https://orcid.org/0000-0001-6835-4508
Figures
Abstract
Myanmar is one of the most vulnerable countries to climate change, and its complex geography together with heterogeneous climate and precipitation patterns present major challenges for producing reliable climate change projections. In light of these challenges, high-resolution regional climate models are essential for improving our understanding of climate change and to provide a knowledge base for adaptation strategies. We employed the Weather Research and Forecasting (WRF) model to simulate the present climate (1981–2010), a mid of century (2031–2060) and an end of century (2071–2100) climate for the SSP2-4.5 and SSP5-8.5 scenarios. We tested out different domain settings and show that large domains are needed to accurately model the climate and, particularly, precipitation in Myanmar. The past climate is validated against station data and satellite based products, and the model demonstrates good skill in representing the climate over Myanmar, with the exception of a dry bias in the southern Ayeyarwady Delta. Generally, the model underestimates precipitation at the end of the rainy season in October, which is related to a mismatch in the atmospheric circulation, moisture availability, and therefore, moisture transport into Myanmar. The climate projections show distinct increases in 2m-temperature, with warming of 0.9 to 2.7 °C for the mid-century in an SSP2-4.5 to end of the century under the SSP5-8.5. Our simulations project that in April the temperature in the Dry Zone in the centre of the country increases disproportionally with a warming of up to 3.6 °C for the SSP5-8.5 end of century simulation, while for all other scenarios the strongest increase is found in May. Changes in precipitation show a non-significant wetting in the Dry Zone and a significant drying in the Shan Hills and the Tanintharyi Region for the two periods in the SSP2-4.5 and the mid-century simulation under SSP5-8.5 scenario. For the end of century simulation under the SSP5-8.5 pathway a general wetting of the north western part including the Dry Zone in the range of 40 to 60% is projected. Even if the annual sum shows an increase in precipitation, this is not true for all the months. Especially, January, July, August and November are months which are projected to have less precipitation in all future scenarios compared to present climate.
Citation: Messmer M, González-Rojí SJ, Nay Chi MA, Leonard S (2026) High-resolution regional climate modeling over Myanmar using WRF: Historical validation and future projections under different shared socioeconomic pathways. PLOS Clim 5(2): e0000820. https://doi.org/10.1371/journal.pclm.0000820
Editor: Jingyu Wang, Nanyang Technological University, SINGAPORE
Received: September 13, 2025; Accepted: January 6, 2026; Published: February 12, 2026
Copyright: © 2026 Messmer et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The Weather Research and Forecasting (WRF) model V3.8.1 is freely available online and can be downloaded from the users’ page: https://www2.mmm.ucar.edu/wrf/users/download/get_sources.html (last access: 13 September 2025). The post-processed outputs for precipitation and temperature from our WRF experiments to reproduce the figures are available at: https://doi.org/10.5281/zenodo.17113911. ERA5 data can be downloaded freely from the Copernicus Climate Change Service (C3S). Aphrodite can be downloaded from: https://www.chikyu.ac.jp/precip/english/downloads.html. IMERG can be downloaded from the Earth Observing System Data and Information System (EOSDIS) from NASA (https://doi.org/10.5067/GPM/IMERG/3B-HH/06, Huffman et al., 2019). MSWEP can be downloaded based on the instructions provided at: https://www.gloh2o.org/mswep/ (last access: 13 September 13 2025).
Funding: The authors acknowledge the funding from the Swiss Agency for Development and Cooperation (SDC) towards the Centre for Development and Environment (CDE) at the University of Bern. Martina Messmer was further supported by the Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (Early PostDoc.Mobility grant no. P2BEP2-181837). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Myanmar exhibits complex topography, bordered by the Bay of Bengal and the Andaman Sea, and traversed by the Ayeyarwady River system, which originates in the Himalayas, flows south through the central Dry Zone, and discharges into the Ayeyarwady Delta. The pronounced elevation gradients induce orographic precipitation along the hill and plateau slopes, whereas the central region remains comparatively arid. In contrast, coastal areas receive substantially higher rainfall due to their proximity to marine moisture sources. Consequently, precipitation across Myanmar is highly heterogeneous. The country is also strongly influenced by the South Asian monsoon, with rainfall typically commencing in May and persisting through October.
Myanmar is ranked among the top 10 countries in the world most affected by weather and climate extremes during the last decades, including extreme temperatures, droughts, cyclones, storm surges, and extreme precipitation events [1]. At the same time, Myanmar has limited capability to prepare and adapt to extreme weather events due to low economic power [1]. Additionally, climate change strongly affects the agricultural sector, which is the most important one in Myanmar, as it makes up for around a third of the gross domestic product, but also the majority of the population depends mainly on agriculture [2]. Therefore, it is vital to understand how anthropogenic climate change will affect the country.
To provide insights into the potential impacts of climate change in Southeast Asia and Myanmar, multiple climate model simulations are available from the Coupled Model Intercomparison Project (CMIP) Phase 6. The number of studies investigating climate change over Myanmar is very limited. The comparison of 18 CMIP6 models over Southeast Asian countries unveiled that most of these models show a cold bias of 3 up to 8 °C over Myanmar in the historical simulation compared to observational data, while some others show a hot bias along the coasts and the hilly regions of Myanmar [3]. Furthermore, the models show a considerable wet bias over Myanmar, except for a dry bias over the west coast along the Bay of Bengal. Only MPI-ESM1-2-LR shows a small general dry bias over whole Myanmar. Temperature projections of these models show a more or less uniform increases of 1.28 and 1.88 °C at mid century and 1.80 and 3.36 °C at the end of the century over Myanmar for the shared socioeconomic pathways (SSP) 2-4.5 and 5-8.5, respectively [3]. For precipitation, future projections show a gradual increase from 10 to 40% over the course of the 21st century, with little differences between both SSPs [3]. Another study [4], using historical simulations of 12 different CMIP6 models, reports highly heterogeneous correlations between simulated and observed precipitation across Myanmar, with the strength of agreement in annual and seasonal precipitation varying substantially among the models. Only MPI-ESM1-2-HR can roughly reproduce the gridded precipitation patterns based on gridded rain gauge data available from the Global Precipitation Climatology Centre (GPCC) for both annual and seasonal scales [4].
