JGR–Atmospheres

In the context of China's “dual carbon” goal, emissions of air pollutants are expected to significantly decrease in the future. Thus, the direct climate effects of black carbon (BC) aerosols in East Asia are investigated under this goal using an updated regional climate and chemistry model. The simulated annual average BC concentration over East Asia is approximately 1.29 μg/m3 in the last decade. Compared to those in 2010–2020, both the BC column burden and instantaneous direct radiative forcing in East Asia decrease by more than 55% and 80%, respectively, in the carbon peak year (2030s) and the carbon neutrality year (2060s). Conversely, the BC effective radiative forcing (ERF) and regional climate responses to BC exhibit substantial nonlinearity to emission reduction, possibly resulting from different adjustments of thermal-dynamic fields and clouds from BC-radiation interactions. The regional mean BC ERF at the tropopause over East Asia is approximately +1.11 W/m2 in 2010–2020 while negative in the 2060s. BC-radiation interactions in the present-day impose a significant annual mean cooling of −0.2 to −0.5 K in central China but warming +0.3 K in the Tibetan Plateau. As China's BC emissions decline, surface temperature responses show a mixed picture compared to 2010–2020, with more cooling in eastern China and Tibet of −0.2 to −0.3 K in the 2030s, but more warming in central China of approximately +0.3 K by the 2060s. The Indian BC might play a more important role in East Asian climate with reduction of BC emissions in China.

Stratiform and convective precipitation are known to be associated with distinct isotopic fingerprints in the tropics. Such rain type specific isotope signals are of key importance for climate reconstructions derived from climate proxies (e.g., stable isotopes in tree rings). Recently, the relation between rain type and isotope signal in present-day climate has been intensively discussed. While some studies point out the importance of deep convection, other studies emphasize the role of stratiform precipitation for strongly depleted isotope signals in precipitation. Uncertainties arise from observational studies due to data scarcity while modeling approaches with global climate models cannot explicitly resolve convective processes and rely on parameterizations. High-resolution climate models are particularly important for studies over complex topography and for the simulation of convective cloud formation and organization. Therefore, we applied the isotope-enabled version of the high-resolution climate model from the Consortium for Small-Scale Modeling (COSMOiso) over the Andes of tropical south Ecuador, South America, to investigate the influence of stratiform and convective rain on the stable oxygen isotope signal of precipitation (δ18OP). Our results highlight the importance of deep convection for depleting the isotopic signal of precipitation and increasing its deuterium excess. Due to the opposing effect of shallow and deep convection on the δ18OP signal, the use of a stratiform fraction might be misleading. We therefore propose to use a shallow and deep convective fraction to analyze the effect of rain types on δ18OP.

The eddy-covariance (EC) method assumes a homogeneous underlying surface. However, recent studies increasingly examining on EC measurements across diverse surfaces, raising concerns about measurement precision and accuracy. This study evaluates the impacts of altering the emission height and rate on the EC measurements through utilizing an artificial source emission system. The results demonstrated a significant impact of changes in the emission height and rate on the EC measurements. Higher emission height may lead to the underestimation of the measured EC fluxes, attributed to the variations in the footprint area and turbulent transport. Traditional data quality control methods may discard effective EC data during sudden changes in the emission rate. Therefore, to secure effective data and accurately observe emissions, it was practical to analyze the auxiliary factors, such as environmental factors, such as vapor pressure deficit (VPD). An unresolved issue would persist with the correction of the EC method for accurately capturing the actual emission signals when there was a sudden increase in the data range or deviation. Furthermore, comparing the footprint model estimations with the actual emissions demonstrated the necessity of footprint analyses, offering a valuable reference for the data calibration when the uncertainties arose owing to inhomogeneous underlying surfaces. Although EC fluxes across the three averaging periods indicated no significant differences, the footprint model suggested that 15-min interval was the optimal. Further validation experiments are required for the EC measurements in locations with complex source conditions to enhance our understanding of land-atmosphere flux exchange.

