Feed aggregator

In this study, we investigate the representation of the global mean energy balance components in 10 atmospheric reanalyses, and compare their magnitudes with recent reference estimates as well as the ones simulated by the latest generation of climate models from the 6th phase of the coupled model intercomparison project (CMIP6). Despite the assimilation of comprehensive observational data in reanalyses, the spread amongst the magnitudes of their global energy balance components generally remains substantial, up to more than 20 Wm−2 in some quantities, and their consistency is typically not higher than amongst the much less observationally constrained CMIP6 models. Relative spreads are particularly large in the reanalysis global mean latent heat fluxes (exceeding 20%) and associated intensity of the global water cycle, as well as in the energy imbalances at the top-of-atmosphere and surface. A comparison of reanalysis runs in full assimilation mode with corresponding runs constrained only by sea surface temperatures reveals marginal differences in their global mean energy balance components. This indicates that discrepancies in the global energy balance components caused by the different model formulations amongst the reanalyses are hardly alleviated by the imposed observational constraints from the assimilation process. Similar to climate models, reanalyses overestimate the global mean surface downward shortwave radiation and underestimate the surface downward longwave radiation by 3–7 Wm−2. While reanalyses are of tremendous value as references for many atmospheric parameters, they currently may not be suited to serve as references for the magnitudes of the global mean energy balance components.

Earth’s energy imbalance (EEI) is a fundamental metric of global Earth system change, quantifying the cumulative impact of natural and anthropogenic radiative forcings and feedback. To date, the most precise measurements of EEI change are obtained through radiometric observations at the top of the atmosphere (TOA), while the quantification of EEI absolute magnitude is facilitated through heat inventory analysis, where ~ 90% of heat uptake manifests as an increase in ocean heat content (OHC). Various international groups provide OHC datasets derived from in situ and satellite observations, as well as from reanalyses ingesting many available observations. The WCRP formed the GEWEX-EEI Assessment Working Group to better understand discrepancies, uncertainties and reconcile current knowledge of EEI magnitude, variability and trends. Here, 21 OHC datasets and ocean heat uptake (OHU) rates are intercompared, providing OHU estimates ranging between 0.40 ± 0.12 and 0.96 ± 0.08 W m−2 (2005–2019), a spread that is slightly reduced when unequal ocean sampling is accounted for, and that is largely attributable to differing source data, mapping methods and quality control procedures. The rate of increase in OHU varies substantially between − 0.03 ± 0.13 (reanalysis product) and 1.1 ± 0.6 W m−2 dec−1 (satellite product). Products that either more regularly observe (satellites) or fill in situ data-sparse regions based on additional physical knowledge (some reanalysis and hybrid products) tend to track radiometric EEI variability better than purely in situ-based OHC products. This paper also examines zonal trends in TOA radiative fluxes and the impact of data gaps on trend estimates. The GEWEX-EEI community aims to refine their assessment studies, to forge a path toward best practices, e.g., in uncertainty quantification, and to formulate recommendations for future activities.

Carrier phase integer ambiguities must be determined for BDS-3/inertial navigation system (INS) tightly coupled (TC) integration to achieve centimetre-level positioning accuracy. However, cycle slip breaks the consistency of the integer ambiguities. Conventional multi-frequency cycle slip methods use the pseudorange; thus, requiring improvement when applied to kinematic situations. Furthermore, a concise and nonprior information-dependent model is crucial for real-time processing. In this study, an inertial-aided BDS-3 cycle slip detection and repair (I-CDR) method was developed. First, a BDS-3/INS TC model with I-CDR was created. The ionospheric delays were modelled as part of the TC states; therefore, they could be estimated and eliminated. Investigations were conducted on the effects of carrier phase noise, residual ionosphere delay, and INS-predicted position error on combined cycle slip detection (CCD) accuracy. The optimal CCDs under various frequency available configurations were determined. The effectiveness of I-CDR was demonstrated using land vehicle test data. The false alarm ratio was less than 1.0%, and the missed detection ratio was almost zero even in situations with challenging abundant 1-cycle slips in random epochs. Furthermore, the right determination ratio reached 100%. In addition, BDS-3 signal loss-recovery cases were simulated, and all cycle slips for all satellites could be repaired within 40s. I-CDR exhibits outstanding cycle slip detection and repair performance for dense 1-cycle slip and signal loss-recovery cases, demonstrating its suitability for BDS-3/INS TC integration.