The spatial resolution of CMIP models is around 100–300 km, but considering the complex geography of Myanmar and Southeast Asia, climate model simulations at higher spatial resolution are needed. There are few studies that downscale such global models with a regional climate model to a resolution of around 60 to 25 km horizontal grid spacing [5–7, e.g.,]. Using the Weather Research and Forecasting (WRF) model to downscale a 10-year period in the present and at mid century, a study [5] found that the model effectively captures temperature patterns. However, it exhibits a cold bias in the maximum temperature and a warm bias for the lowest temperatures, and a dry bias along the western boundaries of the country in the rainy season. For their mid century simulation under a medium emission scenario (A1B), they found a temperature increase of around 1 °C for the maximum temperature and 1.5-2 °C for the minimum temperatures, with the strongest increases projected for the rainy season and March-April, respectively. Annual precipitation changes are projected to increase in all Myanmar in the range of 10% [5]. A later study investigates the WRF model’s ability to simulate the climate variability over Southeast Asia using ERA-Interim data from 1991–2015 at a 27 km spatial resolution [6]. The model captures the general precipitation patterns, with a wet bias in DJF, and a dry bias along the western coast due to a reduced moisture flux transport from the ocean in JJA, as well as a wet bias over eastern Myanmar [6]. Finally, there is a study that analyses projected precipitation changes over the Southeast Asia region of the Coordinated Regional Climate Downscaling Experiment (CORDEX) using 14 different regional climate model simulations at a spatial resolution of 25 km under the Representative Concentration Pathways (RCP) 4.5 and 8.5 [7]. The historical period shows accurate precipitation patterns, again with a dry bias along the west coast and a wet bias inland of Myanmar in the rainy season [7]. The RCP8.5 projections indicate a strong gradual increase in precipitation during DJF, and a wetting trend in the boreal summer. The RCP4.5 shows relatively small precipitation changes, but a significant increase along Myanmar’s northwestern coast [7].
While previous studies often use ensembles of different models, they all lack the high-resolution necessary to capture the complex and heterogeneous precipitation patterns in Myanmar. To better assess the vulnerability of Myanmar to climate change in a regionally more detailed way, high-resolution (≤ 5km) simulations are crucial. Setting up such high-resolution regional climate model simulations needs to be done with care, as the chosen domain size, nesting ratios and physical parameterization options affect the performance of the simulations [8–10]. This is especially true for domains that are not yet well tested, as it is the case for Myanmar. Especially, simulations that focus on precipitation are sensitive to changes in domain size but also horizontal and vertical resolution, whereby the domain size is the most dominant factor [9]. The size of the domain is particularly important when simulating precipitation at a convection-permitting scale [10]. On the one hand, smaller domains support a more accurate simulation of extreme precipitation [8, 10]. On the other hand, larger domains allow better capture of precipitation on a larger scale [10]. Smaller domains tend to enhance the hydrological cycle within the domain on seasonal scales by increasing the transport of moisture into the domain [8].
To address the existing gap in high-resolution model simulations over Myanmar, we employ the WRF model to dynamically downscale the Bay of Bengal and therefore large parts of Myanmar at a spatial resolution of 5 km, using a convection-permitting setting. Convection-permitting simulations have shown an increase in the accuracy of precipitation patterns, intensity and timing during the day [11–15]. We downscale the global climate simulation MPI-ESM1-2-HR for the present (1981–2010), mid century (2031–2060) and end of century (2071–2100) for the SSP2-4.5 and SSP5-8.5.
In the following, we will describe the models and data used in more detail, before we move on to investigate the added value of using 5 km instead of 25 km for such a complex terrain. We also provide a validation of the WRF simulation under present climate conditions using weather station, satellite-based and reanalysis data. We highlight the atmospheric circulation under present climate condition, as this is important to capture precipitation patterns correctly, and finally, we provide results on temperature and precipitation changes under global warming. The study is wrapped up by a conclusion and an outlook.
Materials and methods
WRF model
We employed the Weather Research and Forecasting (WRF) model version 3.8.1 [16] to dynamically downscale global climate model simulations over the Bay of Bengal and surrounding landmasses. Since the main region of interest is Myanmar, we selected the MPI-ESM1-2-HR model of the Coupled Model Intercomparison Project Phase 6 [17, 18]. It has been shown that this model is able to roughly reproduce the annual and seasonal precipitation patterns under present climate [4]. This lead to the assumption that potentially also the atmospheric circulation over the area is captured reasonably well. As initial and boundary conditions for WRF we used six hourly data of the MPI-ESM1-2-HR model at a 100 km spatial resolution, as this is the CMIP model with high enough temporal resolution (6 hours) and because it shows the least precipitation bias over Myanmar compared to other CMIP6 simulations. We selected physical parameterization options in WRF based on extensive sensitivity studies on nesting ratios and different cumulus, microphysics, long wave radiation and planetary boundary layer (PBL) schemes that previously were performed in other tropical and mountainous regions, such as East Africa [14] and Peru [15]. The two regions, East Africa and Peru, show similarities to Southeast Asia, as all three regions are characterized by complex topography and a rain regime that is influenced by monsoonal precipitation. It has already been shown that the optimal scheme for East Africa provides a good agreement for Peru as well, indicating that the optimal parameterization options can be transferred to other regions [15]. This transfer has further been shown between South Africa and Southeast Asia [6] and also our results for Myanmar prove that such a transfer is possible. Additionally, the parameterization options that are chosen here are rather standard and are often used for regional climate modeling over Southeast Asia with WRF [5–7]. The following parameterization options were used: Grell-Freitas for cumulus [19, only for D1], the WRF single-moment 6-class scheme for microphysics [20], the rapid radiative transfer model (RRTM) for long-wave radiation [21], Dudhia short-wave scheme for short-wave radiation [22], Yonsei University Scheme for the PBL [23] and Noah-MP for the land surface [24]. We use a nested domain approach with a parent domain of 25 km and a child domain of 5 km horizontal resolution. We cover 30 years under present climate from 1981–2010 and two future periods, one for the middle of the 21st century (2031–2060) and the other at the end of the 21st century (2071–2100). For the two future periods, we used two different SSPs, that is, the middle of the road scenario SSP2-4.5 and the high emission scenario SSP5-8.5 [25]. The scenarios used and the corresponding global warming level with respect to the present (1981–2010) and preindustrial (1850–1900) climate simulation of the global model are summarized in Table 1.
We ran three different simulations for the present climate with different domain extents to understand the sensitivity of climate in the Dry Zone to the chosen domain size. Here we call those domain designs. The first domain design (“Dry Zone") covers only the Dry Zone of northern Myanmar, where the outer domain is not much larger than the inner one (Fig 1A). The second design (“Dry Zone with outskirts") allows much more space between the parent and the nested domain, while the second domain is almost the same size as in the first domain design (Fig 1B). The last domain design (“Bay of Bengal") covers a larger part of the Bay of Bengal (Fig 1C). The different domain locations and sizes allow to better understand the sensitivity of the results with respect to the selection of the parent domain and its nest.
