Interannual Rainfall Variability Research Articles

Routine Climate Monitoring in the State of Hawai‘i: Establishment of State Climate Divisions

Abstract The Hawaiian Islands have some of the most spatially diverse rainfall patterns on earth, affected by prevailing trade winds, midlatitude disturbances, tropical cyclones, and complex island topography. However, it is the only state in the U.S. that does not have assigned climate divisions (boundaries defining climatically homogeneous areas), which excludes it from many national climate analyses. This study establishes, for the first time, official climate divisions for the State of Hawai‘i using cluster analysis applied to monthly gridded rainfall data from 1990 to 2019. Twelve climate divisions have been identified: two divisions were found each for the islands of Kaua‘i (Leeward Kaua‘i and Windward Kaua‘i), O‘ahu (Waianae and Ko‘olau), and Maui County (Leeward Maui Nui and Windward Maui Nui), and six divisions were identified for Hawai‘i Island (Leeward Kohala, Windward Kohala, Kona, Hawai‘i Mauka, Ka‘u, and Hilo). The climate divisions were validated using a statewide area-weighted division-average rainfall index which successfully captured the annual cycle and interannual rainfall variations in the statewide average rainfall series. Distinct rainfall seasonality features and interannual/decadal variability are found among the different divisions; Leeward Maui Nui, Leeward Kaua‘i, Kona, and Hawai‘i Mauka displayed the most significant rainfall seasonality. The western Hawai‘i Island divisions show the most significant long-term decreasing trends in annual rainfall during the past 100 years (ranging from -2.5% to -5.0% per decade). With these climate divisions now available, Hawai‘i will have access to numerous operational climate analyses that will greatly improve climatic research, monitoring, education and outreach, as well as forecasting applications.

Trend analysis and change-point detection of monsoon rainfall in Uttarakhand and its impact on vegetation productivity

This study analyzes the long-term spatio-temporal changes and trend analysis in rainfall using the data from 1901 to 2020 and its impact on vegetation from 2000 to 2020 across districts of Uttarakhand. The Pettitt test was employed to detect the abrupt change point in time frame, while the Mann-Kendall (MK) test was performed to analyze the rainfall trend. Results show that the most of the districts exhibited significant negative trend of rainfall in monsoon, except two districts. Out of 13 districts, 4 districts recorded noteworthy rainfall declining trend for the monsoon season at 0.05% significance level, while the insignificant negative trend of rainfall was detected for 7 districts of Uttarakhand. Furthermore, the significant negative trend (-2.23) was recorded for overall monsoon rainfall of Uttarakhand. Based on the findings of change detection, the most probable year of change detection was occurred primarily after 1960 for most of the districts of Uttarakhand. A significant decline rainfall was detected after 1960 while after 1970 interannual variability of rainfall was recorded to be increased.  The analysis of month wise cumulative gross primary productivity (GPP) for 13 districts with rainfall trends shows that there is significant impact of rainfall trend on GPP during month of June and it gradually reduces for subsequent monsoon months. It was observed that the GPP of region is increasing at rate of 9.1 gCm-2d-1 in the region since 2000. Based on sensitivity analysis, the GPP of cropped area of region is more sensitive towards rainfall than forest area of Uttarakhand.

Correction: The Southern Annular Mode: its influence on interannual variability of rainfall in North Australia

Correction: The Southern Annular Mode: its influence on interannual variability of rainfall in North Australia

Monsoon Mission Coupled Forecast System version 2.0: model description and Indian monsoon simulations

Abstract. We present the Monsoon Mission Coupled Forecast System version 2.0 (MMCFSv2) model, which substantially upgrades the present operational MMCFSv1 (version 1) at the India Meteorology Department. The latest 25 years (1998–2022) of retrospective seasonal coupled hindcast simulations of the Indian summer monsoon with April initial conditions from Coupled Forecast System Reanalysis are discussed. MMCFSv2 simulates the tropical wind, rainfall, and temperature structure reasonably well. MMCFSv2 captures surface winds well and reduces precipitation biases over land, except over India and North America. The dry bias over these regions remained like in MMCFSv1. MMCFSv2 captures significant features of the Indian monsoon, including the intensity and location of the maximum precipitation centers and the large-scale monsoon circulation. MMCFSv2 improves the phase skill (anomaly correlation coefficient) of the interannual variation of Indian summer monsoon rainfall (ISMR) by 17 % and enhances the amplitude skill (normalized root mean square error) by 20 %. MMCFSv2 shows improved teleconnections of ISMR with the equatorial Indian and Pacific oceans. This 25-year hindcast dataset will serve as the baseline for future sensitivity studies of MMCFSv2.

