Monsoon Onset in Bangladesh: Reconciling scientific and societal perspectives

Abstract

People of Bangladesh depend on the monsoon, and many of them depend on the timing of the monsoon onset for their livelihoods. In this study, we are motivated by the idea of providing information to these people about the monsoon onset that they can use. In order to provide information about the monsoon onset, we firstly have to define it. Once we have a definition, we can apply it to our data. However, a problem arises. There are many different methods to define the monsoon and in Bangladesh these different methods can give very different results. To judge which scientific definition might be appropriate, we carried out a large-scale questionnaire and asked around 1200 people in rural Bangladesh about how they define the monsoon. The people defined the monsoon and its start date differently from region to region and also within regions. Thus, we need a robust way to compare them people’s perceptions with the scientific information. Hence, this method needs to take into account local meteorological conditions, as we want to compare them with local perceptions. This entails using high-resolution data to identify the monsoon onset. When we use high-resolution data, we often come across the problem of false onsets, which give large variations in the scientific time series. In order to extract local information from high-resolution data, we develop a method that yields scientific time series of the monsoon onset with reduced false onsets at the level of individual grid points. In theory, this gives us local estimates of the monsoon onset that we can compare with the peoples perceptions. To compare the local perceptions with the scientific times series, we start by constructing probability mass functions around the answers given by the people in the questionnaire survey. We refer to the functions as modified triangular distributions since they are based on triangular distributions, which are often used in risk analysis and cost projections in project management. With these distributions we simulate artificial time series from the people’s perceptions. We use these simulations to compare with the scientific time series from the different monsoon definitions. This comparison is mostly qualitative. In order to measure this comparison, we use log-likelihoods to construct a score. We use this score to investigate which scientific time series best compares with the people’s perceptions about the monsoon onset. The results give us an idea of which scientific monsoon definitions might be appropriate to use in a dialogue with the people of Bangladesh. We clearly found that involving the people in this process gives us a better understanding of how they perceive the monsoon. This understanding can help to guide research in a direction that will more likely result in useful information for these people. This process is particularly important in regions where people are particularly vulnerable to present climate variability and future change.