POD, FAR, and CSI

The probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI) are examples of measures used to evaluate binary (yes/no) forecasts or classification tasks.

For example, if A is the number of forecasts that rain would occur when it subsequently did occur (forecast = yes and observation = yes, often called hits), B is the number of forecasts of rain when rain did not occur (forecast = yes and observation = no, also known as false alarms), and C is the number of forecasts of no rain when rain did occur (forecast = no and observation = yes, also known as missed events), then:

POD = A/(A + C)

FAR = B/(A + B)

CSI = A/(A + B + C)

For perfect forecasts, POD = CSI =1.0 and FAR = 0.0. POD and FAR scores should be presented as pairs.

These scores have been rediscovered and renamed over time in a variety of contexts and disciplines, leading to confusion. For example, POD was also called prefigurance in meteorology, while it is named recall in information science and machine learning, and sensitivity and true positive rate in the medical literature. CSI is also commonly known as the threat score. For more information, see Brooks et al. (2024).

Reference:

Brooks, H. E., M. L. Flora, and M. E. Baldwin, 2024: A rose by any other name: On basic scores from the 2 × 2 table and the plethora of names attached to them. Artif. Intell. Earth Syst., 3, e230104, https://doi.org/10.1175/AIES-D-23-0104.1