This paper focuses on the retrieval of refractivity fields from GNSS measurements by means of least-squares collocation. Collocation adjustment estimates parameters that relate delays and refractivity without relying on a grid. It contains functional and stochastic models that define the characteristics of the retrieved refractivity fields. This work aims at emphasizing the capabilities and limitations of the collocation method in modeling refractivity and to present it as a valuable alternative to GNSS tomography. Initially, we analyze the stochastic models in collocation and compare the theoretical errors of collocation with those of tomography. We emphasize the low variability of collocation formal variances/covariances compared to tomography and its lower dependence on a-priori fields. Then, based on real and simulated data, we investigate the importance of station resolution and station heights for collocation. Increasing the network resolution, for example, from 10 to 2 km, results in improved a-posteriori statistics, including a 10% reduction in the error statistic for the retrieved refractivity up to 6 km. In addition, using additional stations at higher altitudes has an impact on the retrieved refractivity fields of about 1 ppm in terms of standard deviation up to 6 km, and a bias reduction of more than 3 ppm up to 3 km. Furthermore, we compare refractivity fields retrieved through tomography and collocation, where data of the COSMO weather model are utilized in a closed-loop validation mode to simulate tropospheric delays and validate the retrieved profiles. While tomography estimates are less biased, collocation captures relative changes in refractivity more effectively among the voxels within one height level. Finally, we apply tomography and collocation to test their capabilities to detect an approaching weather front. Both methods can sense the weather front, but their atmospheric structures appear more similar when the GNSS network has a well-distributed height coverage.

Deformation phenomena on Earth are inherently three dimensional. With SAR interferometry (InSAR), in many practical situations the maximum number of observations is two (ascending and descending), resulting in an infinite number of possible displacement estimates. Here we propose a practical solution to this underdeterminancy problem in the form of the strapdown approach. With the strapdown approach, it is possible to obtain “3D-global/2D-local” solutions, by using minimal and largely undisputed contextual information, on the expected driving mechanisms and/or spatial geometry. It is a generic method that defines a local reference system with transversal, longitudinal, and normal (TLN) axes, with displacement occurring in the transversal-normal plane only. Since the orientation of the local frame is based on the physics of the problem at hand, the strapdown approach gives physically more relevant estimates compared to conventional approaches. Moreover, using an a-priori uncertainty approximation on the orientation of the local frame it is possible to assess the precision of the final estimates. As a result, appropriate cartographic visualization using a vector map with confidence ellipses enables an improved interpretation of the results.

Accurate information on satellite antenna phase center offsets (PCOs) and phase variations (PVs) is indispensable for high-precision geodetic applications. In the absence of consistent pre-flight calibrations, satellite antenna PCOs and PVs of global navigation satellite systems are commonly estimated based on observations from a global network, constraining the scale to a given reference frame. As part of this estimation, flatness and zero-mean conditions need to be applied to unambiguously separate PCOs, PVs, and constant phase ambiguities. Within this study, we analytically investigate the impact of different boresight-angle-dependent weighting functions for PV minimization, and we compare antenna models generated with different observation-based weighting schemes with those based on uniform weighting. For the case of the GPS IIR/-M and III satellites, systematic differences of 10 mm in the PVs and 65 cm in the corresponding PCOs are identified. In addition, new antenna models for the different blocks of BeiDou-3 satellites in medium Earth orbit are derived using different processing schemes. As a drawback of traditional approaches estimating PCOs and PVs consecutively in distinct steps, it is shown that different, albeit self-consistent, PCO/PV pairs may result depending on whether PCOs or PVs are estimated first. This apparent discrepancy can be attributed to potentially inconsistent weighting functions in the individual processing steps. Use of a single-step process is therefore proposed, in which a dedicated constraint for PCO-PV separation is applied in the solution of the normal equations. Finally, the impact of neglecting phase patterns in precise point positioning applications is investigated. In addition to an overall increase of the position scatter, the occurrence of systematic height biases is illustrated. While observation-based weighting in the pattern estimation can help to avoid such biases, the possible benefit depends critically on the specific elevation-dependent weighting applied in the user’s positioning model. As such, the practical advantage of such antenna models would remain limited, and uniform weighting is recommended as a lean and transparent approach for the pattern estimation of satellite antenna models from observations.

