A 1961–2024 daily hydrometeorological dataset for the Yarlung Zangbo basin of the southern Tibetan Plateau from high-resolution observation and model integration Authors: He Sun, Tandong Yao, Fengge Su, Bowen Zheng, Yi Zhang Published: 2026 ABSTRACT Understanding runoff changes in the Yarlung Zangbo (YZ) basin is crucial for water resource management on the Tibetan Plateau (TP), but has been hindered by sparse data. This study provides a comprehensive hydrometeorological dataset for the YZ basin spanning 1961–2024, including both daily meteorological variables and hydrological components. The dataset integrates observations from 26 meteorological stations, 15 hydrological stations, and hydrological model simulations. Specifically, it includes: (1) daily meteorological variables (precipitation, mean, maximum, and minimum temperature, and wind speed) at both 26 meteorological stations and a 10-km gridded scale; and (2) daily total runoff and its components (snowmelt, glacier runoff, and rainfall runoff) at the 10-km gridded scale, derived from a physically based hydrological model. The meteorological fields were extended using a random forest–based machine learning approach combined with ERA5 reanalysis data, while the hydrological variables were generated from physically based hydrological model simulations. To facilitate direct application of hydrological station data, the gridded daily hydrological data for 1961–2024 were further routed to 15 hydrological stations. The generated dataset was validated against in situ meteorological at different scales and hydrological observations and compared with other meteorological products, demonstrating good applicability in the YZ. This dataset is available at the National Tibetan Plateau Data Center. The dataset enables basin-wide simulations of hydrological processes across multiple scales and provides a valuable resource for assessing water cycle dynamics and their responses to climate and cryospheric changes in the YZ basin. Background & Summary Hydrological processes in high-mountain regions are significantly influenced by climatic and cryospheric changes. The Tibetan Plateau, referred to as the Asian Water Tower, provides freshwater to approximately 2 billion people1,2. The high elevation, complex terrain, severe weather, and inaccessibility of the TP hinder direct meteorological observations, restricting insights into climate and hydrology interactions. Situated in the southern Tibetan Plateau (TP), the Yarlung Zangbo (YZ) river basin is the world’s highest river, averaging over 4600 meters in elevation, and is the TP’s largest river basin, covering approximately 250,000 km2. It serves as the primary freshwater source for the Tibet Autonomous Region and acts as a conduit for moisture transport from the Indian Ocean to the plateau’s interior. It is also a significant transboundary river shared by multiple countries. The YZ basin features diverse climatic systems, and the westerly system, along with varying glacier and snow conditions. Therefore, it serves as a typical watershed for studying the climate–cryosphere–hydrological processesover the TP. Research indicates that since the mid-1960s, the YZ basin has undergone a significant temperature increase of 0.3–0.4 °C per decade, along with minor rises in precipitation and runoff3–8. However, substantial uncertainties remain in the spatiotemporal variability of climate and runoff within the YZ basin, primarily due to the uneven spatial coverage of hydrometeorological stations and data discontinuity. The China Meteorological Administration (CMA) operates 16 national meteorological stations in the YZ basin, but their central and eastern valley concentration limits insights into climatic changes in upstream and downstream regions9. Current gridded datasets from satellite observations, reanalysis, and regional climate models exhibit significant differences in magnitude and spatiotemporal patterns across the YZ basin10–14. This underscores the pressing need to enhance meteorological observations in both the upstream and downstream regions of the YZ basin. Numerous studies have employed land surface hydrological models to analyze the YZ basin’s hydrological response to climate change3,6,7,10,15–20. However, differences among existing studies primarily arise from the variations in meteorological forcing data, model structures, and parameterization schemes. Across the YZ basin, gauge, satellite, and reanalysis products exhibit considerable discrepancies in mean annual precipitation, ranging from 360 to 1300 mm11, which can induce runoff simulation errors exceeding 100% when employed as hydrological model inputs. The reliability of variables employed in model parameter calibration strongly influences the accuracy of simulated runoff and the understanding of runoff components. Ref. 9 estimated that, using an uncalibrated and unvalidated hydrological model due to insufficient hydrometeorological data, glacier runoff accounted for approximately 45% of the total annual runoff in the Nuxia-Pasighat sub-basin of the YZ basin’s downstream region from 1980 to 2000. Ref. 20 validated the Variable Infiltration Capacity–Glacier (VIC-Glacier) model using new rain gauge and streamflow data, along with glacier mass balance and snow cover observations. The revised analysis revealed that glacier runoff accounted for approximately 19% of the total runoff in the NX-BXK sub-basin from 1971 to 2020. Owing to error compensation effects within the basin, hydrological simulations may perform well at the entire basin scale but still exhibit considerable deviations at individual hydrological cross-sections. This highlights the necessity of achieving a more uniform spatial distribution of hydrometeorological stations. In recent years, to address the uncertainty of meteorological data over the YZ basin, a