Integrated assessment of flood and drought hazards for current and future climate in a tributary of the Mekong river basin

Hydro-meteorological extremes, either an acute lack of water or excess of it, are characterized by spatial–temporal variability and manifest as droughts or floods , respectively. They have the potential to cause severe damage to the environment (Masud et al. 2015), have a significant effect on the hydrology and water resources of watersheds (Aalijahan et al. 2021), and cause socio-economic impacts (Marcos-Garcia et al. 2017). Between 2005 and 2014, 83% of the recorded disasters, 39% of the recorded human deaths and 70% of the documented damages were linked to weather, water and climate (WMO 2018), respectively. With climate change, hydro-meteorological hazards are likely to be more frequent and severe, and influence a greater area. A warmer climate will intensify the hydrological cycle, resulting in the spatial and temporal redistribution of global water resources (Chen & Sun 2017).

With almost one billion people living in flood-prone areas (Kittipongvises et al. 2020), flooding is one of the most destructive natural hazards in the world, affecting 45% of the global population (Hammond et al. 2014; Kittipongvises et al. 2020) and causing billions of dollars of damage each year (Davenport et al. 2021). Several studies have shown that precipitation extremes will increase in the frequency and magnitude in the future (Zhu 2013; Rudra et al. 2015; Burke & Stott 2017; Ohba & Sugimoto 2019). An increase in the extreme precipitation will be manifested as an increase in the flood risk.

Problems caused by flooding are increasingly exacerbated by increased water vapour in the atmosphere (Morita 2011; Aalijahan et al. 2023), a higher frequency of intense rainfall events (Chitwatkulsiri et al. 2021) and changes in land use, increasing surface runoff, resulting in greater flood extents and depths (Khan et al. 2018; Penny et al. 2023). In particular, South Asia is one of the most flood-prone regions in the world due to its natural geography (Shah et al. 2020).

On par with flooding, between the 2008 and 2018 period, global drought resulted in economic losses of about 8.42 billion USD (CRED 2018). A study by Lesk et al. (2016) showed that between 1964 and 2007, drought resulted in an estimated loss of 1,820 million metric tons of cereal globally. There are several types of droughts: meteorological, agricultural, hydrological, groundwater, ecological, socio-economical, etc. (Mishra & Singh 2010; Zargar et al. 2011; Crausbay et al. 2018). Under climate change, global agricultural droughts are projected to increase (Zhao & Dai 2017), several studies in Asian river basins have similar conclusions (Nam et al. 2015; Lu et al. 2016; Li et al. 2017; Kwon & Sung 2019).

Floods and droughts are common natural hazards with dire consequences in Thailand. The probability of drought in any given year is 0.45 for Thailand, the highest among Asia–Pacific countries (Pandey et al. 2007). While the droughts of 2004–2005 caused an estimated damage of about 220 million USD (Wichitarapongsakun et al. 2016), the devastating drought of 2015–2017 resulted in a damage of 3,300 million USD (EM-DAT 2019). On the other hand, flooding is also a serious issue in Thailand, experiencing 69 major floods since 1985 (Singkran 2017). The historic flooding of 2011–2012 affected 16 million people in 64 provinces out of 77 and caused a damage of about 45.7 billion USD (DDPM 2015). The World Bank estimated it to be the fourth most costly natural disaster in the world from 1995 to 2011 (Kittipongvises et al. 2020). Other studies in Thailand have shown that extreme precipitation is likely to increase in the future (Singhrattna et al. 2012; Komori et al. 2018; Igarashi et al. 2019), consequently increasing the flood hazard.

