Skip to content
English
Go to EarthScan
  • There are no suggestions because the search field is empty.

What is probabilistic modeling?

Random variables and probability distributions are the building blocks of probabilistic models

After reading this article, you will learn:

Probabilistic modeling

Probabilistic modeling is an approach that incorporates randomness to calculate a large set of possible outcomes. Probability models go beyond what has occurred within the available historical data by taking into account new circumstances and uncertainties in order to avoid underestimating risk. 

Probabilistic models represent variables within the model as probability distributions, rather than discrete values. In terms of decision-making, probabilistic modeling incorporates uncertainty about future events, while still providing enough information to judge whether or not an event is likely, and make decisions on this basis.

Probability distributions

For each physical metric dataset, various statistics (e.g., maximum or integrated values) are calculated over a historical period (e.g., 1980-2020). These data are then combined to create probability distributions for a given metric (e.g., maximum temperature) at a given location (e.g., Glasgow, Scotland) over the entire period. 

Probability distributions are a statistical function that describe the full range of possible values and potential outcomes for a specific metric (such as maximum temperature), over a specific range (in the case of the multiple futures model, this is geographical location and time). 

Return periods

Once the probability distributions have been constructed, we use other statistical methods to extract key information - the return periods. Return periods are a method climate scientists and risk management experts use to visualize probability. Return periods, also called recurrence intervals or repeat intervals, estimate the average time between events of a similar intensity occurring. This probability is often given in years e.g. 10-year flood, 50-year flood, and describes the average time between events of a similar magnitude. 

In EarthScan, return periods represent a set of specific probabilities related to the severity of each climate hazard event in question. Return periods calculated from our model input data are iteratively refined to ensure the corresponding return levels are representative for each metric.

Learn more about return periods in EarthScan.