The shading indicates the topography in meters above sea level using the WRF topography Global Multi-resolution Terrain Elevation Data (GMTED2010) provided by USGS. The black box represents the extent of the nested domain. A: for the Dry Zone of Myanmar, B: the Dry Zone with outskirts, C: the Bay of Bengal. The violet stippled line in panel C indicates the extent of D2 of panel A and B. Additionally, different stations are marked which are used for the analysis. The base layer of the map was created using the open-source vector map data from Natural Earth data.
Weather station data
To better understand the validity of our climate model simulation, we use weather station data from Myanmar, provided by the Department of Meteorology and Hydrology of Myanmar for the period 1981–2010. We use six stations (Nr. 1-6 in Table 2) to show the sensitivity of the model domain to precipitation and temperature. We use monthly minimum and maximum temperature and monthly precipitation sums. Seven additional stations (Nr. 7-13 in Table 2) are used to validate the much larger second domain of the Bay of Bengal simulation (Fig 1C).
The elevation height is provided for the closest grid point of domain 1 (D1) and domain 2 (D2) of the Bay of Bengal domain design and the location of the station in meters above sea level.
Observation based gridded precipitation products
We use three satellite-based products to be able to compare precipitation patterns rather than just point measurements of the weather station data, but these data sets also allow to check the quality of the weather station data themselves.
The Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (IMERG) is a state-of-the-art precipitation estimation algorithm developed by NASA [26]. IMERG merges precipitation data from multiple satellites and thus, provides global precipitation estimates with high temporal and spatial resolution. The available period is from June 2000 until present. The algorithm integrates microwave and infrared sensor data to produce half-hourly precipitation estimates on a 0.1-degree grid (approximately 11 km). We use the final run of version 6, which is the product at research quality.
The Multi-Source Weighted-Ensemble Precipitation (MSWEP) dataset is a global precipitation product that integrates precipitation information from gauge, satellite, and reanalysis data sources [27]. MSWEP provides high-resolution precipitation estimates on a 0.1-degree grid (approximately 11 km) with a 3-hourly temporal resolution [27]. This dataset is available from 1979 until present. Thanks to the optimal combination of different precipitation measurement techniques, accurate and consistent precipitation estimates are obtained across the globe.
Additionally, we use the gridded precipitation dataset for Asian Precipitation—Highly Resolved Observational Data Integration Towards Evaluation of water resources (APHRODITE) [28]. It is a gridded precipitation data set with daily temporal resolution and is the highest spatial resolution product for Asia [28]. We use the version V1101 for the years 1981–2007 over the region “Monsoon Asia” at a spatial resolution of 0.25-degree grid (approximately 27 km). The data set is built based on a dense network of quality-controlled rain-gauge data, but it must be kept in mind that the accuracy of the dataset depends on the amount of station data in the region of interest [29].
Reanalysis data ERA5
To better understand potential biases in the atmospheric circulation of WRF, we need to compare it to another gridded dataset that has consistent values for wind and moisture at different height levels. Therefore, we employ the latest reanalysis product of the European Centre for Medium-Range Weather Forecasts (ECMWF), ERA5. ERA5 uses 4D-Var data assimilation and the cycle 41r2 of the Integrated Forecasting System (IFS) to obtain model output at 137 vertical levels and at a 0.25×0.25° resolution [30]. We specifically use the vertically integraded moisture and wind in x- and y-direction at 850 hPa. We consider the climatology of 1981–2010 to overlap with our present climate period.
Calculation of mean absolute error and root mean squared error
To better compare the performance of the different domain designs, we compute the mean absolute error (MAE) and root mean squared error (RMSE) between the weather station data and the grid point closest to it in the simulations of the different domain designs for the period 1981–2010. Since we are comparing climate change simulations with weather station data, we calculate the mean value for each month in the weather station data and compare it to the 30 values of each month of the climate model simulation. Finally, we compute an annual average over these values. We used this approach, as we cannot make a year-by-year comparison between the weather station data and the climate model data, as we must expect different weather in the climate model compared to reanalysis. On the climatological scale however the errors should be as small as possible. Using the MAE and the RMSE, we obtain a measure for the performance of the different simulations. MAE focuses more on the performance of the mean value, while the RMSE more strongly penalizes outliers. For the comparison between the different domain designs we focus on the high-resolution 5 km domain, as this is the domain of interest for us.
Results and discussion
Domain sensitivity and added value of high resolution data (25 vs 5 km resolution)
The comparison of the different resolutions of domain 1 and 2 of the three different domain designs compared to the weather station data shows a clear improvement for the higher resolution for monthly maximum temperature (Fig 2 and S1 in S1 File). The overestimation of the coarser domains is not surprising as the topography is resolved much more precisely in the 5 km domain, and therefore the elevation of the grid point containing the station is much closer to the elevation of the weather station data (Table 2, last column). This is particularly clear for Hakha, a station located in mountainous terrain. For Monywa, Hsipaw (Fig 2), Gangaw, Shwebo and Mandalay (Fig S1 in S1 File.) the coarser resolved domains show a clear underestimation of the maximum temperature during June to October, which are months in the rainy season. Albeit the high resolution domains (5 km) show a strong improvement in maximum temperature in terms of their bias compared to the coarser domain, overall the domain design of the Bay of Bengal (Fig 1C) outperforms the others. This better representation of both temperature and precipitation can be associated with a more accurate representation of moisture fluxes over that region. The comparison of the vertically integrated moisture fluxes along the boundaries of the innermost domain of the Dry Zone domain design highlights that both smaller domain designs import moisture through advection across both the southern and northern boundaries, and an export of moisture towards the east (not shown). However, for D2 of the Bay of Bengal domain design, the advection of moisture occurs primarily through the southern and to a smaller part through the western boundary, and it is exported through the northern and eastern boundaries.
The panels A-C show the monthly maximum temperature in degrees Celsius and panels D-F show the monthly precipitation sum in millimeters for the years 1981–2010. A and D: Hakha, B and E: Monywa, and C and F: Hsipaw. The orange colors indicate the domain covering the Dry Zone (Fig 1A), the pink colors indicate the domains of the Dry Zone with outskirts (Fig 1B) and the teal colors indicate the domain covering the Bay of Bengal (Fig 1C). The bright colors indicate the coarser domain of 25 km resolution and the dark colors represent the high resolution 5 km domain. The grey boxes represent the weather station data, whereas the grey shading shows the full range of the weather station data for each month for better comparison with the other datasets. The boxes represent the 25th to 75th percentile, while the whiskers extend to the value that is no more than 1.5 times the interquartile range away from the box. The comparison for the rest of the weather stations can be found in Fig S1 in S1 File.