There have been many methods that examine flood and drought hazards: hydrologic–hydraulic modelling approach (Vojinovic et al. 2016; Jacob et al. 2020), hydrographs (Samu & Akıntuğ 2020), frequency ratio (Kongmuang et al. 2020), Flood and Drought Disasters (FDD) index (Guan et al. 2021), Normalized Difference Vegetation Index (NDVI) and LSWI (Chandrasekar et al. 2010; Navarathinam et al. 2015) and the modified Mann–Kendall (MMK) (Gemmer et al. 2008; Zhang et al. 2015). However, lately Geographical Information Systems (GIS) techniques are increasingly used in flood hazard, with the Standardized Precipitation Index (SPI) and Standardized Precipitation Index (SPEI) used in drought assessments (Nawai et al. 2015; Prabnakorn 2020; Prabnakorn et al. 2016, 2021; Shao & Kam 2020). MCDA-GIS is a combination of GIS with multi-criteria decision analysis (MCDA). GIS-based MCDA is a method increasingly used over the past two decades for flood risk and hazard assessment (Tang et al. 2018; Shadmehri Toosi et al. 2019, 2020; Kittipongvises et al. 2020) that provides a simple, effective and accurate means to investigate spatial distributions and characteristics using basic GIS analysis and mapping (Paquette & Lowry 2012; Samanta et al. 2016; Shadmehri Toosi et al. 2020). It is an especially useful method in areas where there are limited data (Shadmehri Toosi et al. 2020). MCDA-GIS provides more flexibility for decision makers to evaluate factors that cause flooding (Nigusse & Adhanom 2019), providing significant advantages over different methods in overcoming decision-making difficulties (Vavatsikos et al. 2019).

Modelling multiple climate hazards together is rarely done. However, Ming et al. (2015) provided a quantitative approach of multi-hazard risk assessment to assess crop losses caused by high winds and flooding. Typically, droughts and flooding are treated independently due to challenges, associated with their subjective nature and their coupled dynamics. Recently, Brunner et al. (2021) have suggested that droughts and floods should be studied in a joint framework to help learn about fast event transitions. For example, Forzieri et al. (2016) modelled multiple hazards (flooding, drought, heatwaves) over Europe using the Overall Exposure Index (OEI), which works under the assumption that hazards are mutually non-exclusive and the Change Exposure Index (CEI), which expresses the number of hazards at a given baseline period. Tabari et al. (2021) agrees with Forzieri et al. (2016), stating that few studies and combined and projected flood and drought risk owing to large uncertainties, non-linerities and complex spatial-temporal dynamics. Tabari et al. (2021) also develops a global framework to encapsulate changes in flood and drought hazards using Extreme Value Distribution (GEV), SPEI and SPI. Nevertheless, both the above studies failed to combine the hazards within one map/entity. More recently, Yang et al. (2023) have combined the drought and flood status with crop yield through a correlation approach using SPEI to identify the risk from each hazard through both linear and probabilistic analyses. The study concluded that there were uncertainties within the models used and the level of risk classification.

To conclude, floods and droughts are significant disasters with severe implications to the society. It is important to understand and quantify them, especially with the ever disheartening climate change outlook, so that measures can be designed to mitigate them. Consequently, this study considers the joint spatial patterns of the flood and drought hazards in the Mun River basin, Thailand, for present and future climate scenarios. While previous studies have looked at either flood or drought hazard separately, this study presents its novelty by providing a non-modelling approach to joint hazard analysis considering both flood and drought. The methodology, using MCDA-GIS, is applicable to other studies as it is not limited by detailed data needs as found with hydrological/hydraulic modelling. The study will be useful to identify the areas where both single flood and drought hazards are located, as well as areas at risk from combined hazards, which could be a basis for implementing measures to improve resilience and reduce the associated risk in the future against both hydro-meteorological extremes commonly occurring in basins across the globe.

Data sources and GIS tools used to create the driving map can be observed in Table 1. Land use classes were identified according to the land use maps supplied by the Land Development Department (LDD) of Thailand and were the ones identified by Penny et al. (2021) paddy fields, field crops, perennials and orchards, other agriculture, forest, water bodies, marsh and swamp, urban and miscellaneous. From land use data using the ‘Distance tool’, drainage density could also be calculated. Soil data were classified according to the Harmonized World Soil Database, a global soil database framed within a GIS, which contains up-to-date information on world soil resources, classification range from combinations of clay, loam and sand. A slope map was created using the surface toolbox within ArcMap 10.6.1 (ESRI) and an elevation map provided by ALOS-JAXA.

The study utilizes the gridded rainfall and temperature dataset prepared by Khadka et al. (2022). The observed rainfall data from 43 stations (acquired from Thai Meteorological Department) were spatially interpolated to 0.25-degree grids using the inverse distance weighting (IDW) method. The maximum and minimum temperature datasets from the Climate Prediction Center (CPC) Global land surface air temperature analysis (Fan & van den Dool 2008) were used after comparison with other data products to ensure accuracy and consistency.

