The MaxEnt Methodology in Generating Multiple Hazard Maps for Floods

The article Evaluation of multi-hazard map produced using MaxEnt machine learning technique, published las March in the journal Nature, proposes the elaboration of maps of multiple hazards from the use of the MaxEnt* methodology (Maximum Entropy) for the identification of areas sensitive to large-scale hazards.

As explained in the Nature article: “Maximum entropy has been used successfully to assess different types of natural hazards such as landslides, floods, gully erosion and soil salinity, due to its fast and easy implementation and robust mathematical functions and theoretical background”.

The objetives of this study are:

     (1) Explore the ability of the MaxEnt model to predict the spatial occurrence of floods, landslides, and gully erosion.

     (2) Better understand the relationships between these processes and their control factors.

     (3) Design a methodological perspective to prepare a combined multi-hazard map for land use planning and hazard mitigation.

According to the authors, events such as gully erosion, landslides and floods are physical phenomena, active in geological times but uneven in time and space. Humans can dastically modify natural ecosystems, and accelerate them. For example, deforestation, unsustainable agricultural management, or human-made construction can increase soil mobilization and sediment transport, leading to extreme processes of land degradation.

The combination of physical phenomena, modification of ecosystems by man and climate changes can constitute serius threats, in which it is common for several dangers to occur at the same time in the same place, which end up resulting in disasters or catastrophes.

Research has shown that using the MaxEnt machine learning technique, individual maps of susceptibility to floods, landslides and gully erosion, and subsequently a combination of them, can be developed to estimate a probability map of multiple hazards (MHPM), which will allow an integrated assessment and consequently better planning and preparataion to reduce risks. It is therefore an effective predictive method.

* MAXENT is the acronym for Maximum Entropy, an algorithm that has been adapted for the construction of potential distribution models by Steven J. Phillips and colleagues (Phillips et al., 2004, 2006, 2008).