Simulated Tree‐Grass Competition in Drylands Is Modulated by CO2 Fertilization

AbstractInterannual variability in climatic drivers can have a strong impact on dryland ecosystem functioning globally. While interannual variations in dryland ecosystem processes are mainly driven by rainfall, other global change drivers such as CO2 fertilization and rising temperatures can play an increasingly important role for these ecosystems. Yet, the high complexity of dryland ecosystems makes it difficult to unravel the individual and interactive impacts of these different drivers. In this work we study the impacts of interannual climatic variability on the dryland ecosystems of the Sudano‐Sahel region for the period 1950–2022. By using a dynamic vegetation model (LPJ‐GUESS v4.0), we show that the year‐to‐year variability in dryland ecosystems that originates from interannual variability in rainfall is modulated by effects of CO2 fertilization, which can strongly impact woody encroachment and resource competition between vegetation types. In contrast, we found that increased temperatures had a relatively low influence at the ecosystem level. We also show that this response varies with aridity subtype, depending on the amount and type of woody cover. By untangling the impacts of climatic drivers on dryland vegetation, this study helps us to understand the different sensitivities of dryland ecosystems to climatic variability under global change.

The Southern Annular Mode: its influence on interannual variability of rainfall in North Australia

The Southern Annular Mode: its influence on interannual variability of rainfall in North Australia

Importance of land use and rainfall in the dynamics of the Caatinga vegetation cover in Northeastern Brazil

Many factors, such as climate, land use, and cultural and social aspects, influence the vegetation cover dynamics. We aimed to analyze the vegetation cover dynamics in three municipalities of the Brazilian semi-arid region (Aroeiras, Boqueirão, and Congo, state of Paraíba, Brazil) and to check for the influence of land use and inter-annual variations in rainfall. We selected 18 Landsat satellite images between 1995 and 2017. We performed supervised classification based on pre-established regions of interest: Closed or conserved vegetation, Sparse or disturbed vegetation, Exposed soil, Water bodies, clouds, and shadows, according to color, shape, and texture criteria. In addition, we collected data from the IBGE's Agricultural Census and annual rainfall. Variations in land use and rainfall variations worked synergistically on vegetation cover dynamics. In the municipality of Congo, very wet years were associated with higher percentages of closed vegetation and decreased soil exposure. High percentages of land use were related to a higher cover of sparse vegetation and exposed soil. In Boqueirão, we observed the opposite phenomenon, with drier years associated with higher percentages of closed vegetation and exposed soil. In Aroeiras, we identified no clear relationship between rainfall and variations in vegetation cover. Similar to that observed in the other municipalities, the abandonment of agricultural and cattle ranching activities increased dense vegetation. Our results show that the analysis of land use and vegetation cover dynamics in the semi-arid region represents a complex task ranging from regional factors, such as climate, characterized by inter-annual rainfall variations, to local factors, such as land use intensity and demographic/social aspects.

Assessing long-term trends in vegetation cover change in the Xilin River Basin: Potential for monitoring grassland degradation and restoration

Under the influence of climate change and human activities, the problem of grassland degradation is becoming increasingly severe. Detection of changes in vegetation cover is crucial for a better understanding of the interaction between humans and ecosystems. This study maps changes in vegetation cover using the Google Earth Engine (GEE). We used 36 years of Landsat satellite imagery (1985–2020) in the Xilin River Basin, China, to classify grassland conditions and validated the results with field observation data. The overall classification of the model accuracy assessment was 83.3%. The Dynamic Reference Vegetation Cover Method (DRCM) was adopted to remove the effect of interannual variation of rainfall, allowing to focus on the impact of human activities on vegetation cover changes. The results identify five categories of vegetation cover changes: significantly increased, potentially increased, stable, potentially decreased, and significantly decreased. The reference level is derived from the most persistent land surface coverage across different grassland types and all years. Overall, 9.3% of the study area had a significant increase in vegetation cover, 14.2% a potential increase, 48.6% of the area showed a stable vegetation condition, 9.8% showed a potential decrease, and 18.1% a significant decrease in vegetation cover. The largest proportion of combined potential and significant reduction was 35.2% for desert grassland, where the vegetation faced the most severe reduction. This study will provide a basis for identifying grassland degradation and developing scientific management policies.