The Chinese Loess Plateau (CLP) in northern China is home to one of the most prominent loess records in the world, reflecting past eolian dust activity in East Asia. However, their interpretation is hampered by ambiguity in the origin of loess-forming dust and an incomplete understanding of the circulation forcing dust accumulation. In this study, we used a novel modeling approach combining a dust emission model FLEXDUST with simulated back trajectories from FLEXPART to trace the dust back to where it was emitted. Over 21 years (1999–2019), we modeled back trajectories for fine (∼2 μm) and super-coarse (∼20 μm) dust particles at six CLP sites during the peak dust storm season from March to May. FLEXPART source-receptor relationships are combined with the dust emission inventory from FLEXDUST to create site-dependent high-resolution maps of the source contribution of deposited dust. The nearby dust emission areas were found to be the main source of dust to the CLP. Dust deposition across the CLP was found to predominantly occur via wet removal, with also some super-coarse dust from distant emission regions being wet deposited following high-level tropospheric transport. The high topography located on the downwind side of the emission area plays an essential role in forcing the emitted super-coarse dust upward. On an interannual scale, the phase of the Arctic Oscillation in the preceding winter was found to have a strong association with the spring deposition rate on the CLP, while the strength of the East Asian Winter Monsoon was less influential.

The cooling effect of the ocean on the Southern California coastal zone is investigated using a high-resolution (4-km) gridded surface meteorological data set (gridMET) of daily maximum temperature (Tmax), with focus on summer mean conditions, taken as the July–August–September (JAS) average. An empirical orthogonal function analysis reveals a coastal mode of JAS temperature covariability, distinct from a more energetic inland mode, that captures Tmax averaged across the Southern California coastal plain. The coastal mode temperature correlates significantly with, and has similar amplitude to, regional sea surface temperature (SST). High (low) summer land and sea surface temperatures, as well as inversion layer temperature differences, are associated with decreases (increases) of northerly coastal wind speeds and coastal cloudiness. The number of extreme heat days on land increases as regional SST increases (4.3 days °C−1), with heat wave days 10 times more likely during peak warm versus cool coastal mode years. The coastal zone was notably warmer and heat wave days peaked during the well documented marine heat wave events of 2014/15 and 2018 off Southern California. The marine variability associated with the coastal mode also has strong expression off the Baja California peninsula, presumably due to strong covarying winds in that area. As in previous studies, higher ocean temperatures are attributed to weaker summer winds, with associated reductions in ocean surface heat loss, coastal upwelling, and cloudiness.

Black carbon in snow (BCS) is a crucial parameter in Earth System modeling, as it influences global radiative balance. Here, simulated BCS from Coupled Model Intercomparison Project Phase 5 and 6 (CMIP5 and CMIP6) that provided BCS as a model output were evaluated. In comparison with global BCS observations, CMIP5/6 models successfully reproduced long-term historical trends linked to human activities, but struggled capturing decadal variability caused by natural climate variability. CMIP6 models NorESM2-MM, NorESM2-LM, and TaiESM1 yielded the most accurate simulations of BCS concentration with modest overestimation of <50%, while the four CESM2 models underestimated concentrations by up to ∼80%. These errors effectively balanced for the CMIP6 multi-model ensemble mean (MME), which had a relative error (RE) of −37%. However, the CMIP5 MME was less reliable due to extreme overestimation by up to 8,000% in the three MIROC models. The significant BCS concentration errors in the MIROC and CESM2 models were linked mainly to errors in handling of BC in snow processes. Conversely, marked improvements in NorESM, the only common to both CMIP5 and CMIP6, were due to improved simulation of black carbon deposition. BCS errors significantly impacted radiative forcing estimates, particularly at the poles, where model errors reached several thousandfold. CMIP6 exhibited superior results compared to CMIP5, achieving global MME RE of −33% in radiative forcing estimates. However, it's worth noting BCS output is currently limited, with only seven models available for each of CMIP5 and CMIP6 here. Additional models simulating BCS are desirable in the next CMIP generations.

Accurate initialization of CONUS convective-scale forecasting requires a proper estimate of all resolved scales. This study further develops and examines a simultaneous multiscale data assimilation (MDA) approach in EnVar with modulated cross-scale and cross-variable covariances. The method is examined using 10 retrospective cases with the assimilation of both in situ and radar reflectivity observations (hereafter, SimMDA). The necessity of the modulated and therefore weakened cross-covariances in simultaneous MDA for CONUS convective-scale forecasting is first demonstrated. The relative benefits of increasing the decomposed-scale number with increased computational cost in SimMDA are also discussed. The impact of the further developed simultaneous MDA method is revealed by comparing it with a commonly adopted DA approach (Baseline), which separately assimilates in situ and reflectivity observations using individual single-scale localization. During DA cycling, SimMDA improves analysis accuracy for temperature and reflectivity and reduces biases in all variables compared to Baseline. SimMDA yields significantly better forecasts than Baseline for most lead times. Additional experiments are conducted to attribute such improvements in a case study. Specifically, an experiment the same as Baseline except using simultaneous MDA for reflectivity assimilation enhances cold pools and inflows and thus improves storms by making larger-scale increments. An experiment the same as Baseline except using simultaneous MDA for in situ assimilation more properly constrains small-scale covariances, leading to more reasonable correlations along the front and more accurate moisture near the dryline and consequently improved analyses and forecasts. Both effects together largely contribute to the overall improvements of SimMDA compared to Baseline.