Table S1 in S1 File shows very clearly that the MAE and RMSE is the smallest for the second domain of the Bay of Bengal domain design and therefore consistently performs the best. Additionally, the second domain of the Bay of Bengal domain design is within the range of the maximum temperature of the weather station data in 76 to 95% of all months for the different 6 stations (not shown). For the Dry Zone design, this is between 67 to 95% and for the Dry Zone with outskirts it is between 66 to 95% (not shown).
In terms of monthly precipitation sums, the stations Monywa, Hsipaw (Fig 2), Gangaw, Shwebo and Mandalay (Fig S1 in S1 File) show a clear overestimation for the coarse domains (25 km) of the Dry Zone and Dry Zone with outskirts domain designs for June to September. Generally, it becomes clear that the larger the domain one and the higher resolved the grid, the better the monthly precipitation captured by the model. The reason for this improvement could lie in the fact that with the large domain an additional large part west of the domain is downscaled by WRF and, therefore, the moisture transport into the domain is more accurate. It has previously been shown that the accuracy of precipitation does not only depend on capturing the correct large-scale atmospheric circulation, but also on the hydrological characteristics, which are very sensitive to the size of the domain [8]. With the large domain, we are able to capture larger-scale precipitation patterns, as also described by a previous study on sensitivity of domain size to precipitation [10]. This kind of precipitation is responsible for correctly capturing the climatology of the region, while we might miss some of the extreme precipitation [8, 10]. For the high-elevated station Hakha, the two domain designs Dry Zone and Dry Zone with outskirts show a clear overestimation in precipitation in the high resolution (5 km) domain from June to September, while the coarser domains perform much better. There is a general underestimation in October for all model output. These results might indicate that the small domain 2 results in an overactive hydrological cycle as it was seen for small domains in [8] and, therefore, the precipitation is overestimated during the rainy season.
The first domain in all three different domain designs (Fig 1A-C) shows an underestimation in the maximum temperature at Monywa, Hsipaw, Gangaw, Shwebo and Mandalay from June to September and also shows a clear overestimation in precipitation in these months. This is related to the fact that with more precipitation, more of the heating can be used for latent heating and less for sensible heating, which strongly reduces the maximum temperature.
Table S2 in S1 File indicates again that the MAE and RMSE are smallest for the second domain of the Bay of Bengal domain design, except at the weather station Hsipaw, but for this station the Bay of Bengal domain design shows the second best performance. Hence, the validation against weather station data clearly highlights that the Bay of Bengal domain performs best of all the different domain designs. Thus, the following analysis will focus only on the second domain of the Bay of Bengal domain design (Fig 1C).
Comparison to observational and gridded data
To understand how well WRF captures the annual cycle of precipitation, we compare monthly precipitation sums of the station data, WRF and MSWEP for the years 1981–2010, and IMERG for the period 2001–2023 (Fig 3 and S2 in S1 File) at the location of different stations listed in Table 2. Note that IMERG does not cover the same exact period as the other datasets due to limitations on its availability. However, a comparable long-enough period was employed in the comparison. Generally, IMERG shows much higher values than weather station data and MSWEP, which is related to the fact that MSWEP is a combined gauge, satellite and reanalysis product [27], so a closer representation to station data must be expected. Alternatively, the period that we are investigating in IMERG (2000–2023) could be wetter than our reference period (1981–2010).
The panels A-C show the monthly precipitation sum in millimeters for the years 1981–2010, for A: Dawei, B: Mandalay and C: Pathein. The grey boxes indicate the weather station data, whereas the grey shading shows the full range of the weather station data for each months for better comparison with the other datasets. The green boxes indicates the 5 km domain of the Bay of Bengal domain design, the red boxes indicate MSWEP data and the orange boxes represent IMERG data. The boxes represent the 25th to 75th percentile, while the whiskers extend to the value that is no more than 1.5 times the interquartile range away from the box. The comparison for additional weather stations can be found in Fig S2 in S1 File.
In the dry season, from December to April, all data sets show very little precipitation for all stations with some interannual variability, which is to be expected. WRF is too dry in April for most of the stations, except for Pathein, and also in October and November, except for Hakha. The underestimation of precipitation in these months might indicate a misrepresentation of the monsoon onset and cession over most of Myanmar (see the next section for more information on circulation). For all the other months WRF aligns very well with the station data and MSWEP. Mingaladon and Mrauk-U show a slight underestimation, while Pathein shows quite a substantial underestimation of precipitation during the whole wet season. Note that here we compare a climate simulation with real observations, so a slight deviation in the climatology can be expected. However, Pathein, located in the southwestern Ayeyarwady Delta, shows a clear underestimation and therefore, it might be an indication that the model is missing an important driver of precipitation in that region.
To better understand the precipitation patterns over Myanmar in our simulation, we further compare the present climate (1981–2010) to the Aphrodite data (1981–2007) and ERA5 (1981–2010). Fig 4 reveals that WRF compares well to Aphrodite and even better to ERA5. While Aphrodite seems to lack a bit the topographic structure of the country, particularly in the Chin Hills, this is well represented by WRF and ERA5. Also, comparing the weather station data to the gridded data, there is a good agreement, as shown in the previous section and Fig 3. It becomes obvious that the coastal areas, such as Mrauk-U, Pathein and Mingaladon are underestimated in terms of mean annual precipitation sums in WRF. Especially Pathein shows a strong underestimation for monthly precipitation data as shown in Fig 3. Based on Fig 4, the whole lower Ayeyarwady Delta region shows lower precipitation in WRF than in ERA5, Aphrodite and the station data. It seems to be a common problem in global and regional climate models that precipitation along the west coast of Myanmar is underestimated compared to observational datasets [3, 5–7]. This underestimation stems from the rainy season and, therefore, we also look into differences in circulation to identify potential misrepresentations of atmospheric circulation and moisture transport.
The climatological mean of the annual precipitation sums are displayed for A: WRF for the period 1981–2010, B: Aphrodite data set for the period 1981–2007, and C: for ERA5 for the period 1981–2010. Overlayed are the values for the 13 stations that are presented in Fig 1 and Table 2. The open-source Myanmar State and Regional Boundaries MIMU v9.4 dataset provided all the state and region boundaries and the Natural Earth data provides the countries’ outline.
