Vulnerability concept: According to the definition of the Intergovernmental Panel on Climate Change (IPCC), vulnerability to climate change can be synopsized as the “degree to which a system is susceptible to and unable to cope with adverse effects of climate change, including climate variability and extremes” (IPCC, 2001). IPCC advices that a system’s vulnerability is highly dependent on its exposure sensitivity and adaptive capacity to weather extremes. Although a thorough interpretation of vulnerability components is provided by IPCC, it still remains difficult to define the multifaceted nature of vulnerability. Both natural and social scientists agree that vulnerability is multi-dimensional and differential, which means that it is perceived differently across the physical space and between various social groups (Cardona et al., 2012). It is also scaled and time-dependent because various socioeconomic and biophysical impacts, unequal in magnitude, may appear at the same time. Moreover, it is highly dynamic because the impacts may occur instantaneously or cumulatively over the years. While the fuzzy nature of the vulnerability is acknowledged, efforts are on to define an understanding of the boundaries of a vulnerable system. The social sciences typically look at the inherent social and economic differences that will limit communities in developing countries to cope with external pressures. Typical indicators include information in the areas of socio-demographics, economic wealth, infrastructural facilities and information access. It is also frequent that in developing regions, such indicators are assessed with methodological tools used for poverty analysis since a given set of adverse phenomena such as weather extremes may decrease consumption below the poverty level.
3.1. Study areas and vulnerability to climate change Bangladesh is being repeatedly exposed to a multitude of natural threats and disasters like floods, soil and water salinity and extended drought periods, mainly driven by the country’s unique geophysical and climatic conditions. The vast Tibetan Plateau is drained through a massive river network that enters Bangladesh from the North, spreads across the country, and finally ends up in the Bay of Bengal. Recurring, intense monsoon rain events augment the situation and often lead to floods, primarily in the southern lowland areas (World Bank, 2010). On the other hand, a shortage of rainfall along with its uneven distribution and great evaporative demand, particularly in the northwest of Bangladesh, results in extended drought events both in space and time. The frequency, as well as the intensity of extreme events, is anticipated to get enhanced by climate change, as repeatedly noted in the literature (Biswas et al., 2009; Nguyen, 2006; Winston et al., 2010). For Bangladesh, climatic changes may trigger more intense flooding, for instance, by more intense snow-melt in the Tibetan Plateau outside the country or more erratic and intense monsoon rain events. Also, delays to the start of the monsoon rains and rising sea levels allow a further intrusion of saline waters upstream, leading to more frequent and enhanced drought and salinization effects (MoEF, 2009, Winston et al., 2010). Therefore, we have selected the agrarian regions of Rajshahi and Barisal in the North-Western and Southern parts of the country as pilot areas, which particularly suffer from drought and flood-saline occurrences, respectively. Within each region, three districts (Upazilas) were chosen, namely Godagari, Tanore and Gomastapur in Rajshahi and Kalapara, Amtoli and Patharghata in Barisal. The selected drought prone Rajshahi region and the three districts are presented in Fig. 1 as below: As presented in Fig. 1, the Godagari district is adjacent to the Ganges river, which comes through Bangladesh from the eastern states of India while Tanore and Gomastapur districts are more distant from it to the north.
3.2. PCA indicators of vulnerability indices in Bangladesh Three groups of vulnerability indicators were considered for the purpose of this study corresponding to exposure, sensitivity and adaptive capacity, respectively. We sought for reliable data sources that could offer a good coverage for the study areas, as some of the specifically local data were found to be scarce and/or unreliable. The selection of the indicators was based on an extensive literature review, consultation with experts, a household survey and knowledge gained along field visits in the study areas. The exposure group encompassed a set of biophysical and technical indicators, which are well recommended by the literature as representative ones. Namely, the temperature, precipitation and irrigation volume were the most selected parameters in developing regions of Africa and South Asia to interpret exposure conditions (Adger, 2006; Filmer and Pritchett, 2001; Opiyo et al., 2014). The exposure indicators are often deployed on steady time intervals (e.g. semi-annual, annual) of past observations so as to better ascribe any potential trends. In our case, the exposure indicators were based on 30 years of observations (1978–2008) or model simulations for the selected pilot areas. The indicators were developed in the framework of an international research and development project (RiceClima, 2014). The biophysical data required for the exposure assessment (Table 1) were obtained from two sources. The annual and seasonal mean temperatures and rainfall data were obtained from historic meteorological records at the Bangladesh Meteorological Office (BMO) via the Center for Geographical and Information Services (CEGIS, 2014). CEGIS has also estimated the yield loss metrics for the premonsoon (Aus) and monsoon (Aman) cropping seasons, and the net irrigation requirement (NIR) metrics for all seasons. The locally adapted GIS-based DRAS crop model and relevant climatic records were introduced to simulate crop growth and water use in the studied regions (CEGIS, 2013b, 2014). We note that winter (Boro) season yield losses were not simulated since all the yield would be lost in both regions without substantial supplementary irrigation in that season. The following initially selected biophysical variables presented high multicollinearity, and were therefore excluded from the analysis: - Temperature and precipitation in the monsoon season with most of the exposure indicators - Indicated levels of potential yield loss without irrigation for the pre-monsoon (Aus) period with most of the exposure indicators