The geopotential height observations from the Aura Microwave Limb Sounder show that the quasi-6-day wave (Q6DW) events with westward zonal wavenumber 1 (W1) in the Northern Hemisphere (NH) mesosphere and lower thermosphere (MLT) during September 2013 and 2014 had a prolonged lifetime of ∼45–50 days and reached their maximum amplitudes after the September equinox, while the climatological Q6DW-W1 is completely dissipated in the background atmosphere before the equinoxes. The Eliassen-Palm flux diagnostic results indicate that Q6DW-W1 during September 2013 and 2014 obtained an additional source at ∼60°-80°S and ∼40–50 km as the September equinox approached, which is related to the baroclinic/barotropic instability in this region. Further investigation on the background atmosphere reveals that several stratospheric minor warming (SMW) events occurred in the Antarctic region during September 2013 and 2014 due to the enhancement of wavenumber 1 activities, accompanied by the increase in the stratospheric temperature, changes in the shape of the polar vortex and the reversal in the zonal mean circulation. The strong planetary wave breaking during the September 2013 and 2014 Antarctic SMW events significantly weakened the strength of the polar night jet (PNJ) and made its peak height descend below ∼35 km, which generated the baroclinic/barotropic instability at the new upper boundary of the PNJ (∼60°-80°S and ∼40–50 km) by an anomalous double-jet configuration in the background winds. This unusual instability provided additional wave source and energy for the trans-equatorial propagation of Q6DW-W1, which finally led to prolonged wave activities in the NH MLT region.

Aerosol single-scattering albedo (SSA) is the most critical factor for the accurately calculating of aerosol radiative effects, however, the observation of vertical profiles of SSA is difficult to realize. Current assessments of aerosol radiative effects remain uncertain because of the lack of long-term, high-resolution vertical profiles of SSA observations. High-resolution SSA vertical profiles were observed in a semi-arid region of Northwest China during winter using a tethered balloon. The observed SSA vertical profiles were used to calculate the aerosol direct radiative forcing and radiative heating rates. Significant differences in the calculated radiative forcing were found (e.g., a 48.3% relative difference for the heating effect in the atmosphere at 14:00) between the observed SSA profiles and the constant assumption with SSA = 0.90. Diurnal variations in the vertical distribution of SSA decisively influenced direct radiative forcing of aerosols. Furthermore, high-resolution vertical profiles of absorbing aerosols and meteorological parameters provide robust observational evidence of the heating effect of an elevated absorbing aerosol layer. This study provides a more accurate calculation of aerosol radiative forcing using observed aerosol SSA profiles.

Emission sources influencing high particulate air pollution levels and related mortality in India have been studied earlier on country-wide and sub-national scales. Here, we use novel data sets of emissions (for 2019) and observations created under the Carbonaceous Aerosol Emissions, Source Apportionment, and Climate Impacts network in India (Venkataraman et al., 2020, https://doi.org/10.1175/bams-d-19-0030.1) in WRF-Chem simulations to evaluate drivers of high PM2.5 levels during episodes and in airsheds with different pollution levels. We identify airsheds in “extreme” (110–140 μg/m3), “severe” (80–110 μg/m3) and “significant” (40–80 μg/m3) exceedance of the Indian annual ambient air quality standard (National Ambient Air Quality Standards [NAAQS]) of 40 μg/m3 for PM2.5. We find that primary organic matter and anthropogenic mineral matter (largely coal fly-ash) drive high PM2.5 levels, both annually and during high PM2.5 episodes. PM2.5 episodes are driven by organic aerosol in north India (Mohali) in wintertime but are additionally influenced by mineral matter and secondary inorganics in central (Bhopal), south India (Mysuru) and eastern India (Shyamnagar). Across airsheds in exceedance of the NAAQS and during high PM2.5 episodes, primary PM2.5 emissions arise largely from the residential sector (50%–75%). Formal sector emissions (industry, thermal power and transport; 40%–55%) drive airshed and episode scale PM2.5 exceedance in northern and eastern India. Agricultural residue burning emissions predominate (50%–75%) on episode scales, both in northern and central India, but not on annual scales. Interestingly, residential sector emissions strongly influence (60%–90%) airsheds in compliance with the NAAQS (annual mean PM2.5 < 40 μg/m3), implying the need for modern residential energy transitions for the reduction of ambient air pollution across India.