The climatological mean of the atmospheric circulation is shown as the wind arrows and strength at 850 hPa (Fig 5). The flow direction of the wind during the rainy season matches very well the circulation represented in ERA5 over Myanmar. Here we representatively show July, but this is true for June to September, i.e., the rainy season. The exact location of the low pressure system is not equal, but the atmospheric flow in the lower atmosphere is consistent between ERA5 and WRF. Also the precipitable water is comparable between WRF and ERA5 from June to September. October is the month in which WRF underestimates precipitation in most of the stations, and here we see some differences between WRF and ERA5. The circulation in the lower free atmosphere is showing wind advection from inside the continent towards the Bay of Bengal in Myanmar in WRF, while there is still some transport from the sea into the Ayeyarwady Delta and the Dry Zone further north in ERA5. Hence, there is also a difference in the moisture availability between WRF and ERA5, which shows rather dry conditions throughout Myanmar in WRF, while there is much more moisture in the low-lands in ERA5. This is an explanation for the underestimation in precipitation at the end of the monsoon season in most of the stations shown in Fig 3 and S1 in S1 File. This mismatch in atmospheric circulation that leads to a dry bias was also found in another study [6]. The dry bias along the west coast of Myanmar seems to be very dominant in both global [3] and regional climate models [5–7]. The recent effort to obtain grid resolutions of around 10 km in global climate models might help to reduce this bias, as the whole atmosphere is resolved at high-resolution and could resolve some biases. Nevertheless, this bias seems to be persistent in such high-resolution global climate models [31]. Therefore, the interpretation of our results along the western coast of Myanmar must be done with caution.
The arrows provide the wind direction and strength of the wind at 850 hPa and the color shading indicates the monthly average precipitable water in A: and C: for WRF and in B: and D: for ERA5. In A: and B: the month July is presented and in C: and D: October is shown. The reference wind arrow is depicted below the panel. The wind arrows are shown for every 10th and 2nd grid point in both x- and y-direction for WRF and ERA5, respectively. The base layer of the map was created using the open-source vector map data from Natural Earth and the Natural Earth data provides the countries’ outline.
What is the effect of climate change?
Based on our previous evaluations, the WRF model has high skill in representing the climatology of the present climate. Therefore, we present projections for temperature and precipitation for the future under the two emission scenarios SSP2-4.5 and SSP5-8.5 and mid and end of century. The present climate simulation (Fig 6A and D) shows nicely the topography of the region, where the higher elevated regions such as the Chin Hills in the north western part of the domain and the Shan Hills in the north east of the domain. These higher elevated regions show up to 10 °C colder annual mean temperature than the dry lowland in the center (Dry Zone), where annual mean temperatures reach up to 26-27 °C.
The panels A and D show the annual mean 2-metre temperature for the period 1981–2010. Panels B-C and E-F show the difference between the future climate and the present climate. B: the difference with respect to SSP2-4.5 mid century, C: the difference with respect to SSP2-4.5 end of century, E: the difference with respect to SSP5-8.5 mid century, and F: the difference with respect to SSP5-8.5 end of century. The open-source Myanmar State and Regional Boundaries MIMU v9.4 dataset provided all the state and region boundaries and the Natural Earth data provides the countries’ outline.
The two mid-century simulations show an average increase in annual temperature of around 0.9 to 1.1 °C compared to the present climate, whereby the strongest increase in warming is projected for the northern part of the domain, while the smallest changes are seen in coastal regions. The maximum increase in temperature is projected for May in the two mid-century scenarios, where the model projects average temperature increases of 1.3 to 1.9 °C with strongest warming in the Sagaing state of up to 1.8 and 2.5 °C (S3B and Fig S3E in S1 File). These two simulations correspond to a global temperature increase of around 1 and 1.3 °C (see Table 1) compared to the climate in 1981–2010, and most closely represents a climate that is compatible with the Paris agreement as these show a warming of 1.7 to 1.9 °C with respect to the preindustrial climate (1850–1900) [32], respectively. The SSP2-4.5 end of century scenario corresponds to a warming of around 1.7 °C with respect to the present climate and 2.4 °C with respect to preindustrial climate. For this scenario, the model projects an average annual increase of around 1.6 °C, with the strongest warming in the northern part of the domain, both in the higher and lower lands and coldest regions along the coasts (Fig 6C). Also, for this scenario, the strongest increase in temperature is projected for May, with an average increase of around 2.2 °C (Fig S3C in S1 File). The highest increase is modeled for the Dry Zone in the center of the country with a warming up to 3.0 °C. The strongest increase in temperature is projected for the SSP5-8.5 end of century pathway, which corresponds to a 3 °C global warming with respect to the present climate (Table 1). This increase is also reflected in the annual mean temperature throughout all Myanmar in our domain, with an annual mean increase of 2.7 °C (Fig 6F). While for the other scenarios May is the most extreme month over Myanmar in our domain, this is not the case for the SSP5-8.5 end of century scenario. April is the month with the highest average increase in the entire domain with 3.3 °C and peak increases of up to 3.6 °C. The highest increase is projected for the whole Dry Zone (Fig S3F in S1 File).
Although for the change in mean temperature, May showed the strongest warming, the monthly maximum temperature projects the strongest increase in June. The two mid-century pathways project an average increase in monthly maximum temperature of 1.5 and 1.8 °C, with peak warming of 2.5 and 2.8 °C (S4B and Fig S4E in S1 File). The SSP2-4.5 end of century scenario shows a domain-wide increase of 2.3 °C and peak increases of 3.2 °C (Fig S4C in S1 File). For the SSP5-8.5 end of the century the model projects the highest average increase of 3.4 °C in May, but the strongest increase in warming is shown for the Tanintharyi Region (Fig S4F in S1 File) with up to 4.4 °C in June.
The projected warming in our simulations is consistent with previous results [3], which investigated 18 different CMIP6 models over Southeast Asia and found an increase of around 1.28 to 1.88 °C at mid century and 1.80 to 3.36 °C at the end of the century over the Myanmar region. These numbers are comparable to the 0.9 to around 2.7 °C warming for the mid century SSP2-4.5 to end-century SSP5-8.5 in our simulations. Note, that the MPI-EMS1-2-HR simulation shows a rather low climate sensitivity and thus, it is one of the models that project a comparable low warming for the future with respect to other global climate simulations [33].
Moving on to precipitation, the present climate (Fig 7A and D) reveals again the topography of the region, where the higher elevated regions such as the Chin Hills in the northwestern part of the domain and the Shan Hills in the northeast of the domain receive a lot more precipitation compared to the central Dry Zone of the country, which receives only around 600 to 1000 mm precipitation per year on average. This is due to the fact that a lot of lifting takes place along the slopes of the hills, leading to cloud formation and precipitation. This leaves only a limited amount of humidity for the Dry Zone. The Tanintharyi Region receives high amounts of annual precipitation sums, which is mainly related to the fact that it is located close to very warm water bodies that constantly supply the air with moisture.