An attractive feature of PPP-RTK is the possibility of reducing the amount of data that needs to be transferred to users. By leveraging the state-space Representation (SSR) of the corrections, the correction provider (i.e., a GNSS network) can consider distinct transfer rates for each of the individual corrections according to their temporal characteristics. Reducing the transfer rates comes at the cost of delivering time-delayed corrections, urging the user to time predict the corrections to bridge the gap between the corrections’ generation time and the positioning time. Consequently, the user Kalman filter needs to be equipped with a strategy to account for the errors caused by such predictions, minimizing the precision loss of the user parameter solutions. In this contribution, we apply a processing strategy for both the network and user filters to handle the latency of corrections. This enables the network to update corrections over longer time-intervals. To have the strategy applicable to regional networks, an ionosphere-weighted model is adopted for the corresponding observations, delivering minimum-variance spatially predicted ionospheric corrections to users. It is shown that certain components of the network filter’s dynamic model are duplicated and should be excluded from processing. To illustrate the performance of the strategy at work, three globally distributed regional networks are employed, and maximum correction latencies to meet different positioning criteria are evaluated. In terms of both the positioning precision and time-to-first-fix (TTFF), the strategy is numerically shown to outperform the user processing case in which the uncertainty of corrections is discarded.

Alongside magnetic reconnection, turbulence is another fundamental nonlinear plasma phenomenon that plays a key role in energy transport and conversion in space and astrophysical plasmas. From a numerical, theoretical, and observational point of view there is a long history of exploring the interplay between these two phenomena in space plasma environments; however, recent high-resolution, multi-spacecraft observations have ushered in a new era of understanding this complex topic. The interplay between reconnection and turbulence is both complex and multifaceted, and can be viewed through a number of different interrelated lenses - including turbulence acting to generate current sheets that undergo magnetic reconnection (turbulence-driven reconnection), magnetic reconnection driving turbulent dynamics in an environment (reconnection-driven turbulence) or acting as an intermediate step in the excitation of turbulence, and the random diffusive/dispersive nature of the magnetic field lines embedded in turbulent fluctuations enabling so-called stochastic reconnection. In this paper, we review the current state of knowledge on these different facets of the interplay between turbulence and reconnection in the context of collisionless plasmas, such as those found in many near-Earth astrophysical environments, from a theoretical, numerical, and observational perspective. Particular focus is given to several key regions in Earth’s magnetosphere – namely, Earth’s magnetosheath, magnetotail, and Kelvin-Helmholtz vortices on the magnetopause flanks – where NASA’s Magnetospheric Multiscale mission has been providing new insights into the topic.

We propose a benchmark for comparing gap-filling techniques used on global time-variable gravity field time-series. The Gravity Recovery and Climate Experiment (GRACE) and the GRACE Follow-On missions provide products to study the Earth’s time-variable gravity field. However, the presence of missing months in the measurements poses challenges for understanding specific Earth processes through the gravity field. We reproduce, adapt, and compare satellite-monitoring and interpolation techniques for filling these missing months in GRACE and GRACE Follow-On products on a global scale. Satellite-monitoring techniques utilize solutions from Swarm and satellite laser ranging, while interpolation techniques rely on GRACE and/or Swarm solutions. We assess a wide range of interpolation techniques, including least-squares fitting, principal component analysis, singular spectrum analysis, multichannel singular spectrum analysis, auto-regressive models, and the incorporation of prior data in these techniques. To inter-compare these techniques, we employ a remove-and-restore approach, removing existing GRACE products and predicting missing months using interpolation techniques. We provide detailed comparisons of the techniques and discuss their strengths and limitations. The auto-regressive interpolation technique delivers the best score according to our evaluation metric. The interpolation based on a least-squares fitting of constant, trend, annual, and semi-annual cycles offers a simple and effective prediction with a good score. Through this assessment, we establish a starting benchmark for gap-filling techniques in Earth’s time-variable gravity field analysis.

Inverting fault geometry and slip distribution simultaneously with geodetic observations based on Bayesian theory is becoming increasingly prevalent. A widely used approach, proposed by (Fukuda and Johnson, Geophys J Int 181:1441–1458, 2010) (F-J method), employs the least-squares method to solve the linear parameters of slip distribution after sampling the nonlinear parameters, including fault geometry, data weights and smoothing factor. Here, we present a modified version of the F-J method (MF-J method), which treats data weights and the smoothing factor as hyperparameters not directly linked to surface deformation. Additionally, we introduce the variance component estimation (VCE) method to resolve these hyperparameters. To validate the effectiveness of the MF-J method, we conducted inversion tests using both synthetic data and a real earthquake case. In our comparison of the MF-J and F-J methods using synthetic experiments, we found that the F-J method's inversion results for fault geometry were highly sensitive to the initial values and step sizes of hyperparameters, whereas the MF-J method exhibited greater robustness and stability. The MF-J method also exhibited a higher and more stable acceptance rate, enabling convergence to simulated values and ensuring greater accuracy of the parameter estimation. Furthermore, treating the fault length and width as unknown parameters and solving them simultaneously with other fault geometry parameters and hyperparameters using the MF-J method successfully resolved the issue of non-uniqueness in fault location solutions caused by the excessively large no-slip areas. In the 2017 Mw 7.3 Sarpol-e Zahab earthquake case study, the MF-J method produced a fault slip distribution with low uncertainty that accurately fit surface observation data, aligning with results from other research institutions. This validated the method's applicability and robustness in real-world scenarios. Additionally, we inferred that the second slip asperity was caused by early afterslip.