Regional weather variability and extremes over Europe are strongly linked to variations in the North Atlantic jet stream, especially during the winter season. Projections of the evolution of the North Atlantic jet are essential for estimating the regional impacts of climate change. Therefore, separating forced trends in the North Atlantic jet from its natural variability is an extremely relevant task. Here, a deep learning based method, the Latent Linear Adjustment Autoencoder (LLAE), is used for this purpose on an ensemble of fully-coupled climate simulations. The LLAE is based on a variational autoencoder and an additional linear component. The model uses detrended temperature and geopotential to predict the component of the zonal wind associated with natural variability. The residual between this prediction and the original wind field is interpreted as the forced component of the jet. The method is first tested for the geostrophic wind for which the forced trend can be obtained analytically from the difference between geostrophic wind computed from detrended and full geopotential. Despite the large variability of the total trends, the LLAE is shown to be effective in extracting the forced component of the trend for each individual ensemble member in both geostrophic and full wind fields. The LLAE-derived forced trend shows an increase in the upper-level zonal wind speed along a southwest–northeast oriented band over the ocean and a jet extension toward Europe. These are common characteristics over different periods and show some similarities to the upper-level zonal wind speed trend obtained from the ERA5 reanalysis.

In ensemble weather forecast, tropical cyclone (TC) tracks sometimes group together into trajectories parting away from each other. The goal of this study is to propose an objective method, based on a robust clustering approach, to detect such separation scenarios in the Japan Meteorological Agency Meso-scale Ensemble Prediction System (MEPS) for three TCs: “Dolphin” (2020), “Nepartak” (2021), and “Meari” (2022). Taking advantage of the independence of the density-based spatial clustering of applications with noise algorithm to the prior choice of the number of clusters, we first describe an objective way to calculate the aggregation distance, by searching the most frequent Euclidean distance between all the tracks. The clustering is then applied to the forecasted tracks, for each initialization time of MEPS (every 6 hr). Separation scenarios exist when the number of clusters is greater than one.

The characteristics of gravity waves in China are investigated through an extensive analysis of high vertical resolution radiosonde observations collected over eight years across 120 stations, and are subsequently compared to those in the United States. These characteristics encompass energy density, intrinsic frequencies, horizontal and vertical wavelengths, as well as vertical and horizonal propagation directions. China and the United States, situated in mid-latitude regions with prominent western topographical features, the Qinghai-Tibet Plateau and Rocky Mountains respectively, demonstrate striking similarities in the generation and distribution of gravity waves. Both landmasses exhibit the strongest gravity waves during winter and the weakest during summer. And within the troposphere, the maximum energy of gravity waves is generated over and immediately downstream of the topographies. In addition, the energy level is amplified in the lower stratosphere. However, unique regional contrasts in summer are result from the differences of summer monsoon influence and the distinct western topographies. The maximum gravity wave energy in summer troposphere is observed over the north side of the Qinghai-Tibet Plateau in China, contrasting with its location downstream of the Rockies in the United States.

As China's land-based anthropogenic emissions are decreasing, the impact of marine shipping emissions (MSEs) on the atmosphere, especially in coastal areas, deserves further attention. This study investigates the impact of MSEs on MDA8 ozone (O3) levels during the warm seasons of 2017 in China, considering different seasons and synoptic patterns. The results indicate that the average impact of MSEs on O3 decreases from offshore to inland, peaking at over 29.0 ppb at sea and 13.8 ppb along the coast of mainland China. Influenced by precursor emissions, meteorology and other factors, MSEs contribute differently to O3 in Beijing-Tianjin-Hebei (BTH), Yangtze River Delta (YRD) and Pearl River Delta (PRD), with contributions of 3.0, 5.2, and 4.9 ppb, respectively, and ranging from 2.7 to 7.3 ppb in 13 coastal port cities. The O3 impacts of MSEs are higher on polluted days than on clean days, especially during onshore winds. In the BTH, MSEs increase O3 by 5.5 ppb on polluted days and 3.0 ppb on clean days with northeast winds from the Bohai Sea. In the YRD, MSEs increase O3 by 9.4 ppb on polluted days and 7.3 ppb on clean days with southeast winds. MSEs significantly increase O3 levels in the PRD by 11.0 ppb on polluted days and 5.0 ppb on clean days with southeast winds. Although the emission inventories, initial and boundary conditions, etc. may introduce uncertainties, our results still provide useful information for O3 pollution management in coastal cities as a reasonable way to track mass contributions.