The panels A and D show the annual precipitation sum for the period 1981–2010. Panels B-C and E-F show the difference between the future climate and the present climate. B: the difference with respect to SSP2-4.5 mid century, C: the difference with respect to SSP2-4.5 end of century, E: the difference with respect to SSP5-8.5 mid century, and F: the difference with respect to SSP5-8.5 end of century. The stippling indicates a significant change based on the non-parametric Mann-Whitney U test at a significance level of 5%. The open-source Myanmar State and Regional Boundaries MIMU v9.4 dataset provided all the state and region boundaries and the Natural Earth data provides the countries’ outline.
On an annual basis, there is only limited change in precipitation projected for the two mid-century simulations. There are some significant reductions in precipitation projected for the Shan Hills of around 10%. This is also true for the Tanintharyi Region where a decline of around 10 to 15% is projected. This reduction could therefore be as high as 500 to 600 mm per year (Fig 7B and E). A similar drying is also projected for the end of century under the SSP2-4.5 pathway (Fig 7C). Additionally, some wetting is projected in the central part of the Kayin state (east of Myanmar). For the 3 °C global warming projected by the end of the century for the SSP5-8.5 pathway, a considerable wetting is modeled for the whole northwestern part of Myanmar including the Dry Zone in the center with an increase of around 40 to 60%, while for the Shan Hills and the Tanintharyi Region there is still a reduction of up to 20% projected (Fig 7F). The increase in the Dry Zone is mainly related to a stronger moisture transport from the Bay of Bengal in the month June. There is a difference in projected change in the Ayeyarwady Delta around the mid-century for the two SSPs, where a significant reduction is projected for the SSP5-8.5 scenario and a slight non-significant increase is shown for the SSP2-4.5 around mid-century (Fig 7B and E). Given that the warming levels of the two mid-century simulations are very similar with 1.0 and 1.3 °C compared to present climate (see Table 1) this might also indicate that decadal variability can play an important role in steering the precipitation amounts in this region. Furthermore, given that the model tends to underestimate precipitation in this region compared to weather station and satellite data, this trend should be considered with care.
It must be noted that these changes on an annual basis can be very different for individual months, especially for the end of century under the SSP5-8.5 (Fig S5 in S1 File). There is a clear decline projected for precipitation in January, July, August, and November, while the increase mainly occurs during June, September, and October. Note that the wetting and the most intense overall warming occurs in June, which indicates high temperatures in combination with high levels of humidity in the future. This combination can potentially be very dangerous for humans, as an increase in the wet-bulb temperature can result in severe health issues [34].
A tendency towards a drying is projected for Myanmar for a 1 to 2 °C global warming compared to present-day climate, which then turns into a general wetting for a 3 °C global warming at the end of the 21st century under the SSP5-8.5 scenario. It must be noted that the Dry Zone seems to show a gradual increase in precipitation during the course of the 21st century and with stronger warming, which is non-significant for all periods and scenarios, except for the end of century under the SSP5-8.5 pathway. This is a new insight into climate change in Myanmar, as previous studies based on global and regional climate models generally project a gradual increase in precipitation with time and warming [3, 5, 7]. The higher level of detail in our precipitation projections for Myanmar is likely owed to the much higher resolution compared to previous studies performed over Southeast Asia.
Conclusion
Myanmar is one of the most vulnerable countries with respect to climate change and weather extremes, yet there are only very few studies that investigate global or regional climate model simulations. Understanding climate change is of uttermost importance to develop suitable adaptation strategies for the local population. High-resolution climate model simulations are needed, because the country is characterized by very heterogeneous precipitation patterns induced by a complex topography and strong influence of the warm waters in the Bay of Bengal and the Andaman Sea. Thus, we employed WRF to simulate a present climate (1981–2010) and future climate simulations for two scenarios – the middle of the road pathway SSP2-4.5 and the high-emission pathway SSP5-8.5 – each for the mid-century (2031–2060) and end of century (2071–2100). The simulations use initial and boundary conditions of one of the CMIP6 models MPI-ESM1-2-HR to dynamically downscale to 25 and 5 km spatial resolution and hourly temporal output.
We tested different domain settings to understand the sensitivity of the domain size to accuracy in 2m-temperature and precipitation. A small domain size for both the parent and the nest (Dry Zone domain design) does not allow to obtain an accurate climatology for temperature and precipitation for the region. This is due to the fact that precipitation is overestimated in large parts of the domain during the rainy season (June to October), particularly in the Dry Zone. This overestimation in precipitation also affects the temperature, as maximum temperatures are significantly underestimated due to the fact that too much energy is used for evaporation of the moisture and, therefore, it cannot be transformed into sensible heat. This bias improves in the nest with a bigger parent domain (Dry Zone with outskirts domain design), and almost vanishes for the Bay of Bengal domain design, which has both a very large parent domain and nest. Hence, the high resolved domain (5 km) of the Bay of Bengal design is used to further investigate the present and future climate simulations.
Comparisons against station and satellite based data indicate that the WRF simulations for the Bay of Bengal domain are very well able to capture the climatology of Myanmar. This is true except for a dry bias in the Ayeyarwady Delta, more precisely in Pathein, and an underestimation in monthly precipitation sums in October, which is related to a misrepresentation of the atmospheric circulation and therefore moisture advection to the region in October. A dry bias compared to observational data along the west coast of Myanmar is something most global and regional climate models present, no matter what their spatial resolution is.
For the future, the model projects an annual warming which is very similar to the global warming level of each of the simulations with respect to present-day climate, reaching from around 0.9 °C for mid century under the SSP2-4.5 to 2.7 °C at the end of the century for an SSP5-8.5. Mountainous terrain and the Dry Zone heat up comparably more than the coastal regions. Although most of the months warm in the range of the annual increase, May shows the strongest warming on average, which is on average around half a degree warmer than the annual value. For all the scenarios, except the SSP5-8.5 end of century, the maximum temperature shows the strongest increase in June overall of Myanmar, which is similar to the average warming of the mean temperature in May. For the SSP5-8.5 end of century the average increase in maximum temperature is in May, but the strongest increase is again in June with up to 4.4 °C of warming.
Annual precipitation changes show a significant drying in the Shan Hills of around 10% and the Tanintharyi Region of around 20% for both mid-century simulations and the end of century simulation under the SSP5-8.5. In the Dry Zone there is a tendency towards a wetting, albeit not statistically significant. The wetting becomes more pronounced and significant with an increase of 40 to 60% for the end of the century and the high-emission pathway SSP5-8.5. The drying in the Shan Hills and Tanintharyi Region remains. However, the changes for the annual precipitation sum can deviate for individual months. There is a clear decline projected for precipitation in January, July, August, and November, while the increase mainly occurs during June, September, and October. The combined effect of disproportional warming and increased precipitation especially in June could potentially affect human health in the region.
