Skip to main content

Mainstreaming and modelling: A systematic review of how gender analysis could inform hydrological modelling

Posted by: Clare Brandon

March 12, 2018

Evangeline Packett, CSIRO Summer Vacation Scholar

L-R, Nicky Grigg, Evie Packett, Joyce Wu and Toby Walmsley at CSIRO Black Mountain Seminar Series
L-R, Nicky Grigg, Evie Packett, Joyce Wu and Toby Walmsley at CSIRO Black Mountain Seminar Series

How can gender be mainstreamed into hydrological modelling processes themselves? There is often a disconnect between technical modelling work and the social science theories of gender and I investigated how these fields could inform each other.

It has been well established in the literature that water management outcomes can have different impacts on men and women. But accounting for those affects in a biophysical model is a complicated process. Gender relationships are cross-cutting, complex and heterogenous and human behaviour can be unpredictable. Social science can be a powerful tool that gives biophysical information context and makes it be more useful. But can it be mobilised in a model?

The foundation thinking of my work was that hydrological models are often used to test water management decisions and these decisions are often hugely significant:

  • they can be decisions surrounding major infrastructure investments that will last decades;
  • or decisions that require making trade-offs between people’s livelihoods;
  • or ensuring drinking water is keeping up with your city’s populations.

The information they provide can be powerful and is often essential. Given these high stakes, it is important to understand what the model output actually means.

  • Models are not perfect reproductions of the real world. The real world is simplified in models for specific purposes. So, understanding model results means understanding how the real system is perceived, defined and idealised.

And what does that mean for interpreting the results?

  • When a model is used to help decide on solutions, what values are being used to measure performance?
  • When a model shows costs and benefits, where are those costs and benefits being felt? Do you know the distribution of them? How do you know that you aren’t entrenching disadvantage for one group or another, how do you know that benefits are being spread equitably?

This information has the potential to change what the model output means and thus, how that information is used for making a decision.

I focused on the modelling steps of problem definition, option modelling and identifying preferred option. I found this method to be hugely useful for structuring my gender analysis of modelling but it is important to note that these steps involve more than just the actual model, they include how the system is perceived and how modelling results are used.

One of the ways I reflected on gender and models was to discover the role of values in modelling. While scientists are supposed to be apolitical actors, we are still individuals with our own perspectives and world views. In 2005, when Nancarrow wrote about her experience with modelling she asked: ‘Why is the researcher’s definition of the problem the right one?’ I wanted to interrogate how systems are perceived and defined within hydrological models, to understand if this could have a gender impact.

Beyond this I looked at the actual building of the model to see if gender analysis could inform any of the steps.

For example, calibration.

Calibrating a model involves adjusting a model’s parameter values until the results of the model adequately reflects the observational data it is trying to mimic. For example, when creating a flood hydrograph, the most important data for the model is the high flow events, so you want to make sure the model is getting those right. But how do you choose what data you want to mimic?

This involves a choice regarding which data is the most important and that choice is not always straightforward. For example, hydrological models are often calibrated to ensure annual average flows are consistent with observations. However, for eco-hydrological applications, literature shows that the most important factors for ecosystem health tend to be flow regimes and flow events[1]. So, depending on what the end use of the model was, the calibration choice will need to be different.

Likewise, there is the possibility that the type or timing of flow that is most important may be different for different people.

In the Chhattis Mauja irrigation system in Nepal, Zwarteveen (1997) found that despite men and women working cooperatively as co-farmers, they tended to have different priorities around water flow conditions[2]. This is because their work made them aware of the role of different flow regimes, men tended to care about water arriving at the start of the season while women also stressed the importance of ongoing access to water throughout the irrigation season. So, if you wanted to model this system, there are different gendered perspectives on which part of the flow is most important for farming. Not only does that affect your choice of calibration method, it also affects what variables or metrics you report and analysise when communicating and interpreting model results.

When a model is being used to help make a decision or choose a solution based on costs or benefits, then ensuring that that decisions have equitable benefits requires knowing something to do with the distribution of costs and benefits. But how would you discover that at any large scale?

Escobar et al. proposed a method of doing so in 2017. They proposed you could construct gender-specific user-profiles that relate a certain livelihood type (i.e. woman, this catchment, working in this industry, of this class) and then examine that user-profile’s relationship with water [3]. Then you could quantify how many people fit that user-profile and where they are (using census data) and thus have the model show disaggregated effects of a water management change.

Obviously, there would be challenges in using this methodology. It would require significant amounts of data to create rigorous, intersectional user-profiles and you would have to be able to estimate how those user groups interacted with water. I examined how indigenous water research and ecosystem models could inform could inform gender focused water modelling.

What is clear is that this process would be very complex. When discussing economic models, Collier (1990) wrote, ‘sometimes gender disaggregation will not add enough [to a model] to be worthwhile. However, for some topics, it will be useful and for others essential’[4].

These methods are complicated. Where gender relationships may not be fully understood, using them may not be possible. Sometimes including social dynamics in a model will not be useful and sometimes it will be impossible due to time and resource constraints.

It is also important to ask how does not including them affect uncertainty and risk analysis? If social dynamics are not included in a model that is designed to aid human development, then other questions need to be asked in terms of how a model is used and understood:

  1. How sure are you that the solutions or the costs and benefits shown are accounting for all people?
  2. Where uncertainty about effect distributions are high, then what other forms of information can be used to minimise this uncertainty?
  3. And where this uncertainty is unacceptably high, and cannot be reduced, then how useful is the model’s output for development decision making?

I think that this uncertainty should be understood as significant and be built into the risk assessment process.

[1] Bunn and Arthington 2002

[2] Zwarteveen 1997

[3] Escobar et al. 2017

[4] Collier, 1990, 145 – 150 : Quoted in Elson 1995


I couldn’t have hoped for my seminar to go better. Not only was it attended by key people at CSIRO and ACIAR who have helped shape my thinking and learning process but also leaders in the industry such as Tony Jakeman. Thus, it was the perfect space to present and discuss the key concepts and ideas I have developed for gender integration in hydrological modelling.

Discussion was lively and reflected the key challenges presented in the work such as, when is modelling the appropriate format for tackling development, when is integrating social components into models justified despite their uncertainty and complexity and what gives modelling impact?

These dialogues reflect the wider debates taking place in the hydrology field. For example, the discussions around Pahl-Wostl et al.’s (2007, 2010, 2011) paradigm shift in water management and Sivapalan et al. (2012) socio-hydrology, which have both argued for a greater focus on the human dimension and complexity in water systems to deal with future challenges.

Excitingly, the discussion moved beyond exploring the challenges and covered the possible applications and lessons of integrating gender into modelling. For instance, how to deal with uncertainty, where gender analysis could be applied to ‘personas’ in software management and what CSIRO can do to analyse our option modelling in the near term (explain). It was a pleasure to see these discussions occur and an honour to be able to present.

Slides Transcript

Recording from Evie’s presentation at the Black Mountain Seminar Series

This work is part of a portfolio of investments supported by the Australian Government addressing the regional challenges of water, food and energy security in South Asia.

The SDIP Phase 2 aims to improve the integrated management of water, energy and food in the Himalayan river basins, addressing climate risk and the interests of women and girls. It seeks to:

  • strengthen practices for regional cooperation
  • generate and use critical new knowledge to enhance regional cooperation
  • improve the regional enabling environment for private sctor engagement.

Find out more about SDIP in CSIRO by visiting our website

Access the SDIP knowledge products

Find more information about international water research at CSIRO