Vegetation response to grazing and drought (13 yr) in a conservation area in the Succulent Karoo, South Africa

Rangelands of arid ecosystems are driven by internal and external environmental controls. Grazing pressure is an important anthropogenic driver, but stochastic environmental events such as rainfall variability and prolonged drought can have profound effects on arid vegetation. We investigated the effect of a prolonged drought and initial high grazing pressure on range condition, perennial vegetation cover, life form cover, vegetation composition and vegetation diversity at five monitoring sites in a conservation area in arid Succulent Karoo vegetation. The response of the rocky mountainous vegetation to drought and grazing differed notably from the response of the sandy plains vegetation. Range condition and perennial vegetation cover reduced drastically in all habitats, although the extent of deterioration was less in the mountainous habitats. Annual vegetation cover was associated with inter-annual rainfall variability. Species richness decreased at three of five sites, whereas species evenness increased at three of five sites. Shannon-Wiener index of diversity and Simpson's diversity index trends were inconsistent. Vegetation composition changed directionally from a quasi-stable state in all five monitoring sites as the drought persisted. The extent of the change was less in mountainous habitats indicating potentially higher resilience than in plains habitats, though grazing pressure was lower in the mountains than on the plains over the study period. The existence of quasi-stable states, together with the observed directional changes in vegetation composition in response to both drought and grazing, indicated that both equilibrium and non-equilibrium vegetation dynamics apply.

Large-Scale Stability and the Greater Horn of Africa Long and Short Rains

Abstract The societies of the coastal regions of the Greater Horn of Africa (GHA) experience two distinct rainy seasons: the generally wetter “long” rains in the boreal spring and the generally drier “short” rains in the boreal fall. The GHA rainfall climatology is unique for its latitude in both its aridity and for the dynamical differences between its two rainy seasons. This study explains the drivers of the rainy seasons through the climatology of moist static stability, estimated as the difference between surface moist static energy hs and midtropospheric saturation moist static energy . In areas and at times when this difference, , is higher, rainfall is more frequent and more intense. However, even during the rainy seasons, on average and the atmosphere remains largely stable, in line with the GHA’s aridity. The seasonal cycle of , to which the unique seasonal cycles of surface humidity, surface temperature, and midtropospheric temperature all contribute, helps explain the double-peaked nature of the regional hydroclimate. Despite tropospheric temperature being relatively uniform in the tropics, even small changes in can have substantial impacts on instability; for example, during the short rains, the annual minimum in GHA lowers the threshold for convection and allows for instability despite surface humidity anomalies being relatively weak. This framework can help identify the drivers of interannual variability in GHA mean rainfall or diagnose the origin of biases in climate model simulations of the regional climate.

Enhanced Soil Moisture Retrieval through Integrating Satellite Data with Pedotransfer Functions in a Complex Landscape of Ethiopia

Remotely sensed soil moisture products potentially provide a valuable resource for monitoring agricultural drought and assessing food security. The agriculture dominated countries of Eastern Africa experience high inter-annual variability of rainfall, but the monitoring and assessment of the predominantly rainfed agriculture systems is hindered by an absence of ground-based observations. This study evaluates the accuracy of three soil moisture products: ASCAT SWI 12.5 km, SMAP soil moisture data 9 km (SPL3SMP_E), and enhanced surface soil moisture map derived through integrating ASCAT SWI and Pedotransfer Functions (PTFs) (ASCAT_PTF_SM), in Ethiopia, through comparison with in situ-observed soil moisture datasets. Additionally, a new water retention PTF, developed for Ethiopian soils, is integrated with a high-resolution soil property dataset to enhance the spatial resolution of the soil moisture product. The results show that the new integrated dataset performs better in terms of unbiased root mean square error (ubRMSE = 0.0398 m3/m3) and bias (0.0222 m3/m3) in comparison with ASCAT SWI 12.5 km (ubRMSE = 0.0.0771 m3/m3, bias = 0.1065 m3/m3). SMAP is found to have limitations during the wet season, overestimating soil moisture. The finer spatial resolution of the data allows for a better depiction of heterogeneity of soil moisture across the landscape and can be used to identify water-related issues and improve hydrological models for agricultural water management.

