Combined Hazard Classification

The cannon2010 (or c10) module provides functions to classify relative hazard based on debris-flow likelihoods and potential sediment volume. This classification is a multi-part process. First, likelihoods and volumes are divided into classes. Then, the likelihood and volume classes are added together for each debris-flow. Finally, these sums are divided into relative hazard classes.

You can use the hazard function to classify relative hazard. The function requires a set of debris-flow likelihoods (on the interval from 0 to 1), and potential sediment volumes (in m^3). These inputs should be numpy arrays, and are often produced by the staley2017 and gartner2014 modules. For example:

from pfdf.models import s17, g14, c10
likelihoods = s17.likelihood(...)
volumes, Vmin, Vmax = g14.emergency(...)
hazard = c10.hazard(likelihoods, volumes)

By default, the function uses the following classification scheme:

Class	Likelihood	Volume (m^3)	Combined Score
1	[0, 0.25]	[0, 10^3]	1-3
2	(0.25, 0.5]	(10^3, 10^4]	4-6
3	(0.5, 0.75]	(10^4, 10^5]	>6
4	(0.75, 1]	>10^5	N/A

where parentheses () indicate an open interval, and brackets [] indicate a closed interval. The above table lists the default thresholds, but you can also use the “p_thresholds”, “v_thresholds”, and “h_thresholds” options to specify different thresholds for the likelihood (p), (v)olume, and (h)azard classifications. For example:

# Use custom thresholds to classify hazards
hazard = c10.hazard(
    likelihoods,
    volumes,
    p_thresholds=[.3, .6, .9],
    v_thresholds = [2E3, 2E4, 2E5],
    h_thresholds = [2,4,6],
)

Advanced users may also be interested in the pscore and vscore functions, which return the individual likelihood and volume classes:

# Classify likelihood and volume individually
p = c10.pscore(likelihoods)
v = c10.vscore(volumes)

And note that these functions also accept custom thresholds:

p = c10.pscore(likelihoods, thresholds=[.3, .6, .9])
v = c10.vscore(volumes, threholds=[2E3, 2E4, 2E5])