Near-future (2021–2050) climate, projected by Khadka et al. (2022), was used in this study. The climate projection (rainfall and temperatures) is based on eight climate models participating in HighResMIPs (Haarsma et al. 2016) of CMIP6 (Table 2). The study has assessed the changes in the climate for the near-future (2021–2050) with respect to the baseline period of 1981–2010. The future projections are available for shared socio-economic pathway (SSP) 5–8.5, which represents the high emission scenario (O'Neill et al. 2017). Bias correction for raw temperature and rainfall data from climate model is carried out using the quantile mapping method (Ines & Hansen 2006).

Results from Khadka et al. (2022) suggest that the maximum and minimum temperature in the study basin will increase by 1.45 °C (0.8–1.9 °C) and 1.54 °C (1.1–1.9 °C) in the near-future compared to the baseline. While there will be no significant changes (0.5%, between −5 and +10%) in the annual rainfall, temporal variations will increase in the future with projected increase by 6–11% and decrease in the summer by 2–8%. Similarly, the ensemble of the selected climate models indicated that the extreme rainfall events, computed as 10-year return values of 1-day maximum and 5-day consecutive maximum rainfall, will increase by 23% (9–40%) in the near-future period. The results are suggestive of significant impacts of climate change on the rainfall extremes.

The weighting for driving factors can be decided via questioning from a series of experts or stakeholders (Alves et al. 2021); however, for this case study, the six chosen driving factors were considered equal weights (Januadi & Nabila 2020).

To characterize the drought events in the basin, SPEI (Vicente-Serrano et al. 2010) has been used. It is a multi-scalar meteorological drought index, which considers climatic water balance [i.e., difference between precipitation (P) and potential evapotranspiration (PET)]. Since it considers both temperature and precipitation, it is suitable for climate change analysis. PET is estimated using the Hargreaves–Samani (Hargreaves & Samani 1982) method. Although the SPEI can be calculated at different timescales, a 12-month timescale is more relevant for climate change studies (Ahmadalipour et al. 2017; Lee et al. 2019) so that short-term rainfall variabilities can be avoided. SPEI is calculated by fitting the data in GEV distribution (Stagge et al. 2015; Khadka et al. 2021).

While the higher resolution version of CNRM-CM6-1-HR shows a reduced flood hazard (Figure 8) compared to the multi-model ensemble, both models agreed that the flood hazard will increase in the basin. This corresponds to where the projected increase in the rainfall is 10%. The results from the two models from EC-Earth Consortium are dissimilar, with Earth3P-HR finding an increase in the ‘very high’ flood hazard area in Nakhon Ratchasima compared to EC-Earth3P. Nevertheless, percentage-wise (Figure 8) the two models have similar results, although the high-resolution version of the model does find a slight increase in the % area covered by ‘very high’, ‘medium’ and ‘low’ flood hazards (Figure 8).

There is a significant variation in the drought hazard level within the basin during the baseline period [Figure 9(a)]. Nakhon Ratchasima (western part) has the higher area under the ‘very high’ hazard category, while Nakhon Ratchasima, Si Sa Ket (eastern part) and Roi Et (north-eastern part) have a significant area under the ‘high’ drought hazard category. Buriram, Surin and Maha Sarakham provinces have a relatively ‘low’ drought hazard in the basin. Overall, it can be observed that the drought hazard in the basin is irrespective of the annual rainfall received by the area. For example, Si Sa Ket receives the highest annual rainfall, but a significant area is under ‘high’ and ‘very high’ drought hazard levels. Drought characteristics and, hence, drought hazards are more affected by the temporal variability of the climatic parameters.