A 900-Year Isotopic Proxy Rainfall Record from Northeastern Botswana

A high-resolution climate archive was reconstructed based on carbon isotope analysis and radiocarbon dating of the Chapman baobab in northeastern Botswana. The Chapman baobab, which exhibited an open ring-shaped structure composed of six stems, collapsed in January 2016 during an intense El Niño event. Two samples belonging to the oldest stems were investigated in order to obtain a proxy rainfall record, which provides insight into the precipitation regime over the last millennium, evincing centennial and decadal scale variability. The results indicate that the Medieval Warm Period was marked by relatively stable precipitation, whereas rainfall variability and drought frequency increased during the Little Ice Age. The investigated area has experienced both wetter and drier conditions in the past. The wettest conditions of the last millennium were registered before 1450 while the driest period occurred in 1835. For southern Africa, inter-annual rainfall variability is mainly associated with sea surface temperatures in the Agulhas Current core region, which determine the east–west displacement of tropical temperate troughs. Previous studies suggested that positive sea surface temperature anomalies in the Mozambique Channel led to an eastward movement of the troughs but the Chapman record demonstrates a westward displacement in the past, causing drought in northeastern South Africa and wetter conditions in the central part of southern Africa. The positive rainfall correlation with SST anomalies reversed after 1900, causing a gradual decrease in precipitation and confirming the current aridity trend for Botswana. The results contribute to a better understanding of the past climate of southern Africa for which paleoclimate reconstructions remain scarce.

Accurate statistical seasonal streamflow forecasts developed by incorporating remote sensing soil moisture and terrestrial water storage anomaly information

Water managers need improved seasonal streamflow forecasts in drought-prone regions like the U.S. Great Plains, where the inter-annual variability in rainfall and streamflow is high and seasonal streamflow forecasting systems are not well developed. Recent research has demonstrated that in-situ soil moisture measurements can be used to produce accurate streamflow forecasts in the Great Plains where rainfall runoff is the predominant source of streamflow. However, an inadequate density of in-situ soil moisture monitoring locations makes producing such forecasts impossible in many locations. Therefore, our objective was to develop and quantify the accuracy of seasonal streamflow forecasts for rainfall-dominated watersheds using remotely sensed data. Precipitation and remotely sensed soil moisture and terrestrial water storage anomaly data were incorporated into a multiple linear regression (MLR) model to produce seasonal streamflow forecasts for five watersheds in the Great Plains. The best MLR models had Nash-Sutcliffe Efficiency (NSE) values ranging from 0.71 to 0.93, indicating good to very good model performance. In one watershed with nearby in-situ soil moisture data, the use of remotely sensed soil moisture data led to greater forecast accuracy than did the inclusion of the sparse in-situ data. Our results show that the use of remote sensing data in MLR models can provide accurate seasonal streamflow forecasts in rainfall-dominated regions which currently lack such information.

Spatiotemporal Variability of Rainfall as an Evidence of Changing Climate over the Nigerian Niger Delta

This study assesses the spatial and temporal variability of rainfall over the Niger Delta area (Benin, Calabar, Port Harcourt and Uyo) in Nigeria between 1990 and 2020. The daily rainfall data used for this study were obtained through secondary source from NASA’s POWER (Prediction of Worldwide Energy Resources). Coefficient of variation, standardized anomaly index (SAI), precipitation concentration index (PCI) and seasonality index (SI) were used to evaluate rainfall variability and seasonality. The results of temporal analysis of annual rainfall indicated that Benin had the highest annual rainfall amount in 2000 and lowest in 1999 while Calabar and Port Harcourt had their highest rainfall amount in 2007 with the lowest occurring in 1991 and 2014 respectively. Uyo had the highest in 2000 and lowest in 2014. SAI also witnessed the presence of interannual variability of rainfall with negative and positive anomalies in 59.46% and 40.54% of the analyzed years, respectively. PCI and SI results implied that the Niger Delta area had strong irregular rainfall distribution. The results confirm the meteorological sensitivity to locations and provide support for the implementation of station and regional meteorological hazards prevention and to support the mitigation of hazards and improve the management practices of environment system in Niger Delta area.