series of studies have developed high-accuracy data products through multi-source data fusion and bias correction methods. Ref. 21 utilized four statistical correction techniques on the APHRODITE extreme precipitation dataset, enhancing the precision of extreme precipitation indices in the YZ basin. This approach reduced the consecutive wet days error to 0.8–3.1 days and limited the heavy precipitation days error to within one day. Ref. 13 developed a high-resolution precipitation product (3-hourly, 5 km) for the YZ basin from 1981 to 2016 by integrating gauge observations, satellite estimates, and reanalysis data through a weighted fusion method. The dataset showed high accuracy at both daily and monthly scales and was validated in hydrological simulations. Ref. 22 employed a dynamic Bayesian model averaging approach to merge ground observations and satellite precipitation data, generating a daily 0.1°-resolution precipitation dataset for the warm seasons from 2014 to 2019, exhibiting substantial accuracy in hydrometeorological changes. Similarly, ref. 14 in 2022, a random forest algorithm was applied to data from 260 stations to refine the ERA5 daily precipitation dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF), resulting in a 10 km resolution dataset for the YZ basin spanning 1951–2020. The results effectively reduced the systematic overestimation of ERA5 precipitation by 10–50%. However, these studies primarily focused on the correction and generation of precipitation or extreme precipitation data, without considering the synergistic effects of other meteorological variables and precipitation on hydrological simulations. More recently, ref. 2 in 2025, a high-resolution (1/30°) near-surface meteorological dataset for the Third Pole region (1979–2020) was developed by integrating ERA5 reanalysis data, high-resolution WRF model simulations, and data from over 600 ground-based stations. The dataset employs a convolutional neural network for downscaling and integrates a random forest–kriging fusion method, encompassing seven variables: air temperature, specific humidity, wind speed, air pressure, shortwave radiation, longwave radiation, and precipitation. Although this dataset demonstrates superior temporal and spatial accuracy compared with ERA5, the lack of meteorological observations in the upper and lower reaches of YZ basin limits its spatial representativeness and the reliability of its application in hydrological modeling for this basin. Given the above issues, in this study, we established an integrated hydrometeorological observation network in the YZ basin by deploying additional five automatic weather stations (AWSs), five rain gauges, and five hydrological sections to complement the existing national meteorological and hydrological stations. This expanded network provides a high-resolution and more evenly distributed monitoring system that spans from the glacierized headwaters to the lower mainstream, covering a wide elevation range. For meteorological data generation, we adopted a multi-source data fusion approach by integrating ERA5 reanalysis data with a random forest–based machine learning method. This approach has been widely used and proven effective in previous studies (e.g., CMFD), particularly for reconstructing meteorological variables in high-altitude, data-scarce regions. Building on these observations, we employed the process-based VIC-Glacier hydrological model to generate continuous gridded daily datasets of key hydrometeorological variables and runoff components at 10 km resolution for 1961–2024. For user convenience, and considering flow routing time, monthly total runoff and its components (rainfall runoff, glacier runoff and snowmelt) at different stations for 1961–2024 are also provided. This newly developed dataset not only fills gaps and harmonizes discontinuous station records but also offers a physically consistent representation of climate-cryosphere-hydrological processes across the basin. The combined observational–modeling framework provides a valuable basis for advancing our understanding of climate variability, glacier and snow dynamics, and runoff formation in the YZ basin, while also supporting reliable water resources assessment and management in this data-scarce but climatically critical region. Methods This study provides three parts of data. The initial section comprises daily meteorological data, such as precipitation (P), mean, maximum, and minimum temperatures (Tmean, Tmax, Tmin), along with wind speed (WS), gathered from 26 stations. The second part includes monthly total runoff and its components (snowmelt, glacier runoff, and rainfall runoff) from 15 hydrological stations, taking into account flow routing time. The third section involved creating a daily gridded hydrometeorological dataset with a 10 km spatial resolution by extending and interpolating data from 26 meteorological and 15 hydrological stations, integrating these variables with ERA5 meteorological data. In this study, we focus on analyzing the observed variations at meteorological and hydrological stations. hydrometeorological observations In the YZ basin, 16 national meteorological stations were established, yet their concentration in the central valleys (Fig. 1) constrains understanding of climate variability in the upstream and downstream region. To address this gap, we deployed three Automatic Weather Stations (AWSs) with T200B rain gauges in the upper reaches at elevations of 4485–5050 m (Table 1), and two additional AWSs in the upstream regions of the Lasa and Rikaze sub-basins, respectively. In addition, five tipping-bucket rain gauges were installed along the lower mainstream of the YZ river at elevations of 820–870 m. Together with the 16 existing national meteorological stations, these form a meteorological observation network