Seismic and electromagnetic (EM) imaging are essential tools for characterizing velocity and conductivity. However, the separate inversion of seismic and EM data is challenging due to the noisy measurements, inadequate data collection, and reliance on prior information, consequently resulting in uncertainty and ambiguity of the solutions. Moreover, the two methods are different in sensitivity and spatial resolution, making it difficult to discover consistencies in the inverted models. Joint inversion of seismic and EM data takes advantage of both methods and significantly improves the imaging capability of subsurface structures. In this paper, we review various coupling strategies for the joint inversion of seismic and EM data and highlight the application advances from 1-D to 3-D inversion. Specifically, we investigate the integration of machine learning techniques to tackle ill-posed inverse problems and showcase their effectiveness in coupling. Following this, we construct a deep-learning-based joint inversion workflow and provide a synthetic test to demonstrate its superiority by applying an attention mechanism, which enhances the model’s capability to focus on specific features within the data. This study proves the potential of integrating artificial intelligence into joint inversion and understanding the deep Earth interior by incorporating multiple geophysical data.

NASA’s Europa Clipper mission is designed to provide a diversity of measurements to further our understanding of the potential habitability of this intriguing ocean world. The Europa mission’s Ultraviolet Spectrograph (Europa-UVS), built at the Southwest Research Institute (SwRI), is primarily a “plume finder” and tenuous atmosphere investigation. The science objectives of Europa-UVS are to: 1) Search for and characterize any current activity, notably plumes; and 2) Characterize the composition and sources of volatiles to identify the signatures of non-ice materials, including organic compounds, in the atmosphere and local space environment. Europa-UVS observes photons in the 55–206 nm wavelength range at moderate spectral and spatial resolution along a 7.5° slit composed of 7.3°×0.1° and 0.2°×0.2° contiguous sections. A variety of observational techniques including nadir pushbroom imaging, disk scans, stellar and solar occultations, Jupiter transit observations, and neutral cloud/plasma torus stares are employed to perform a comprehensive study of Europa’s atmosphere, plumes, surface, and local space environment. This paper describes the Europa-UVS investigation’s science plans, instrument details, concept of operations, and data formats in the context of the Europa Clipper mission’s primary habitability assessment goals.

This review article provides a synthesis and perspective on how weather and climate extreme events can play a role in influencing tipping elements and triggering tipping points in the Earth System. An example of a potential critical global tipping point, induced by climate extremes in an increasingly warmer climate, is Amazon rainforest dieback that could be driven by regional increases in droughts and exacerbated by fires, in addition to deforestation. A tipping element associated with the boreal forest might also be vulnerable to heat, drought and fire. An oceanic example is the potential collapse of the Atlantic meridional overturning circulation due to extreme variability in freshwater inputs, while marine heatwaves and high acidity extremes can lead to coral reef collapse. Extreme heat events may furthermore play an important role in ice sheet, glacier and permafrost stability. Regional severe extreme events could also lead to tipping in ecosystems, as well as in human systems, in response to climate drivers. However, substantial scientific uncertainty remains on mechanistic links between extreme events and tipping points. Earth observations are of high relevance to evaluate and constrain those links between extreme events and tipping elements, by determining conditions leading to delayed recovery with a potential for tipping in the atmosphere, on land, in vegetation, and in the ocean. In the subsurface ocean, there is a lack of consistent, synoptic and high frequency observations of changes in both ocean physics and biogeochemistry. This review article shows the importance of considering the interface between extreme events and tipping points, two topics usually addressed in isolation, and the need for continued monitoring to observe early warning signs and to evaluate Earth system response to extreme events as well as improving model skill in simulating extremes, compound extremes and tipping elements.

The 3D ionosphere structure is of interest in many fields such as radio frequency communication and global navigation satellite system (GNSS) applications. However, the limited temporal and spatial coverage of measurements poses a challenge for 3D electron density modeling. To overcome this challenge, we explore the use of kriging interpolation technique. The kriging interpolation is performed to obtain 3D representation of the ionosphere over electron density measurements retrieved by GNSS radio-occultation (RO) data. RO measurements are first reduced to “shape function,” the ratio of electron density to vertical total electron content (VTEC), aiming to create a background model. Then, the empirical residual semivariogram is analyzed for variation characteristics of the shape functions under different solar geomagnetic conditions. Finally, 3D kriging is adopted for shape function interpolation. Compared to the modeling results without kriging, the maximum root mean square error (RMSE) reduction reaches \(3.4\times {10}^{-4}~\text {km}^{-1}\) , which amounts to \(3.4\times {10}^{11}~\text {el/m}^{3}\) of electron density when VTEC is assumed as 100 TECU. This improvement accounts for 17.8% of root mean square (RMS) of shape function.

