2.1 Study Area
Nakhon Si Thammarat, a southern province in Thailand, flooding is a recurrent event affecting the entire province (Langkulsen et al., 2022). Every year (Feb 2017, Jan 2017, Dec 2018, Dec 2020, Dec 2021, Dec 2022, Feb 2022, Nov 2023, etc ), it causes lives and damages to infrastructure, agricultural production and severely affects local economic development (Centre for Research on the Epidemiology of Disasters (CRED), 2023). This province has been selected as the focus area of our research to tackle challenges associated with extreme precipitation and flooding and to empower policymakers and stakeholders to make informed decisions to enhance resilience and adaptability to a changing climate.
Nakhon Si Thammarat province is characterized by a diverse topography and climate. Bordered by the Gulf of Thailand to the east and the Andaman Sea to the west, this province exhibits a range of elevations. The topography includes low-lying coastal areas along the Gulf of Thailand, which are close to sea level, as well as higher terrain in the central and western regions, reaching elevations of approximately 1800 meters above sea level. The seasonal distribution of rainfall plays a crucial role in shaping the landscape, ecosystems, and overall environmental conditions in Nakhon Si Thammarat, contributing to the region's unique climatic characteristics.
The climate in this region is tropical, typical of Thailand, with distinct variations throughout the year. The average temperatures range from 25°C to 32°C. The province experiences a tropical climate, with warm temperatures prevailing throughout the year. The mean annual precipitation is around 2400 mm. The annual rainfall pattern in Nakhon Si Thammarat displays distinct seasonal variations, with the highest precipitation occurring in October, November, December, and January (ONDJ), collectively contributing to over 60% of the total annual rainfall. In contrast, February experiences the lowest rainfall, representing a drier phase in the climate cycle (Figure 1b). Noteworthy changes in the annual rainfall pattern has been observed, particularly from the year 2000 onwards. There is a discernible upward trend in precipitation during January (Figure 1c). This shift in historical precipitation patterns is significant, indicating an increase in January rainfall compared to previous years.
2.2 Datasets
Daily observed precipitation data from the Nakhon Si Thammarat Province station (552201), obtained from the Thai Meteorological Department (TMD), was used for this study. The station data covered the period from 1980 to 2022. Additionally, the daily precipitation data from Climate Hazards Group Infrared Precipitation with Station (CHIRPS) was acquired from the Climate Engine app (https://app.climateengine.com), covering the period 1981 to 2022. The CHIRPS data was bias-corrected using the station data and it was used to fill the missing data of the observed station data.
In our study focusing on extreme precipitation and its trends as proposed by Expert Team on Climate Change Detection and Indices (ETCCTI) in Nakhon Si Thammarat province, Thailand, we employed the NASA Earth Exchange Global Daily Downscaled Projections archive (NEX-GDDP-CMIP-6). NEX-GDDP-CMIP6 dataset, which serves as a bias-corrected downscaled iteration of Global Climate Models (GCM) (Thrasher et al., 2022). To assess precipitation and its extreme variability in southern Thailand, we obtained the bias-corrected versions of 6 CMIP6 model experiments from source, https://www.nccs.nasa.gov/services/data-collections/land-based-products/nex-gddp-cmip6. These 6 models were also employed by Rojpratak & Supharatid, (2023); Ge et al., (2021); De Silva et al., (2023) and Supharatid et al., (2022) in their previous study focusing on the Southeast Asia (SEA) and Thailand region. Similarly, the NINO3.4 data for same 6 CMIP6 model was obtained from Climate Variability Diagnostics Package (CVDP) (Phillips et al., 2014) from https://www2.cesm.ucar.edu/working_groups/CVC/cvdp/data-repository.html to analyze the relationship between ENSO and annual precipitation. Our investigation spanned historical simulations covering the period from 1980 to 2014 and projected periods extending from 2015 to 2100 under two Shared Socioeconomic Pathways (SSPs): SSP245 and SSP585. The SSP245 (SSP585) is an update of the CMIP5 scenario RCP4.5 (RCP8.5), an additional radiative forcing of 4.5 (8.5) W/m² by the year 2100, with now combined with socioeconomic reasons in CMIP6 future scenario (Thrasher et al., 2022). As, these scenarios form an essential part of climate change study representing the medium (SSP245) and high (SSP585) pathways of future greenhouse gas emissions. Precipitation data was again bias-corrected with the observed precipitation data and further analysis was performed. Table 1 provides a detailed list of the CMIP6 ensemble members employed in our study, thus forming the foundation for our investigation into precipitation patterns, ENSO relationship and extreme variability in the study region.