spanning from the headwaters to the lower reaches, covering an elevation range of 820–5050 m. Owing to the discontinuity of station records (Table 1), the relationship between station observations and the corresponding grid data from ERA5 was used to harmonize and infill the records, thereby extending them into continuous daily series for 1961–2024.Fig. 1(a) Geographic locations of the hydrometeorological observed stations. (b) The instrument of Automatic Weather Station. (c) The instrument of T200B rain gauge. (d) The instrument of Radar Water Level Gauge.Table 1Information of meteorological observations for the YZ basin.Station nameLatitude (°N)Longitude (°E)Elevation (m)VariablesTemporal resolutionPeriodNational Meteorological StationLazi29.0887.604000P,T, WSDaily1978.1‒2023.12Rikaze29.2588.883836P,T, WSDaily1956.1‒2023.12Nanmulin29.6889.104000P,T, WSDaily1991.1‒2023.12Jiangzi28.9189.604040P,T, WSDaily1957.1‒2023.12Nimu29.4390.163809.4P,T, WSDaily1974.1‒2023.12Langkazi28.9690.404432.4P,T, WSDaily1991.1‒2023.12Gongga29.3090.983555.3P,T, WSDaily1991.1‒2023.12Dangxiong30.4891.104200P,T, WSDaily1963.1‒2023.12Lasa29.6691.133648.9P,T, WSDaily1955.1‒2023.12Mozhugongka29.8591.733047.4P,T, WSDaily1991.1‒2023.12Zedang29.2591.763551.7P,T, WSDaily1957.1‒2023.12Jiacha29.1592.583260P,T, WSDaily1991.1‒2023.12Jiali30.6693.284488.8P,T, WSDaily1955.1‒2023.12Milin29.2194.212950P,T, WSDaily1991.1‒2023.12Linzhi29.6694.332991.8P,T, WSDaily1954.1‒2023.12Bomi29.8695.762736P,T, WSDaily1955.1‒2023.12Automatic Meteorological StationMayoumulashan30.5982.465050P,T, WSDaily2022.1‒presentZhuzhucun29.6184.374830P,T, WSDaily2022.1‒presentSaga29.3285.254485P,T, WSDaily2022.1‒presentLindixiang30.5491.364400P,T, WSDaily2022.1‒presentKangmaxian28.4189.564406P,T, WSDaily2023.1‒presentRain GaugeMiri29.4195.40856PDaily2023.1‒presentDexing29.3295.29820PDaily2023.1‒presentBeibeng29.2495.17870PDaily2023.1‒presentXirang29.1895.03840PDaily2023.1‒presentDidong29.2195.09863PDaily2023.1‒present (a) Geographic locations of the hydrometeorological observed stations. (b) The instrument of Automatic Weather Station. (c) The instrument of T200B rain gauge. (d) The instrument of Radar Water Level Gauge. Table 1Information of meteorological observations for the YZ basin.Station nameLatitude (°N)Longitude (°E)Elevation (m)VariablesTemporal resolutionPeriodNational Meteorological StationLazi29.0887.604000P,T, WSDaily1978.1‒2023.12Rikaze29.2588.883836P,T, WSDaily1956.1‒2023.12Nanmulin29.6889.104000P,T, WSDaily1991.1‒2023.12Jiangzi28.9189.604040P,T, WSDaily1957.1‒2023.12Nimu29.4390.163809.4P,T, WSDaily1974.1‒2023.12Langkazi28.9690.404432.4P,T, WSDaily1991.1‒2023.12Gongga29.3090.983555.3P,T, WSDaily1991.1‒2023.12Dangxiong30.4891.104200P,T, WSDaily1963.1‒2023.12Lasa29.6691.133648.9P,T, WSDaily1955.1‒2023.12Mozhugongka29.8591.733047.4P,T, WSDaily1991.1‒2023.12Zedang29.2591.763551.7P,T, WSDaily1957.1‒2023.12Jiacha29.1592.583260P,T, WSDaily1991.1‒2023.12Jiali30.6693.284488.8P,T, WSDaily1955.1‒2023.12Milin29.2194.212950P,T, WSDaily1991.1‒2023.12Linzhi29.6694.332991.8P,T, WSDaily1954.1‒2023.12Bomi29.8695.762736P,T, WSDaily1955.1‒2023.12Automatic Meteorological StationMayoumulashan30.5982.465050P,T, WSDaily2022.1‒presentZhuzhucun29.6184.374830P,T, WSDaily2022.1‒presentSaga29.3285.254485P,T, WSDaily2022.1‒presentLindixiang30.5491.364400P,T, WSDaily2022.1‒presentKangmaxian28.4189.564406P,T, WSDaily2023.1‒presentRain GaugeMiri29.4195.40856PDaily2023.1‒presentDexing29.3295.29820PDaily2023.1‒presentBeibeng29.2495.17870PDaily2023.1‒presentXirang29.1895.03840PDaily2023.1‒presentDidong29.2195.09863PDaily2023.1‒present Information of meteorological observations for the YZ basin. Currently, five national hydrological stations are used to analyze runoff variations in the YZ basin, including Lazi, Yangcun, and Nuxia on the mainstem, and Lhasa and Rikaze on tributaries (Fig. 1, Table 2). Building on this foundation, we established five additional hydrological observation sections. First, a section was set near the Nuxia national hydrological station for data comparison. Second, a MT section was installed in the lower mainstream to address data scarcity that could otherwise lead to considerable biases in hydrological model simulations. Third, to account for human impacts on runoff, a section (QS) was added in the lower mainstream of Lasa river basin. In addition, two sections (YG and BM) were set on two major tributaries in the downstream glacierized area, the Yigong Zangbo and Palong Zangbo, respectively. The combination of the five new sections and the five existing national hydrological stations constitutes a comprehensive hydrological observation network that spans the mainstem, tributaries, and glacier source regions.Table 2Information of hydrological observations for the YZ basin.Station nameLatitude (°N)Longitude (°E)Elevation (m)Temporal resolutionPeriodNational Hydrological StationLazi29.0587.384397Monthly1971.1‒2015.12Yangcun29.2891.883300Monthly1971.1‒2015.12Lasa29.6391.153658Monthly1971.1‒2015.12Rikaze29.2588.883834Monthly1971.1‒2015.12Nuxia29.4794.572917Monthly1971.1‒2015.12Hydrological GaugeNuxia29.4794.572917Daily2024.1‒presentQS29.4490.933586Daily2024.1‒presentYG30.2494.872231Daily2023.1‒presentBM29.995.482670Daily2024.1‒presentMT29.3295.29701Daily2023.1‒present Table 2Information of hydrological observations for the YZ basin.Station nameLatitude (°N)Longitude (°E)Elevation (m)Temporal resolutionPeriodNational Hydrological StationLazi29.0587.384397Monthly1971.1‒2015.12Yangcun29.2891.883300Monthly1971.1‒2015.12Lasa29.6391.153658Monthly1971.1‒2015.12Rikaze29.2588.883834Monthly1971.1‒2015.12Nuxia29.4794.572917Monthly1971.1‒2015.12Hydrological GaugeNuxia29.4794.572917Daily2024.1‒presentQS29.4490.933586Daily2024.1‒presentYG30.2494.872231Daily2023.1‒presentBM29.995.482670Daily2024.1‒presentMT29.3295.29701Daily2023.1‒present Information