For the near-future period, most of the climate models (except CNRM-CM6-1-HR and HadGEM3-GC31-LL) have consensus that the ‘high’ and ‘very high’ hazards area will significantly increase and mostly concentrate on the western part of the basin. The multi-model ensemble also indicates the dire scenario where hazard levels have increased significantly throughout the basin. Table 5 shows that the area under high and ‘very high’ drought hazard levels will increase from 27 and 4% during the baseline period to 43 and 37% during the near-future period (multi-model ensemble average), respectively. All of the climate models (except CNRM-CM6-1-HR) suggest that the area with ‘low’ and ‘very low’ drought hazards will decrease in the near-future. High-resolution version of climate models from Met Office Hadley Centre (MOHC) and EC-Earth Consortium projected a significant increase in the area with a ‘very high’ drought hazard level, while the model from CNRM projected least increase. Spatially, the high-resolution version of the model from CNRM shows a reduced hazard in the eastern part of the basin (where the projected increase in the annual rainfall is 10%); both versions agree that the drought hazard will increase in the western part. The results from two models from EC-Earth Consortium are similar, even though the low-resolution version shows a low hazard in the south-east part. Except the low-resolution version of the model from the MOHC, three models project a considerable increase in the drought hazard in the basin.

Under a climate change scenario, future droughts are projected to be longer (increase of 22%) and severe (increase of 63%), while the joint occurrence probability of the drought event, which exceeds the 10-year return period threshold of duration and severity, is also projected to increase by 37% (Khadka et al. 2022). These are mainly driven by an increase in evapotranspiration (by approximately 5%) and higher temporal variability of the rainfall (increase by approximately 35%). Consequently, areas under ‘high’ and very high drought hazards are projected to increase in the basin. The results concur with the findings by Khadka et al. (2022), which shows that the increase in the magnitude of the extreme events will be much higher than the increase in the average climate.

Although we present how agriculture faces multiple challenges in terms of floods and droughts, we also want to point challenges for the critical infrastructure (CI) (Kumar et al. 2021a). The IPCC states that climate change unequivocally impacts various aspects of the built environment: transport, energy, water/wastewater and communications (Hawchar et al. 2020). Therefore, we argue that it is also essential to gain an understanding of CI vulnerability to climate-related threats (current and future), in order to develop effective strategies to enhance the resilience of CI (Hawchar et al. 2020; Schipper 2020). Floods and droughts can cause both direct (direct damage to physical infrastructure) and indirect effects (hindering supply chains and raw material production) (Kumar et al. 2021a). The current CI within the Mun River basin for the baseline period is situated in ‘high’ multi-hazardous areas, for example, Nakhon Ratchasima, Surin and Si Sa Ket (Figure 10). It is likely in the future with land use changes, CI development (cities, communications, utilities, and industrial and institutional lands) will increase across the catchment (Penny et al. 2021). Steps towards mitigating hydro-meteorological hazards could include nature-based solutions (NBS) (Majidi et al. 2019; Kumar et al. 2021b; Vojinovic et al. 2021). A study by Penny et al. (2023) provides an extension of this research, which looks into NBS for the Mun River basin. Nevertheless, caution must be taken as though adaptation to climate change is necessary, if adaptation strategies fail, this can worsen situations, leading to maladaptation (Schipper 2020).

We recommend that a future study may also consider (i) using additional indices, (ii) different weights’ allocation for the hazard indexes, and (iii) other methods such as NDVI and LSWI, or (iv) investigating the role of antecedent rainfall and soil moisture conditions (Ávila 2015; Navarathinam et al. 2015) in flood hazards. Currently, the present work only considers physiographic variables at a certain point, focusing more on local flooding. However, we also see that the work could be improved by including variables that affect the flood hazard originating from upstream, localized flooding, i.e., flood waters that occur upstream of the catchment but, in fact, affect the flood hazard within the catchment area, for example, stream flow. For this, we recommend to consider studies that have included flow accumulation within their MCDA-GIS methodologies for assessing flood hazard and/or risk, including Nandi et al. 2016; Ozkan & Tarhan 2016; Kabenge et al. 2017; Dash & Sar 2020; Dung et al. 2022; and Baykal et al. 2023. In addition to the above, the work could be taken further by using a hydrologic-hydrodynamic model and or ‘Bluespot’ analysis (Balstrøm 2022; Penny et al. 2023).

We acknowledge that this study could be enhanced by using methodologies such as Saaty AHP and ANP (de Brito et al. 2018), and rather than using equal weights (used in this study), individual weighting or proportional scoring for MCDA-GIS variables should be used. Nevertheless, when compared to outcomes from other researchers, Khadka et al. (2021, 2022) and Prabnakorn (2020), both flood and drought are in agreement with similar studies/forecasts carried out in the Mun River basin. This consolidates/validates the methodology used here.