The influence of biogeochemical cycles, particularly the nitrogen cycle, on near-surface meteorological fields is a critical yet understudied aspect of regional climate modeling. Neglecting such interactions may compromise the accurate representation of vegetation growth and hydrological processes in climate models, consequently affecting the simulated regional near-surface climate conditions. In order to quantify such effects, we coupled the nitrogen-augmented Noah-MP land surface model with the Weather Research and Forecasting (WRF) model v4.1.2 (hereafter WRF-CN) for regional climate modeling. Compared to the default WRF simulation without nitrogen dynamics, the WRF-CN simulated net primary productivity, gross primary productivity (GPP), and leaf area index (LAI) were all higher in the study region. Because WRF underestimated the observed GPP and LAI due to the fixed nitrogen limitation of plant growth, these higher estimations improved WRF-CN's performance in modeling GPP and LAI, which translated into improved simulations of near-surface climate. Specifically, for the 2-m air temperature, compared to WRF, WRF-CN reduced the mean absolute error and root mean square error by 14.45% and 14.19%, respectively, while increased the Nash-Sutcliffe efficiency coefficient by 7.23%, with the most pronounced improvements in the regions dominated by croplands. Our findings shed light on the crucial interactions between biogeochemical processes and near-surface meteorological conditions, emphasizing the significance of incorporating terrestrial nitrogen dynamics in regional climate models. These insights contribute to advancing our understanding of climate system dynamics and improving the accuracy of climate predictions at the mesoscale.

Clouds can be classified into regimes based on their appearance or meteorological controlling factors. The cloud appearance regimes inherently include adjustments to aerosol effects, such as transitions between closed and open cells. Therefore, calculating cloud susceptibilities to aerosols for each cloud-appearance regime individually and then aggregating them excludes much of the cloud adjustment component of the susceptibilities. In contrast, aggregating susceptibilities over regimes defined by cloud-controlling factors includes the full effects of cloud adjustments. Here we compared the susceptibilities of the two kinds of cloud regimes and demonstrated this effect. Overall, increasing cloud droplet number concentration (N d ) consistently correlates to weaker precipitation, higher cloud fraction (CF), and reduced liquid water path, regardless of how the regime is defined. However, their susceptibilities to N d aggregated over cloud-appearance regimes are significantly lower than those aggregated over cloud-controlling factors regimes, with lower-tropospheric stability (LTS) serving as an example to define cloud-controlling factors regimes. This underestimation is more pronounced for CF susceptibility, where the susceptibility for cloud appearance regimes is only 1/4 of the susceptibility for cloud controlling regimes. These findings imply that relying solely on cloud-appearance regimes may underestimate the effective radiative forcing produced by cloud adjustment (ERFaci). Nevertheless, the substantial variability in the magnitude of cloud adjustment across appearance regimes at similar LTS also suggests that a single cloud-controlling factor is not sufficient to fully separate cloud regimes to quantify cloud adjustment. Therefore, identifying a comprehensive set of cloud-controlling factors is essential for accurately quantifying cloud adjustments in future studies.

Snow is the most reflective natural surface on Earth. Since fresh snow on bare sea ice increases the surface albedo, the impact of summer snow accumulation can have a negative radiative forcing effect, which would inhibit sea ice surface melt and potentially slow sea-ice loss. However, it is not well known how often, where, and when summer snowfall events occur on Arctic sea ice. In this study, we used in situ and model snow depth data paired with surface albedo and atmospheric conditions from satellite retrievals to characterize summer snow accumulation on Arctic sea ice from 2003 to 2017. We found that, across the Arctic, ∼2 snow accumulation events occurred on initially snow-free conditions each year. The average snow depth and albedo increases were ∼2 cm and 0.08, respectively. 16.5% of the snow accumulation events were optically thick (>3 cm deep) and lasted 2.9 days longer than the average snow accumulation event (3.4 days). Based on a simple, multiple scattering radiative transfer model, we estimated a −0.086 ± 0.020 W m−2 change in the annual average top-of-the-atmosphere radiative forcing for summer snowfall events in 2003–2017. The following work provides new information on the frequency, distribution, and duration of observed snow accumulation events over Arctic sea ice in summer. Such results may be particularly useful in understanding the impacts of ephemeral summer weather on surface albedo and their propagating effects on the radiative forcing over Arctic sea ice, as well as assessing climate model simulations of summer atmosphere-ice processes.