Table 1 List of Bias-corrected CMIP6 models used
S.N.
|
Model
|
Name
|
1
|
ACCESS-CM2
|
Australian Community Climate and Earth System Simulator - Climate Model version 2 (Australia)
|
2
|
ACCESS-ESM1-5
|
Australian Community Climate and Earth System Simulator - Earth System Model version 1.5 (Australia)
|
3
|
EC EARTH 3CC
|
EC-Earth Climate Model version 3 Coupled Configuration (Europe)
|
4
|
INM-CM5-0
|
Institute of Numerical Mathematics Climate Model version 5.0 (Russia)
|
5
|
MIROC-6
|
Model for Interdisciplinary Research on Climate version 6 (Japan)
|
6
|
FGOALS
|
FGOALS Model (China)
|
2.3 Approach to analysis
Approach to analysis of various model outputs and datasets has been adopted from similar exercises undertaken before. For example, the types of extreme precipitation indices proposed by Zhang et al. (2011) as part of the Expert Team on Climate Change Detection and Indices (ETCCTI) offer ways to interpret the patterns. The ETCCTI is a group of climate experts convened by the World Meteorological Organization (WMO) with the goal of developing a suite of climate change indices that can be used to monitor and assess the impacts of climate change on different regions and sectors. Eight Precipitation indices were used for their study under four categories: duration, absolute, threshold, and percentile-based threshold indices. Table 2 provides specific descriptions of each of these indices. Climpact, an R-based software package was used to calculate the extreme precipitation indices.
Further, the Mann–Kendall (MK) statistical test (Kendall, 1938) was used to assess the monotonic upward or downward trend, and Sen’s slope estimator (Sen, 1968) was used to calculate the magnitude of trend for seasonal, annual precipitation and extreme precipitation’s time series. We have also calculated the change in precipitation and extremes by dividing the study period into the 50s (2025–2055) and 80s (2056–2086) in reference to the historical period (1980-2014) for CMIP6 analysis. Spearman's Correlation (Spearman, 1961), a non-parametric statistical method, is employed to assess and quantify the relationships between observed precipitation and ENSO.
Climpact is specifically designed to calculate climate indicators for various socio-economic sectors, such as health, agriculture, and water resources, utilizing daily temperature and rainfall data (Alexander & Herold, 2016). While it is preferable to employ extensive and complete instrumental observations as the primary data source, Climpact also accommodates the computation of indicators using alternative sources, including remote sensing data from satellites or reanalysis. In this study, Climpact was used to calculate extreme precipitation indices, demonstrating its versatility in climate research. Due to the missing precipitation data, climpact failed the quality control method. Bias correction was performed using CHIRPS dataset to fill the missing data for observed precipitation data.
To address biases between station data and both the CHIRPS dataset and CMIP6 precipitation data quantile mapping was done, adopting the methodologies outlined by Piani et al., (2010) and Dosio & Paruolo, (2011) in this study and employing the techniques they proposed for bias correction of precipitation data. Quantile mapping was employed for bias correction utilizing the fitQmapPTF function in the R programming language, with additional support from the doQmapPTF function. This method involves fitting parametric transformations to the quantile-quantile relationship between observed and modeled precipitation values. The transformation function, represented by tfun, adjusts the distribution of model data to align with observed data. The wet day correction was implemented to equalize the fraction of days with precipitation between observed and modeled data. The formula for the exponential asymptotic transformation, used as one of the predefined transfun options, is
where Po and Pm denote observed and modeled cumulative distribution functions, respectively a and b are coefficients determining the shape, while τ is a time constant influencing the rate at which the transformation approaches its asymptote.
An anomalous occurrence was noted during the initial application of quantile mapping for the entire data period, as it failed to accurately capture the underlying equation for bias correction due to the high monthly variation in precipitation. To address this limitation, a refined approach was adopted, wherein the dataset was divided into individual monthly segments. Subsequently, bias correction was executed separately for each of the 12 months. This month-wise segmentation aimed to enhance the precision of the quantile mapping process, ensuring a more effective correction of biases within the data.
Table 2 List of precipitation indices used
Indices
|
Indicator Name
|
Definitions
|
Units
|
Duration indices
|
CDD
|
Consecutive dry days
|
Maximum number of consecutive days with PR < 1 mm
|
day
|
CWD
|
Consecutive wet days
|
Maximum number of consecutive days with PR ≥1 mm
|
day
|
Absolute indices
|
RX1 day
|
Max 1-day precipitation
|
Monthly maximum 1-day precipitation
|
mm
|
SDII
|
Simple day intensity index
|
Annual total precipitation divided by the number of wet days (defined as PRCP ≥ 0 mm) in the year
|
mm/day
|
PRCPTOT
|
Annual total wet day precipitation
|
Annual total PRCP in wet days (PR ≥ 1 mm)
|
mm
|
Threshold indices
|
R10
|
Number of heavy precipitation days
|
Annual count of days when PRCP ≥ 10 mm
|
Day
|
R30
|
Number of very heavy precipitation days
|
Annual count of days when PRCP ≥ 30 mm
|
Day
|
Percentile-based threshold indices
|
R95p
|
Very wet days
|
Annual total PRCP when PR > 95th percentile
|
mm
|