. Pixels that match one of the specified burn severities will have a value of 1. All other pixels will be 0.

Note that the burn severity descriptions are strings describing the appropriate burn severity levels. The supported strings are: “unburned”, “burned”, “low”, “moderate”, and “high”.

Inputs:

• severity (Raster-like) – A BARC4-like burn severity raster.

• descriptions (str | list[str]): A list of strings indicating the burn severity levels that should be set as True in the returned mask

Outputs:

Raster – The burn severity mask

pfdf.severity.estimate(raster, thresholds=[125, 250, 500])

Estimates a BARC4-like burn severity raster

Estimate severity

estimate(raster)

Estimates a BARC4 burn severity from a raster assumed to be (raw dNBR * 1000). (Refer to the following syntax if you instead have raw dNBR, BARC256, or another burn-severity measure). This process classifies the burn severity of each raster pixel using an integer from 1 to 4. The classification scheme is as follows:

Class	dNBR Range	Level
1	[-∞, 125]	Unburned
2	(125, 250]	Low
3	(250, 500]	Moderate
4	(500, ∞]	High

NoData values are set to 0. Returns a Raster object holding the estimated BARC4 burn severity raster.

Custom thresholds

estimate(raster, thresholds)

Specifies the thresholds to use to distinguish between burn severity classes in the input raster. This syntax should be used whenever the input raster is not (raw dNBR * 1000), and also supports custom thresholds for the (raw dNBR * 1000) case. Note that the function does not check the intervals of raster values when thresholds are specified.

The “thresholds” input should have 3 elements. In order, these should be the thresholds between:

1. unburned and low severity,

2. low and moderate severity, and

3. moderate and high severity.

Each threshold defines the upper bound (inclusive) of the less-burned class, and the lower bound (exclusive) of the more-burned class. The thresholds must be in increasing order.

Inputs:

• raster (Raster-like) – A raster holding the data used to classify burn severity

• thresholds (vector) – The 3 thresholds to use to distinguish between burn severity classes

Outputs:

Raster – The BARC4 burn severity estimate