Effects of climate variability changes on runoff and erosion in the Western European Loess Belt region (NW, France)

Runoff and soil erosion are very pronounced in the Western European Loess Belt. In this study, the distributed physically-based model CLiDE is calibrated, validated, and applied to a catchment of this area (Dun, NW, France) to assess the hydro-sedimentary impacts of climate change scenarios. Despite considerable progress over the last decade in the study of runoff and soil erosion in the context of climate change, the effects of changes in the temporal variability of precipitation remain poorly understood, especially at the scale of a river basin. To examine these relationships more closely, we developed a stochastic weather generator to individually adjust the components that structure the temporal variability of rainfall. The climate scenarios considered represent projections to the year 2100 of the temporal variability of rainfall over NW Europe. The scenarios are based on historical daily rainfall records (1990–2012) and 4 exploratory assumptions: a 50 % decrease in the interannual rainfall regime (scenario 6yD), a 100 % increase in the interannual rainfall regime (scenario 6yI), a 50 % increase in the seasonal rainfall regime (scenario 1yI) and a 50 % increase in the synoptic rainfall regime (scenario 3dI). Simulated daily water and sediment discharges and erosion/deposition maps for each scenario are compared to those simulated for the situation without changes in rainfall. The time series were aggregated over different time intervals to allow for a multi-scale analysis of the differences. The results indicate that the model provides a satisfactory prediction of the catchment's water and sediment discharges, especially over the calibration period. Increased climate variability, whether on a synoptic (3dI), seasonal (1yI) or interannual (6yI) scale, leads to increased runoff and erosion. Increasing the synoptic rainfall variability (3dI) leads to the largest increase in mean annual runoff and erosion. Only the reduction of the interannual rainfall variability (6yD) provokes the decrease of these values.

Factors affecting interannual variation in late summer rainfall in the Red River Delta of Vietnam

Based on station monthly observational statistics, the major rainfall in the Red River Delta (RRD) of Vietnam occurs in late summer (July–September) with conspicuous year-to-year variation. Using the ± 0.8 of the seasonal total rainfall standard deviation as criteria, seven wet and six dry years are identified over the period 1983–2015. In addition to the 70.5% of the seasonal total rainfall contributed by the heavy rainfall days, the distinct heavy rainfall accumulation difference between wet and dry years seems to fundamentally establish these two separated extreme wet and dry groups. As revealed from further analyses, the large variability in rainfall is attributed to the influence of tropical cyclones (TCs) and 7–24- and 30–60-day intraseasonal oscillations (ISOs); in particular, the number of TCs affecting the RRD and rainfall produced by TCs are more (less) during the wet (dry) years, and the amplitudes of ISOs are also enhanced (reduced). In many cases, heavy rainfall days are induced by the combined effect of both ISOs and TCs, while some heavy rainfall events are mainly triggered by ISOs. It is found from the water vapor budget analyses that an anomalous cyclone (anticyclone) dominates over the Indochina Peninsula in wet (dry) years, resulting in more (less) water vapor being transported to the RRD, whereas the anomalous convergence (divergence) of water vapor flux leads to the maintenance of excessive (insufficient) rainfall over the RRD. However, the El Niño–Southern Oscillation (ENSO) forcing shows minor effects on the interannual variation in rainfall in the RRD.

Towards improved seasonal rainfall prediction in the tropical Pacific Islands

The El Niño Southern Oscillation (ENSO) is a major influence on interannual variability of rainfall in stations in the tropical southwest Pacific. Predictions of seasonal rainfall, especially a season or two ahead, are of great value to these countries. This paper therefore examines the correlations over ~ 60 years between seasonal rainfall and 8 ENSO indicators at 16 island stations, allowing for lead times. The results show the influence on rainfall of the position and movement of the South Pacific Convergence Zone (SPCZ) during ENSO events, and that the Southern Oscillation Index (SOI), the sea surface temperature anomaly in the central Pacific (Niño 3.4), and the warm water volume in the eastern Pacific (WWV1) have longer lead times compared to most other ENSO indicators. These indicators can therefore be used with confidence in SCOPIC, a widely used statistical tool for prediction of seasonal rainfall. (As global climate models generally have systematic errors in their depiction of the SPCZ, they cannot yet be used directly to reliably predict seasonal rainfall in this region.) For several sites near the SPCZ, we find that a moderately good forecast of rainfall in both spring and summer can be made from indicators measured in June–July (i.e., 3–6 months in advance.)