of hydrological observations for the YZ basin. To better analyze runoff variations at different sections of the main stream and tributaries in the YZ basin, in addition to the five national hydrological stations and five self-established observation sites, we selected five virtual hydrological stations. These virtual stations are located at the river sections closest to our five AWSs (Fig. 1). Three sites are located in the upper YZ basin (MY, ZZC, and SG), one in the Rikaze sub-basin headwaters (KM), and one in the Lasa sub-basin headwaters (LD). Generation of a 10 km daily meteorological dataset for 1961–2024 A 10 km daily meteorological dataset for the YZ basin was created by extending and interpolating records from 26 stations. This dataset, covering 1961–2024, includes precipitation (P), mean temperature (Tmean), maximum temperature (Tmax), minimum temperature (Tmin), and wind speed (WS), with bias correction applied for consistency and accuracy. This dataset represented an update of existing data from ref. 14 and ref. 20, and was validated by observations from 16 national meteorological stations. The flowchart of the data generation process is presented in Fig. 2.Fig. 2Flowchart of the data generation process. Flowchart of the data generation process. Daily precipitation data from ERA5 for 1961–2020 at 10 km resolution was corrected by ref. 14 in both temporal and spatial dimensions. Due to ERA5’s coarse resolution, spatial correction utilized high-resolution (~10 km) WRF Model data to derive grid-specific annual mean precipitation correction factors, allowing ERA5 to reflect WRF’s detailed spatial distribution. Subsequently, monthly corrections were applied using 580 rain gauges in the monsoon-dominated TP region, including 290 in the YZ basin, alongside a random-forest (RF) machine learning algorithm14. The RF algorithm in this study utilizes inputs comprising (1) geographical features such as longitude, latitude, elevation, slope gradient, and aspect, which affect precipitation distribution, and (2) climatic features from ERA5, including convective available potential energy, lifting condensation level, and total column water vapor, indicative of precipitation generation and development potential. The main contribution of this study lies in deriving grid-specific correction factors for ERA5 data over the YZ river basin14, which were directly used as a continuous background field and corrected in both spatial and temporal dimensions for 1961–2024 in this study. Temperature data from 1961 to 2020 was adjusted using orographic relationships, while wind data was refined through the Inverse Distance Weighting interpolation method20. Temperature data from the 16 CMA stations are adjusted using lapse rates of −0.80 °C/100 m for January–March, −0.70 °C/100 m for April–May, −0.56 °C/100 m for June, −0.40 °C/100 m for July–September, and 0.58 °C/100 m for October–December during the gridding process. Correction factors were derived from the relationships between ERA5 temperature and wind speed data and their bias-corrected counterparts from ref. 20, calculated as multi-year monthly means for each grid during 1961–2020, and subsequently applied to the temperature and wind speed data for 2021–2024. For stations with short observation records (e.g., Miri Station with data available only after 2023), historical precipitation series were not directly extended from observations. Instead, precipitation at these locations was extracted from the reconstructed gridded dataset. These stations were used only for independent validation and were not involved in the RF model training process. The reliability of the reconstructed precipitation dataset was evaluated by comparing it with available gauge observations at multiple temporal scales, including both annual and daily scales. While our study follows a similar methodological framework14, several important improvements and extensions have been made. First, this study specifically focuses on the YZ river basin and aims to generate a consistent meteorological-hydrological dataset. The reconstructed meteorological data are directly used to drive hydrological simulations, enabling not only improved runoff estimation but also a more physically consistent analysis of hydrological processes. Second, a larger number of meteorological and hydrological stations were incorporated in this study. These datasets include newly collected field observations in recent years, which were not available in the previous study, thereby improving the reliability of the correction and reconstruction. Finally, the reconstructed dataset has been extended to cover the period 1961–2024, providing a more up-to-date and comprehensive dataset for the YZ river basin. To the best of our knowledge, this represents one of the most complete long-term meteorological datasets currently available for this region. Generation of a 10 km daily hydrological dataset for 1961–2024 A 10 km daily meteorological dataset was utilized as the input for the VIC-Glacier hydrological model, enabling the simulation of a daily hydrological dataset at a 10 km spatial resolution from 1961 to 2024. This modeling framework enables detailed simulation of water and energy exchanges across a grid mesh that includes soil, vegetation, and the atmosphere23,24. The model incorporates a two-layer energy-balance snow model25, a degree-day glacier melt approach26, and a frozen soil–permafrost algorithm27. The VIC-Glacier model has proven effective in hydrological simulations across multiple high-mountain TP basins3,7,9,19,20,28–30. The modeling framework, utilizing a 10 km × 10 km spatial resolution and model parameters, was adopted from ref. 20. Glacier model parameters, specifically degree-day factors (DDFs), were calibrated using glacier area data from the Randolph Glacier Inventory (RGI) 6.0 