These simulations highlight the added value of high-resolution modeling for such a region. First, the appropriate domain, large enough to resemble the correct strength of the hydrological cycle, is able to capture correctly the climatology of temperature and precipitation of the region. Second, high-resolution is key to accurately represent the heterogeneous precipitation patterns and to project changes on a small-scale structure. While our model simulations show detailed spatial changes, other global and regional climate simulations project a rather uniform increase in both temperature and precipitation for Myanmar. In a next study we will present more extensively how climate change affects extreme weather and climate in the Bay of Bengal region. To further support our results, more convection-resolving simulations over the region would be required.
Supporting information
S1 File. Table S1.
Annual root mean squared error (RMSE) and mean absolute error (MAE) between weather station data and different domain designs for the period 1981–2010 for maximum two-meter temperature. Bold numbers indicate the lowest values for MAE and RMSE for the high-resolution 5 km domain.
Table S2. Annual root mean squared error (RMSE), mean absolute error (MAE), and temporal correlation between weather station data and different domain designs for the period 1981–2010 for precipitation. Bold numbers indicate the lowest values for MAE and RMSE for the high-resolution 5 km domain.
Fig S1. Comparison of different WRF designs with weather station data. The panels A-C show the monthly maximum temperature in degrees Celsius and panels D-F show the monthly precipitation sum in millimeters for the years 1981–2010. A and D: Gangaw, B and E: Shwebo, and C and F: Mandalay. The orange colors indicate the domain covering the Dry Zone (Fig 1A), the pink colors indicate the domains of the Dry Zone with outskirts (Fig 1B) and the teal colors indicate the domain covering the Bay of Bengal (Fig 1C). The bright colors indicate the coarser domain of 25 km resolution and the dark color represent the high resolution 5 km domain. The grey boxes indicate the weather station data, whereas the grey shading shows the full range of the weather station data for each months for better comparison with the other datasets. The boxes represent the 25th to 75th percentile, while the whiskers extend to the value that is no more than 1.5 times the interquartile range away from the box.
Fig S2. Comparison of Bay of Bengal WRF design with weather station, MSWEP and IMERG data. The grey boxes indicate the weather station data, whereas the grey shading shows the full range of the weather station data for each months for better comparison with the other datasets. The green boxes indicates the 5 km domain of the Bay of Bengal domain design, the red boxes indicate MSWEP data and the orange boxes represent IMERG data. The boxes represent the 25th to 75th percentile, while the whiskers extend to the value that is no more than 1.5 times the interquartile range away from the box.
Fig S3. Increase in monthly 2-metre average temperature in Myanmar for different scenarios, time periods and months. The panels A and D show the maximum 2-metre temperature in May and April, respectively, for the period 1981–2010. Panels B-C and E-F show the difference between the future climate and the present climate. B: the difference with respect to SSP2-4.5 mid century for May, C: the difference with respect to SSP2-4.5 end of century for May, E: the difference with respect to SSP5-8.5 mid century for May, and F: the difference with respect to SSP5-8.5 end of century for April. The open-source Myanmar State and Regional Boundaries MIMU v9.4 dataset provided all the state and region boundaries and Natural Earth data provides the countries’ outline.
Fig S4. Increase in monthly 2-metre maximum temperature in Myanmar for different scenarios and periods for June. The panels A and D show the maximum 2-metre temperature in June for the period 1981–2010.Panels B-C and E-F show the difference between the future climate and the present climate. B: the difference with respect to SSP2-4.5 mid century, C: the difference with respect to SSP2-4.5 end of century, E: the difference with respect to SSP5-8.5 mid century, and F: the difference with respect to SSP5-8.5 end of century. The open-source Myanmar State and Regional Boundaries MIMU v9.4 dataset provided all the state and region boundaries and the Natural Earth data provides the countries’ outline.
Fig S5. Projected change in monthly precipitation at different stations. The boxplots show the distribution of the 30 values at each month (x-axis) at different stations (compare Fig 1) for the present climate simulation (grey box), the mid- and end of century simulation for the SSP2-4.5 (light and dark blue boxes, respectively) and the mid- and end of century simulation for the SSP5-8.5 (orange and red boxes, respectively). The boxes represent the 25th to 75th percentile,while the whiskers extend to the value that is no more than 1.5 times the interquartile range away from the box.
https://doi.org/10.1371/journal.pclm.0000820.s001
(ZIP)
Acknowledgments
We thank Christoph C. Raible for providing access to computing resources and for providing storage of the data. We also acknowledge the Swiss National Supercomputing Center (CSCS) under the project s905, s1171, and s1248 for providing the computing hours necessary to run the WRF simulations used in this study.
References
- 1.
Adil L, Eckstein D, Künzel V, Schäfer L. Climate risk index 2025 : Who suffers most from extreme weather events?; 2025. Available from: https://www.germanwatch.org/sites/default/files/2025-02/Climate
- 2. Swe LMM, Shrestha RP, Ebbers T, Jourdain D. Farmers’ perception of and adaptation to climate-change impacts in the Dry Zone of Myanmar. Clim Dev. 2015;7(5):437–53.
- 3. Supharatid S, Nafung J, Aribarg T. Projected changes in temperature and precipitation over mainland Southeast Asia by CMIP6 models. J Water Clim Change. 2021;13(1):337–56.
- 4. Sein ZMM, Zhi X, Ogou FK, Nooni IK, Paing KH. Evaluation of coupled model intercomparison project phase 6 models in simulating precipitation and its possible relationship with sea surface temperature over Myanmar. Front Environ Sci. 2022;10.
- 5. Chotamonsak C, Salathé EP Jr, Kreasuwan J, Chantara S, Siriwitayakorn K. Projected climate change over Southeast Asia simulated using a WRF regional climate model. Atmos Sci Lett. 2011;12(2):213–9.
- 6. Ratna S, Ratnam J, Behera S, Tangang F, Yamagata T. Validation of the WRF regional climate model over the subregions of Southeast Asia: Climatology and interannual variability. Clim Res. 2017;71(3):263–80.
- 7. Tangang F, Chung JX, Juneng L, Supari , Salimun E, Ngai ST, et al. Projected future changes in rainfall in Southeast Asia based on CORDEX–SEA multi-model simulations. Clim Dyn. 2020;55(5–6):1247–67.