Analysis of Observed and Perceived Climate Change and Variability in Gondar Zuria Woreda, Northwestern Ethiopia

Examining the spatiotemporal variability and trends of climate in the context of global climate changes particularly in countries like ours where rainfed agriculture is predominant is indispensable to suggest possible adaptation policies. Therefore, this study aimed at assessing the changes and variability of rainfall and temperature in Gonder Zuria Woreda, Northwestern Ethiopia. The study was conducted using quantitative research design. Climatic data which were collected from 1952 – 2016 were the data source of this study. Questionnaire was used to collect data from 100 local people about their perception on climate changes, and it was presented in frequency and percentage distributions. Coefficient of variation, anomaly index, Precipitation Concentration Index and Seasonality Index were used to analyze the variability of rainfall on annual and seasonal basis whereas Inverse distance weighting interpolation was used to show the spatial variability of rainfall. Furthermore, both non- Mann-Kendall tests and linear regression were used to detect the trends in time-series climatic data for the period of 1952-2016. The result revealed that there is intra-annual/seasonal and inter-annual variability of rainfall. Annual, summer and spring rainfall has decreased with a rate of 21.24, 7.2 and 1.85 mm per decade respectively. The minimum average and maximum rate of change of temperature was found to be 0.19, 0.25 and 0.3oC per decade respectively. The Mann-Kendall trend test result indicated that there is significant increasing trend for minimum, average and maximum temperatures. Most of the respondents (80%), perceived that temperature is increasing and rainfall is declining. Therefore, a sustainable climate risk management approach is recommended to adapt to the ongoing impacts ofclimate variability and climate change.

Nexus between summer climate variability and household food security in rural Mpumalanga Province, South Africa

Ongoing climate changes are likely to impact household food security in rural households that depend on rainfed subsistence agriculture. This paper investigates the relationship between summer climate variability and household food security in rural Mpumalanga, South Africa. We used a household panel data set nested in the Agincourt Health and Socio-Demographic Surveillance System, together with rainfall and temperature data for the summer periods 2006-07 to 2018-19 from three weather stations that surround the study area. We quantified the variability of rainfall using coefficient of variation and the standardised rainfall anomaly index, while temperature variability was reflected by the standardised temperature anomaly. In addition, the Mann-Kendall analysis was applied to detect temporal trends in rainfall and temperature. Longitudinal models accounting for socioeconomic and climate factors were used to estimate the relationship between weather and climate. The results reveal significant impact on food security from high inter-annual rainfall variability through fluctuations in food consumption, dietary diversity, and the experience of hunger. This study offers significant insights on how dietary diversity, food availability and overall food security are positively associated with greater average rainfall through subsistence agriculture as a livelihood strategy. These insights have important implications by suggesting seasonal forecasts to predict periods of potential food insecurity in local communities and can guide government policy and interventions to lessen food insecurity in rural areas.

Comparison of Bias Correction Methods for Summertime Daily Rainfall in South Korea Using Quantile Mapping and Machine Learning Model

This study compares the bias correction techniques of empirical quantile mapping (QM) and the Long Short-Term Memory (LSTM) machine learning model for summertime daily rainfall simulation focusing on precipitation-dependent bias and temporal variation. Numerical experiments using Weather Research and Forecasting (WRF) were conducted over South Korea with lateral boundary conditions of ERA5 reanalysis data. For the spatial distribution of mean summertime rainfall, the bias-uncorrected WRF simulation (WRF_RAW) showed dry bias for most of the region of South Korea. The WRF results corrected by QM and LSTM (WRF_QM and WRF_LSTM, respectively) were improved for the mean summer rainfall simulation with the root mean square error values of 0.17 and 0.69, respectively, which were smaller than those of the WRF_RAW (1.10). Although the WRF_QM performed better than the WRF_LSTM in terms of the summertime mean and monthly precipitation, the WRF_LSTM presented a closer interannual rainfall variation to the observation than the WRF_QM. The coefficient of determination for calendar-day mean rainfall was the highest in the following order: the WRF_LSTM (0.451), WRF_QM (0.230), and WRF_RAW (0.201). However, the WRF_LSTM had a limitation in reproducing extreme rainfall exceeding 50 mm/day due to the few cases of extreme precipitation in training data. Nevertheless, the WRF_LSTM better simulated the observed light-to-moderate precipitation (10–50 mm/day) than the others.