for the 2000s–2010s in the YZ basin, and validated with mass balance observations from the Gurenhekou and Parlung No. 4 glaciers. Ref. 20 studied 94 glaciers. The VIC model parameters for runoff simulation in non-glacierized areas, comprising the infiltration parameter (B_inf), depths of the first and second soil layers (D1, D2), maximum baseflow velocity (Dsmax), fraction of Dsmax initiating non-linear baseflow (Ds), and fraction of maximum soil moisture triggering non-linear baseflow (Ws), were validated using observed monthly streamflow data from five national hydrological stations and snow cover fraction estimates20. The parameter B_inf typically ranges from 0 to 0.4, while D2 ranges from 0.5 to 1.0. The DDFs range from 9.2 to 10.97 in the upper reaches and from 6.5 to 6.8 in the middle and lower reaches. The VIC-Glacier model effectively simulates streamflow, glacier area, glacier mass balance, and snow cover dynamics in the YZ basin, as evidenced by the strong correlation between simulation results and observations. Data Records The National Tibetan Plateau Data Center provides open access to extended and interpolated records from 26 meteorological and 15 hydrological stations, along with a daily hydrometeorological dataset at 10 km resolution for 1961–2024 in the YZ basin (10.11888/Atmos.tpdc.303008, 10.11888/Atmos.tpdc.303008). The dataset is available at the National Tibetan Plateau Data Center31. Daily meteorological dataset at 26 stations is provided in an Excel file named Meteorological data_daily.xlsx, with each sheet corresponding to a meteorological station. Each sheet comprises 8 columns detailing year, month, day, precipitation (mm), mean temperature (°C), maximum temperature (°C), minimum temperature (°C), and wind speed (m/s) from 1961 to 2024. The gridded dataset is available in plain text file format, comprising 3,380 grids, each labeled according to its geographic coordinates as fluxes_latitude_longitude. Each file comprises 12 columns detailing year, month, day, precipitation, mean temperature, maximum temperature, minimum temperature, wind speed, total runoff (mm), snowmelt (mm), glacier runoff (mm), and rainfall runoff (mm). To enhance user convenience and consider flow routing time, we offer monthly total runoff data and its components (m3/s) from 1961 to 2024 for 15 stations in an Excel file, Streamflow_month.xlsx, with each sheet representing a hydrological station. Each sheet includes six columns for year, month, total runoff (m3/s), glacier runoff (m3/s), snowmelt (m3/s), and rainfall runoff (m3/s). An accompanying document provides details on the latitude, longitude, and elevation of each meteorological and hydrological station. Technical Validation Precipitation and temperature The adjusted precipitation and temperature dataset was evaluated at the point scale using observations from 16 national meteorological stations at the daily, monthly, and mean annual scales for 2021–2023 (Fig. 3; Table 3). At the daily scale (Fig. 3a,b), the relative bias (RB, %) of precipitation exhibits substantial spatial variability, ranging from −47.05% to 45.15%. Most stations show positive biases, indicating an overall overestimation of precipitation, particularly at stations such as Mayoumulashan and Lazi, where the relative bias exceeds 35%. In contrast, a few stations (e.g., Beibeng) exhibit significant underestimation. This pattern suggests that the adjustment procedure effectively enhances precipitation estimates in high-altitude regions, whereas larger uncertainties persist in downstream, precipitation-rich areas, likely due to increased spatial variability and complex hydrometeorological processes. For temperature, the bias ranges approximately from −1.02 °C to 1.36 °C, with both positive and negative deviations distributed across stations, suggesting no strong systematic bias but noticeable local discrepancies.Fig. 3Comparison between observations and gridded precipitation and temperature at the daily (a,b), monthly (c,d), and mean annual (e,f) scales for 2021–2023. The y-axes in panels (a) and (b) correspond to station names.Table 3Comparison of daily observations with gridded precipitation and temperature in the YZ basin.Station namePrecipitationTemperatureRelative Bias(%)Bias (°C)Dangxiong6.62−0.53Lazi38.17−0.74Linzhi9.350.52Rikaze1.970.78Nimu28.27−0.51Zedang−1.940.51Nanmulin−0.61−1.01Mozhugongka11.980.49Gongga13.66−0.49Jiangzi29.750.63Langkazi11.431.04Jiali13.101.02Bomi22.350.98Jiacha19.04−1.02Lasa18.86−1.01Milin6.490.24Saga−2.410.93Lindixiang−3.760.43Mayoumulashan45.151.36Kangmaxian16.551.07Zhuzhucun16.74−0.27Beibeng−47.05Dexing24.44Didong−3.35Miri5.69Xirang17.95 Comparison between observations and gridded precipitation and temperature at the daily (a,b), monthly (c,d), and mean annual (e,f) scales for 2021–2023. The y-axes in panels (a) and (b) correspond to station names. Table 3Comparison of daily observations with gridded precipitation and temperature in the YZ basin.Station namePrecipitationTemperatureRelative Bias(%)Bias (°C)Dangxiong6.62−0.53Lazi38.17−0.74Linzhi9.350.52Rikaze1.970.78Nimu28.27−0.51Zedang−1.940.51Nanmulin−0.61−1.01Mozhugongka11.980.49Gongga13.66−0.49Jiangzi29.750.63Langkazi11.431.04Jiali13.101.02Bomi22.350.98Jiacha19.04−1.02Lasa18.86−1.01Milin6.490.24Saga−2.410.93Lindixiang−3.760.43Mayoumulashan45.151.36Kangmaxian16.551.07Zhuzhucun16.74−0.27Beibeng−47.05Dexing24.44Didong−3.35Miri5.69Xirang17.95 Comparison of daily observations with gridded precipitation and temperature in the YZ basin. Precipitation bias demonstrates a clear seasonal pattern at monthly scale (Fig. 3c,d). Biases are relatively small during the early months of the year but become strongly negative during the mid-year period (June–August), indicating underestimation during the peak precipitation season. The large error bars further suggest considerable