Formulation and implementation of necessary drought and flood adaptation plans in the basin has paramount importance in making the society more resilient. Provision of high efficiency irrigation systems by the integrative use of surface and groundwater resources could be a major step to combat the negative impacts of droughts in the near-future. Combining temporary storage solutions for excess flood waters in the open fields and increasing swamp/wetland areas could help store flood waters and recharge groundwater aquifers, in turn, combating drought.

The study has assessed individual and joint hazard levels of floods and droughts in the Mun River basin, Thailand, for baseline (1981–2010) and the near-future period (2021–2050). The flood hazard is computed using six physiographic and climatic factors. The method used for the flood hazard is useful for data-scarce regions, particularly when hydrologic and hydraulic modelling is not available and when rapid assessment is required. Similarly, the spatial drought hazard was assessed using the multi-variate approach that combines three drought characteristics (duration, severity and frequency). A joint hazard map is created by combining the information of drought and flood hazards.

Overall, there is a limited variation in the locations of flood hazard; however, between the baseline and near-future period, the majority of climate models suggest an increase in the ‘high’ and ‘very high’ flood areas. The area under ‘high’ and ‘very high’ flood hazard levels increases from 11 and 22% during the baseline period (1981–2010) to 16 and 24% during the near-future period (2021–2050), respectively. Flood hazard increases are observed in lower elevations in the east of the basin especially in Roi Et and Si Sa Ket, correlating to areas where increased extreme rainfall is predicted and along Mun River and its tributaries within Nakhon Ratchasima and Buri Ram.

Drought hazard during the baseline period shows an uneven spatial distribution within the basin. While majority of Nakhon Ratchasima and Si Sa Ket have the ‘high’ drought hazard, Buriram, Surin and Maha Sarakham have low levels, not following the average rainfall trends, which decrease from east towards west. This shows that the drought hazards are more defined by the temporal variability of the climatic parameter (particularly rainfall) than the spatial variability. Results clearly show that the drought hazard levels as well as their spatial extents will increase in the near-future period. The ensemble of climate models suggests that area under ‘high’ and ‘very high’ drought hazard levels will increase from 27 and 4% during the baseline period (1981–2010) to 43 and 37% during the near-future period (2021–2050), respectively.

When combining flood and drought hazards together, compared to the baseline period, the multi-model ensemble finds an increased area of 155% in the overall hazard across the catchment, this is most prominent in the west – Nakhon Ratchasima and Buri Ram, and north-eastern provinces – Khon Kean, Roi Et and Maha Sarakham. In addition, the CI in the basin, particularly situated in Nakhon Ratchasima, Surin and Si Sa Ket, is also at risk from hydro-meteorological hazards. Thus, adaptation for multi-hazard within multiland uses, potentially in the form of NBS, is needed across the catchment not just for the primarily agricultural regions but for areas where the CI is located. Findings present a grim outlook on future floods and droughts in the basin, and this is particularly important as the majority of the population and land area are tied to agriculture, which at present are mostly rainfed cultivation systems. Thus, the results show that the agriculture sector in the basin is highly vulnerable.

To summarize, our research delivers an integrated assessment of flood and drought hazards within the Mun River basin, providing an important step in mitigating risks in multi-hazard environments, the results of which can be employed to test scenarios of measures to reduce the impacts of drought and flooding, and the purpose of which can not only improve predictions but also help inform decision-making. We have reaffirmed that multi-criteria analysis in the GIS environment can be extremely useful in real-world water-related applications.

The presented work is conducted under the project titled ‘Integrated Management of Flood and Drought in the Mun River basin, Thailand’. The authors would like to acknowledge the funding agency the Natural Environment Research Council (NERC) under NERC COP26 Adaptation and Resilience Project Scoping Call. Some of the processed climatic data were accessed from the research project ‘Enhancing Resilience to Future Hydro-meteorological Extremes in the Mun River basin in north-eastern Thailand (ENRICH)’, funded by the National Research Council of Thailand (NRCT) and NERC.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.