Modelling the human factor: sociohydrology and flooding

I've talked a lot about the human factor on this blog: floods don't become disasters until people get involved. This got me wondering if anyone had done any work on modelling human interaction with floods, and of course they have: social scientists armed with differential equations are a menace.

Sociohydrology is a relatively new approach which attempts to model the interactions between social and hydrological systems, assuming that humans are an integral part of the hydrological system. Some sociohydrological models attempt to take human behaviour into account when societies interact with floods, and this is what I want to look at today.

A loop diagram showing the complex interactions between factors affecting flood impacts on settlements in flood plains.Source: Di Baldassare et al. (2013)

One paper conceptualises the interactions between floodplain communities and floods, and attempts to quanitify a variety of factors which affect these (politics, wealth, level of technology etc.) to identify how people respond to floods (e.g building protection vs. moving away from the river) and the subsequent effects on flood impact. Unsurprisingly, the levee effect emerges as building higher levees results in a greater shock when a flood occurs. Conversely, vulnerability can actually decrease when more regular flooding occurs: people are more aware of flood risks and more used to coping with floods, a phenomenon called the adaptation effect.

The adaptation effect also appears in another example, where a model of coastal flooding in Bangladesh in communities protected by polders. As well as hydrological factors such as tidal surge levels, the model incorporates terms for flood protection infrastructure, how flooding affects agriculture, economic factors such as wage levels and crop prices, and the concept of "social memory", which is a measure of community awareness of flood risk.

There is, um, quite a lot going on here, but this conceptual model of the dynamics of flood risk in a coastal community in Bangladesh shows the wide range of factors affecting risk and vulnerability. From Yu et al (2017).

Interestingly, the authors draw on game theory to model how community members interact and contribute to polder maintenance. While the community as a whole is incentivised to maintain the polders to reduce flood risk, individuals can seek other economic opportunities less vulnerable to flood than agriculture ("defectors", in the model above), and act as 'free-riders', benefiting from maintenance without contributing; however, "cooperators" can ostracise defectors who fail to fulfil communal responsibilities. All these factors can be adjusted in the model to run various scenarios.

A polder (artificial island) in southwest Bangladesh. In reality, many polder communities collectively manage their flood protection infrastructure with considerable success. Photo: Tanmoy Bhaduri

What emerges from the model is that the resilience of the community is sustained best not with the highest flood protection and fewest floods, but when there are occasional, moderate floods. This is because without regular flooding, community flood awareness decreases to the point that there is limited motivation to maintain the polders. The "best" level of flooding also varies based on how desirable it is to seek other economic opportunities, reflecting the vulnerability of polder communities to economic change.

Can models like these help us make better decisions about managing flood risk? With the important caveat that none of the models described here are designed to influence policy-making, I do wonder about the implications. Certainly many current approaches to flood risk management fail to take into account the adaptation effect and the importance of risk awareness to long-term resilience. That said, it may be difficult to convince many communities of the importance of enduring occasional flooding to improve their coping capacity!

Mabel disapproves of reducing the complex heterogenity of human societies to a series of mathematical parameters.

More generally, I am a bit sceptical of attempts to draw broad lessons from models of human behaviour, which is rarely rational and highly dependent on context. Communities respond very differently to flood risk depending on social, cultural and economic factors, and capturing these details would require both enormous amounts of data and extremely complex models for each individual community.

Still, I'm fascinated to see how this field develops in the future: human activities are affecting the hydrological cycle like never before, and it's clear that new ways of modelling and managing how societies interact with water will be vital for future approaches to flood risk management.