variability and uncertainty during these months. Similarly, temperature bias remains consistently negative throughout the year, generally ranging from −2 °C to −4 °C, indicating a systematic underestimation of temperature across all months, with larger deviations observed in winter and late autumn. Strong agreement between observations and gridded data is also observed at the mean annual scale (Fig. 3e,f). Precipitation shows a high correlation coefficient (CC = 0.98) with a relative bias of 14%, indicating that the dataset effectively captures spatial variability despite moderate overestimation. Temperature also demonstrates strong consistency, with a correlation coefficient of 0.96 and a mean bias of approximately 1 °C. The linear relationships in both cases confirm the overall reliability of the adjusted dataset at the annual scale. Such discrepancies are likely related to differences in data sources, spatial resolution (10 km for gridded datasets), and the complex topography and climatic heterogeneity of the YZ river basin. First, the dataset used in this study has been validated and demonstrates high reliability in reproducing regional climate variability. As shown in Fig. 3f, the correlation coefficient (CC = 0.96) indicates a strong consistency between the observed and reanalysis temperature, suggesting that the dataset captures the temporal variability well despite the presence of bias. Second, the large temperature bias can be partly attributed to the complex geographical and climatic characteristics of the YZ basin. The basin spans a vast area across the TP and extends from high-altitude cold regions to relatively warmer and humid downstream regions. Specifically, the basin covers multiple climate zones. Such strong spatial heterogeneity leads to significant temperature gradients. When averaged at the basin scale over multiple years, these differences can introduce systematic bias in the comparison. Third, scale mismatch between observations and reanalysis data is another important source of error. The observed temperature is typically measured at a single meteorological station (point-scale), whereas ERA5 data represent an area-average over a 10 km grid. In regions with complex terrain (e.g., steep elevation gradients in the TP), this spatial representativeness difference can lead to considerable discrepancies between station observations and gridded data. In addition to point-based validation, the generated gridded dataset was also evaluated against the Fifth Generation ECMWF Reanalysis (ERA5) and the China Meteorological Forcing Dataset (CMFD), which are widely used datasets at both global and national (China) scales. Figure 4 presents the spatial distribution of mean precipitation and temperature across the YZ river basin derived from YZ observations, CMFD, and ERA5 datasets. Overall, the three datasets exhibit broadly consistent spatial patterns, indicating a coherent representation of the regional climate structure. For precipitation, all datasets capture a clear southeast–northwest gradient, with higher precipitation in the downstream southeastern region and lower values over the upstream YZ basin. This pattern is consistent with the influence of the Indian monsoon and orographic effects along the Himalayan range. CMFD and ERA5 show comparable spatial structures, although slight differences in magnitude are observed, particularly in regions with complex topography. For temperature, a pronounced elevation-dependent pattern is evident across all datasets, with lower temperatures over the high-altitude upstream regions and higher temperatures in the lower-elevation downstream areas.Fig. 4Comparison of spatial distributions of precipitation and temperature in the YZ river basin based on YZ, CMFD, and ERA5 datasets. Comparison of spatial distributions of precipitation and temperature in the YZ river basin based on YZ, CMFD, and ERA5 datasets. Although the three datasets exhibit consistent large-scale spatial patterns, notable differences exist in both spatial distribution and magnitude based on multi-year mean values. The ERA5 precipitation generally overestimates the YZ dataset by approximately 63%, with local differences in the southeastern part of the basin exceeding120%. In contrast, CMFD shows relatively closer agreement with YZ but still exhibits a slight overestimation of about 5–15% in most areas. Compared to YZ dataset, ERA5 tends to underestimate temperature by about 1 °C, particularly in high-elevation regions, with differences of approximately 1–3 °C. CMFD shows better agreement overall but still presents a slight warm bias of about 0.5–1.5 °C in downstream glacier areas. These differences are more pronounced in regions with strong elevation gradients. Given that the newly established meteorological stations are primarily situated in the upstream high-altitude region (Lazi hydrological station, LZ) and the downstream high-precipitation region (Nuxia hydrological station–outlet, NX-BXK), where the strongest spatial heterogeneity is observed, these two sub-basins were selected as representative regions. The multi-year mean seasonal patterns of precipitation and temperature from the three datasets (YZ, CMFD, and ERA5) were subsequently compared to assess their performance under contrasting climatic conditions (Fig. 5).Fig. 5Mean monthly pattern of precipitation and temperature at LZ and BXK stations based on YZ observations, CMFD, and ERA5 datasets. Mean monthly pattern of precipitation and temperature at LZ and BXK stations based on YZ observations, CMFD, and ERA5 datasets. All precipitation datasets capture a pronounced seasonal cycle in these sub-basins, with more than 70% of annual precipitation concentrated during the wet season (June–September) and relatively low precipitation during the remaining months. However, substantial differences