- 8. Bhaskaran B, Ramachandran A, Jones R, Moufouma-Okia W. Regional climate model applications on sub-regional scales over the Indian monsoon region: The role of domain size on downscaling uncertainty. J Geophys Res. 2012;117(D10).
- 9. Chu Q, Xu Z, Chen Y, Han D. Evaluation of the ability of the Weather Research and Forecasting model to reproduce a sub-daily extreme rainfall event in Beijing, China using different domain configurations and spin-up times. Hydrol Earth Syst Sci. 2018;22(6):3391–407.
- 10. Du S, Zhuo L, Kendon EJ, Han D. Exploring suitable domain size for high-resolution urban rainfall simulation. Urban Clim. 2025;61:102489.
- 11. Ban N, Schmidli J, Schär C. Evaluation of the convection-resolving regional climate modeling approach in decade-long simulations. J Geophys Res Atmos. 2014;119(13):7889–907.
- 12. Giorgi F, Torma C, Coppola E, Ban N, Schär C, Somot S. Enhanced summer convective rainfall at Alpine high elevations in response to climate warming. Nature Geosci. 2016;9(8):584–9.
- 13. Kendon EJ, Stratton RA, Tucker S, Marsham JH, Berthou S, Rowell DP, et al. Enhanced future changes in wet and dry extremes over Africa at convection-permitting scale. Nat Commun. 2019;10(1):1794. pmid:31015416
- 14. Messmer M, González-Rojí SJ, Raible CC, Stocker TF. Sensitivity of precipitation and temperature over the Mount Kenya area to physics parameterization options in a high-resolution model simulation performed with WRFV3.8.1. Geosci Model Dev. 2021;14(5):2691–711.
- 15. González-Rojí SJ, Messmer M, Raible CC, Stocker TF. Sensitivity of precipitation in the highlands and lowlands of Peru to physics parameterization options in WRFV3.8.1. Geosci Model Dev. 2022;15(7):2859–79.
- 16.
Skamarock WC, Klemp JB, Dudhia J, Gill DO, Barker DM, Duda MG, et al. A description of the advanced research WRF Version 3. NCAR Technical Note NCAR/TN-475+STR. 2008;https://doi.org/10.5065/D68S4MVH
- 17. von Storch JS, Putrasahan D, Lohmann K, Gutjahr O, Jungclaus J, Bittner M, et al. MPI-M MPIESM1.2-HR model output prepared for CMIP6 HighResMIP; 2017. Available from.
- 18. Gutjahr O, Putrasahan D, Lohmann K, Jungclaus JH, von Storch J-S, Brüggemann N, et al. Max Planck institute earth system model (MPI-ESM1.2) for the high-resolution model intercomparison project (HighResMIP). Geosci Model Dev. 2019;12(7):3241–81.
- 19. Grell GA, Freitas SR. A scale and aerosol aware stochastic convective parameterization for weather and air quality modeling. Atmos Chem Phys. 2014;14(10):5233–50.
- 20. Hong SY, Jade Lim JO. The WRF single-moment 6-class microphysics scheme (WSM6). Asia-Pacific J Atmos Sci. 2006;42(2):129–51.
- 21. Mlawer EJ, Taubman SJ, Brown PD, Iacono MJ, Clough SA. Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J Geophys Res. 1997;102(D14):16663–82.
- 22. Dudhia J. Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model. J Atmos Sci. 1989;46(20):3077–107.
- 23. Hong S-Y, Noh Y, Dudhia J. A new vertical diffusion package with an explicit treatment of entrainment processes. Month Weather Rev. 2006;134(9):2318–41.
- 24. Niu G-Y, Yang Z-L, Mitchell KE, Chen F, Ek MB, Barlage M, et al. The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. J Geophys Res. 2011;116(D12).
- 25. Riahi K, van Vuuren DP, Kriegler E, Edmonds J, O’Neill BC, Fujimori S, et al. The shared socioeconomic pathways and their energy, land use, and greenhouse gas emissions implications: An overview. Global Environ Change. 2017;42:153–68.
- 26.
Huffman GJ, Bolvin DT, Braithwaite D, Hsu K-L, Joyce RJ, Kidd C, et al. Integrated multi-satellite retrievals for the global precipitation measurement (GPM) mission (IMERG). Advances in Global Change Research. Springer International Publishing; 2020. p. 343–53. https://doi.org/10.1007/978-3-030-24568-9_19
- 27. Beck HE, Wood EF, Pan M, Fisher CK, Miralles DG, DijkAIJMv , et al. MSWEP V2 Global 3-hourly 0.1° precipitation: Methodology and quantitative assessment. Bull Am Meteorol Soc. 2019;100(3):473–500.
- 28. Yatagai A, Kamiguchi K, Arakawa O, Hamada A, Yasutomi N, Kitoh A. APHRODITE: Constructing a long-term daily gridded precipitation dataset for Asia based on a dense network of rain gauges. Bull Am Meteorol Soc. 2012;93(9):1401–15.
- 29.
Maeda M, Yasutomi N, Yatagai A, for Atmospheric Research Staff (Eds) NC. The climate data guide: APHRODITE: Asian precipitation – Highly-resolved observational data integration towards evaluation of water resources. 2023. https://climatedataguide.ucar.edu/climate-data/aphrodite-asian-precipitation-highly-resolved-observational-data-integration-towards
- 30. Hersbach H, Bell B, Berrisford P, Hirahara S, Horányi A, Muñoz-Sabater J, et al. The ERA5 global reanalysis. Quart J Royal Meteoro Soc. 2020;146(730):1999–2049.
- 31. Moon J-Y, Streffing J, Lee S-S, Semmler T, Andrés-Martínez M, Chen J, et al. Earth’s future climate and its variability simulated at 9 km global resolution. Earth Syst Dynam. 2025;16(4):1103–34.
- 32.
United Nations Framework Convention on Climate Change (UNFCCC). Paris agreement; 2015. Adopted at the twenty-first session of the conference of the parties (COP 21), Paris; 12 December 2015. Available from: https://unfccc.int/sites/default/files/english_paris_agreement.pdf
- 33.
Gutiérrez JM, Jones RG, Narisma GT, Alves LM, Amjad M, Gorodetskaya IV, et al. Atlas. In: Climate change 2021 : The physical science basis. Contribution of working group I to the sixth assessment report of the intergovernmental panel on climate change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L.Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K.Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press; 2021. Available from: https://interactive-atlas.ipcc.ch/
- 34. Raymond C, Matthews T, Horton RM. The emergence of heat and humidity too severe for human tolerance. Sci Adv. 2020;6(19):eaaw1838. pmid:32494693