The Antarctic Oscillation (AAO), which is the main mode of extratropical circulation in the Southern Hemisphere, also has a substantial effect on the Northern Hemisphere climate. We investigated the influence of the early AAO on the frequency of late severe pollution events (SPEF) in the Beijing–Tianjin–Hebei region (SPEFBTH) of China during winter. The results show that the winter (December–January–February) SPEFBTH is negatively correlated with the AAO from the previous autumn (August–September–October). The controlling mechanism can be briefly described as follows: the autumn AAO is positively correlated with the mid-latitude sea surface temperature (SST) in the South Atlantic Ocean. The SST preserves the autumn anomaly signal into the following winter. This anomalous SST regulates changes in the tropical western Indian Ocean–Intertropical Convergence Zone (IN–ITCZ) via air–sea coupling. Subsequently, as a response to the IN–ITCZ anomalies, anomalous wave trains are excited in the upper troposphere from the tropical western Indian Ocean to East Asia. In addition, the local meridional circulation is modulated; therefore, the circulation field and other meteorological elements favorable for the SPEFBTH appear and exacerbate the SPEFBTH. This study describes a new physical mechanism for the pathway of the AAO influence on subsequent SPEFBTH and finds a predictable source in the Southern Hemisphere air–sea system.

Tropospheric ozone (O3) is an important greenhouse gas that is also hazardous to human health. The formation of O3 is sensitive to the levels of its precursors NOx (≡NO + NO2) and peroxy radicals, for example, generated by the oxidation of volatile organic compounds (VOCs). A better understanding of this sensitivity will show how changes in the levels of these trace gases could affect O3 levels today and in the future, and thus air quality and climate. In this study, we investigate O3 sensitivity in the tropical troposphere based on in situ observations of NO, HO2 and O3 from four research aircraft campaigns between 2015 and 2023. These are OMO (Oxidation Mechanism Observations), ATom (Atmospheric Tomography Mission), CAFE Africa (Chemistry of the Atmosphere Field Experiment in Africa) and CAFE Brazil, in combination with simulations using the EMAC atmospheric chemistry—climate model. We use the metric α(CH3O2) together with NO to investigate the O3 formation sensitivity. We show that O3 formation is generally NOx-sensitive in the lower and middle tropical troposphere and is in a transition regime in the upper troposphere. By distinguishing observations impacted by lightning or not we show that NO from lightning is the most important driver of O3 sensitivity in the tropics. NOx-sensitive chemistry predominates in regions without lightning impact, with α(CH3O2) ranging between 0.56 and 0.82 and observed average O3 levels between 35 and 55 ppbv. Areas affected by lightning exhibit strongly VOC-sensitive O3 chemistry with α(CH3O2) of about 1 and average O3 levels between 55 and 80 ppbv.

Passive microwave radiometers onboard satellites rely on the received upwelling radiation to retrieve precipitation, which is a mixed signal from the surface, atmosphere and precipitation hydrometeors. Liquid precipitation droplets increase the upwelling radiation from the surface at lower frequencies, while ice particles cause a decrease in upwelling radiation at higher frequencies. The task of the retrieval algorithm is to identify the precipitation phase and to isolate the signal of precipitation from that of the surface. This study develops a machine learning method to retrieve rainfall and snowfall rates based on observations from the Microwave Hydrometer Sounder and Microwave Temperature Sounder onboard FY-3E. Self-organized mapping (SOM) is selected to classify the precipitation and underlying surface types, and an artificial neural network (ANN) is subsequently used to relate the brightness temperature to the precipitation rate for the clusters derived from the SOM. The half-hour product of the Integrated Multi-Satellite Retrieval for Global Precipitation Measurement (IMERG) is used to train the ANN. To address the issue that number of heavy precipitation samples are not enough for training, the simulation of radiative transfer for TOVS is used as a supplement to heavy rain samples. The SOM-ANN algorithm outperforms the IMERG and Goddard profiling algorithm (GPROF) retrieval products in both rainfall and snowfall retrieval. Compared with the hourly observations at ∼4,400 stations during a 2-year period, the root mean square errors of SOM-ANN proposed here are 1.06 and 0.34 mm/hr for the rainfall and snowfall rates, which are better than those of IMERG (1.23 and 0.42 mm/hr) and GPROF (1.22 and 0.44 mm/hr).