in magnitude are observed. In the upstream LZ sub-basin, the YZ dataset shows markedly lower summer precipitation compared to both ERA5 and CMFD. Specifically, ERA5 overestimates YZ precipitation dataset by approximately 40–150% in summer, while CMFD shows overestimations of about 80–90%. In contrast, YZ indicates higher precipitation than CMFD during the spring season, with increases ranging from 30–77%. A similar pattern is observed in the downstream NX–BXK sub-basin. During the wet season (June–September), ERA5 and CMFD both overestimate precipitation relative to YZ dataset, with biases of approximately 80–90% for ERA5 and 17–43% for CMFD. In the dry season (winter–spring), YZ shows higher precipitation than CMFD by about 35–50%, whereas ERA5 substantially overestimates precipitation, with differences reaching 80–99%. For temperature, all datasets reproduce a clear seasonal cycle, with higher temperatures in summer and lower temperatures in winter. However, notable differences in magnitude are evident when using the YZ dataset as a reference. In the LZ sub-basin, the YZ dataset indicates higher temperatures during the winter–spring period by approximately 1.1–4.5 °C compared to ERA5 and CMFD, while showing lower temperatures in summer by about 0.4–1.5 °C. This suggests that ERA5 generally exhibits a pronounced cold bias throughout the year, particularly in winter, with deviations reaching 2–4.5 °C, whereas CMFD shows relatively smaller differences, generally within ±2 °C. In the NX–BXK sub-basin, YZ temperature dataset is generally close to CMFD during the summer months, with differences typically within ±1 °C, indicating good agreement under warm conditions. However, during the winter–spring period, YZ shows substantially lower temperatures than CMFD by approximately 1.5–5.8 °C, highlighting a clear seasonal discrepancy. Compared to ERA5, the YZ dataset consistently presents lower temperature values across all seasons, with differences of about 1.1–1.5 °C. Total runoff To evaluate the model’s performance in simulating runoff at different scales, simulations were firstly compared against observed streamflow at five national hydrological stations for 1971–2015 due to their long-term records (Fig. 6). 1971–2000 was selected as the calibration period and 2001–2015 the validation period based on the observed monthly streamflow for 1971–2015. The simulated monthly streamflow accurately reproduced the observed magnitudes and seasonal variability at all hydrological stations, demonstrating consistent performance across the basin. NSE values ranged from 0.71 to 0.88, indicating good model reliability, while RB values between −6% and 4% suggest low systematic bias (Fig. 7a,b). Furthermore, the model successfully captured the timing and amplitude of peak flows as well as low-flow conditions, highlighting its ability to represent hydrological processes under varying flow regimes and across different spatial scales (Fig. 6).Fig. 6Time series of observed and simulated runoff at five national hydrological stations for 1971–2015.Fig. 7Model performance evaluation based on NSE and RB at hydrological stations during calibration and validation periods. Time series of observed and simulated runoff at five national hydrological stations for 1971–2015. Model performance evaluation based on NSE and RB at hydrological stations during calibration and validation periods. Similarly, validation using monthly observations from five field-deployed hydrological stations during 2024–2025 further confirmed the model’s reliability. Validation based on monthly observations from five field-deployed hydrological stations during 2024–2025 shows that NSE values range from 0.75 to 0.82, with RB between −8% and 6%. Simulated runoff is slightly higher than observations at the mainstream stations NX and MT by approximately 3–4%. In contrast, runoff at glacier-fed stations BM and YG is underestimated by about 4–6%, which may reflect an underrepresentation of glacier melt contributions. Overall, the agreement between simulations and observations across stations indicates consistent model performance (Fig. 7c). Given the consistent agreement between simulated and observed streamflow, a 10 km daily hydrological dataset for 1961–2024 was generated in the YZ basin. Based on this framework, streamflow records from observed hydrological stations were gap-filled and extended, yielding continuous daily series for 1961–2024 to overcome discontinuities in the observational periods. This provides a complete and continuous spatial depiction of streamflow along the river from upstream to downstream. Mean annual runoff of the YZ basin exhibited a distinct increasing trend from upstream to downstream for 1961–2024 (Fig. 8). In the upper reaches, from MY to LZ, mean annual runoff remained relatively small and stable, with values of 13 × 108 m3, 41 × 108 m3, 50 × 108 m3, and 59 × 108 m3, respectively. In the middle reaches, marked spatial contrasts were observed. The KM and RKZ tributaries contributed only 2 × 108 m3 and 15 × 108 m3, respectively, indicating limited inflow to the main stream. In contrast, the Lasa river made substantial contributions, with runoff increasing from 23 × 108 m3 at LD to 96 × 108 m3 at LS and 112 × 108 m3 at QS. Along the main stream, runoff increased to 290 × 108 m3 at YC and 610 × 108 m3 at NX, reflecting the significant influence of large tributaries. In the lower reaches, runoff accumulated rapidly. The YG and BM, located in the largest glacierized area of the YZ basin, recorded mean annual runoff of 100 × 108 m3 and 65 × 108 m3, respectively. The main stream further increased to 1000 × 108 m3 at MT and reached 1400 × 108 m3 at BXK, representing the final concentration of water resources within the basin. Overall, runoff in the YZ basin increased progressively along its course, characterized by relatively stable values in the upper reaches, rapid growth in the middle reaches, and pronounced accumulation in the lower reaches.Fig. 8Annual timeseries of total runoff for 1961–2024 in the YZ basin and decadal averages of streamflow at 15 hydrological stations. Number in the YZ basin map is mean annual streamflow (108 m3) for 1961–2024 at each hydrological station. Annual timeseries of total runoff for 1961–2024 in the YZ basin and decadal averages of streamflow at 15 hydrological stations. Number in the YZ basin map is mean annual streamflow (108 m3) for 1961–2024 at each hydrological station. Total runoff in the YZ basin increased at a rate of 21.2 m3 s−1 per decade for 1961–2024, with clear spatial variability. Runoff trends generally intensified downstream: upstream stations (MY, ZZZ, SG, LZ) showed modest increases (0.6–3.3 m3 s−1 per decade), middle reaches (LS, QS) exhibited moderate growth (6.7–7.9 m3 s−1 per decade), while some tributaries (KM, RKZ) experienced slight declines (−0.4 to −1.7 m3 s−1 per decade). In contrast, downstream stations (MT, BXK) recorded substantial increases, reaching 40.1 and 21.2 m3 s−1 per decade, respectively. Decadal variations relative to the 1960s further highlight distinct spatial–temporal patterns: upper and middle reaches experienced notable declines during the 1970s–1990s (−10% to −35%), followed by gradual recovery and positive anomalies (7%–10%) in recent decades; small tributaries such as KM and RKZ showed persistent decreases, remaining negative throughout the period. Meanwhile, the main stream at YC and NX remained below baseline levels despite partial recovery. In contrast, the lower reaches displayed sustained increases, with glacier-fed tributaries (YG, BM) and main stream stations (MT, BXK) exhibiting significant positive anomalies, reaching up to 41% in the 2020 s. Overall, these results indicate long-term declines with partial recovery in the upper and middle basin, alongside pronounced and sustained runoff enhancement in downstream regions. Runoff components We divided the total runoff into three components: rainfall runoff, snowmelt runoff, and glacier runoff. Glacier runoff encompasses all water produced in the glacierized region, comprising rainfall, snowmelt, and ice melt. Figure 9 illustrates the spatial distribution of the mean annual total runoff and its components (rainfall runoff, glacier runoff, and snowmelt) in the YZ basin from 1961 to 2024. The total runoff was higher in the southeastern and downstream regions, showing a clear spatial gradient, while the runoff values in the eastern and central areas were relatively lower. Rainfall runoff distribution mirrored total runoff patterns, as it was the primary driver of the basin’s runoff. Glacier runoff was primarily concentrated in the eastern region of Parlung Zangbo Glacier, while snowmelt runoff was mainly concentrated in the high-altitude areas, particularly in the eastern and central mountainous regions, where snowmelt contributed significantly to runoff.Fig. 9Spatial pattern of mean annual total runoff and its three components (rainfall runoff, glacier runoff and snowmelt) during 1961–2024 in the YZ basin. Spatial pattern of mean annual total runoff and its three components (rainfall runoff, glacier runoff and snowmelt) during 1961–2024 in the YZ basin. The 1961–2024 daily hydrometeorological dataset for the YZ basin provides a robust foundation for a wide range of scientific and practical applications in hydrology, climatology, ecology, disaster risk assessment, and hydropower development across the southern TP. The continuous series of precipitation, temperature, and runoff variables enable quantitative simulations of the hydrological cycle and water balance across multiple temporal and spatial scales. The long-term coverage from 1961 to 2024 provides an exceptional foundation for detecting hydroclimatic trends and assessing the impacts of global warming on cryosphere and hydrosphere interactions in the YZ basin. By capturing the transition from relatively stable hydroclimatic conditions in the mid-20th century to the accelerated warming and wetting phases of recent decades, the dataset allows for detailed analyses of how climate change reshapes precipitation regimes, snow and glacier melt processes, and runoff composition across the TP. The dataset supports basin-wide simulations of runoff generation, routing, and partitioning under varying climatic and cryospheric conditions. It can be directly used to calibrate and validate process-based hydrological models such as VIC, SWAT, and HBV, thereby improving the simulation of rainfall–runoff relationships, snow and glacier melt dynamics, and extreme hydrological events. In addition, the dataset provides consistent boundary and forcing data for land surface and ecosystem models, facilitating coupled analyses of cryosphere-hydrosphere-biosphere feedbacks. When integrated with remote-sensing products such as NDVI or LAI, it enables investigations into the hydroclimatic controls on vegetation dynamics, ecosystem productivity, and alpine grassland degradation. Beyond individual scientific domains, this dataset bridges observation gaps in the southern TP, particularly in the downstream regions of the YZ basin. Its open-access design facilitates integration with regional and global reanalysis products, climate projections, and Earth system model outputs. The dataset is therefore expected to play a critical role in regional water–climate–ecosystem assessments, CMIP model evaluation, and the broader objectives of the Third Pole Environment (TPE) research program. In summary, this long-term dataset establishes a benchmark for future multi-sphere modeling of the TP, providing a scientific foundation for integrated studies of water, energy, and ecosystem evolution under changing climate conditions.