The spherical radial basis function (SRBF) approach, widely used in gravity modeling, is theoretically surveyed from a viewpoint of random field theory. Let the gravity potential be a random field which is represented as an integral functional of another random field, namely an isotropic Gaussian random field (IGRF) on a sphere inside the Bjerhammar sphere with the SRBF as the integral kernel. When the integration is approximated by a discrete sum within a local region, one gets the widely applicable SRBF model. With this theoretical study, the following two findings are made. First, the IGRF implies a Gaussian prior on the spherical harmonic coefficients (SHCs) of the gravity potential; for this prior the SHCs are independent with each other and their variances are degree-only dependent. This should be reminiscent of two well-known priors, namely the power-law Kaula’s rule and the asymptotic power-law Tscherning-Rapp model. Second, the IGRF-SRBF representation is non-unique. Benefiting from this redundant representation, one can employ a simple IGRF, e.g., the simplest white field, and then design the SRBF accordingly to represent a potential with desired prior statistical properties. This can simplify the corresponding SRBF modeling significantly; to be more specific, the regularization matrix in parameter estimation of the SRBF modeling can be chosen to be a diagonal matrix, or even the naïve identity matrix.

In recent years, coseismic velocity from high-rate global navigation satellite systems (GNSS) carrier phase data has been widely utilized to estimate instrumental seismic intensity, thereby guiding earthquake early warning and emergency response. However, using carrier phase data only yields displacement, displacement increment, and average velocity but not instantaneous velocity at the epoch level. In large earthquakes, using average velocity over a brief time span (e.g., 1 s) to quantify instantaneous coseismic velocity is less reliable for recovering accurate deformation dynamics, especially for the near-field region. In this study, we first introduce GNSS raw Doppler-based instantaneous velocity into seismology, expanding carrier phase-based traditional GNSS seismology. We also propose a new integrated GNSS velocity estimation method that employs a Kalman filter to integrate raw Doppler-based instantaneous velocity and carrier phase-based average velocity. The GNSS data from shake table experiments and two real-world earthquake events (i.e., the 2016 Mw 6.6 Norcia earthquake and the 2011 Mw 9.1 Tohoku-oki earthquake) are used to investigate the impact of high-rate GNSS raw Doppler on capturing coseismic velocity waveforms and predicting instrumental seismic intensity. The simulated sine wave experiment results indicate that the accuracy of instantaneous and average velocity for the 1 Hz sampling rate case is 1.20 cm/s and 12.67 cm/s, respectively. A similar case holds for the simulated quake wave experiment. The retrospective analysis of the ultra-high-rate (20 Hz) GNSS data for the Norcia earthquake shows the average velocities exhibit more aliasing and have a smaller peak ground velocity value than instantaneous velocities in all cases (i.e., 1, 2, 4, 5, 10, and 20 Hz). For the 2011 Mw 9.1 Tohoku-oki earthquake, results show that incorporating raw Doppler data enhances the consistency between the GNSS intensity map and the United States Geological Survey intensity map for near-field regions. Therefore, high-rate GNSS RD data as it becomes more widely available should be incorporated into data processing of high-rate GNSS seismology to capture more accurate instantaneous coseismic velocity waveforms and predict more realistic instrumental seismic intensity in future analyses.

Global Navigation Satellite Systems (GNSS) applications require computation of the geometric range between the satellite vehicle at the time-of-signal transmission and the receiver antenna location at the time-of-signal reception. This computation requires attention to the frames of reference due to the rotation of the Earth-Centered Earth-Fixed (ECEF) frame during the time-of-signal propagation. Three range computation approaches are commonplace and will be discussed herein. The first is the Global Positioning System Interface Control Document recommendation to rotate the ECEF frames to a common reference time. The other two are forms of the Sagnac correction. The Sagnac derivations already in the literature are either limited to stationary receivers or lack the connection between the Earth-centered inertial (ECI) and ECEF frames. Neither form of the Sagnac correction exactly reproduces the geometric range. They are approximations. The literature does not currently contain an analysis of the error involved in using either form of the Sagnac correction. This article makes two contributions: (1) it presents derivations for both forms of the Sagnac correction that are valid for moving receivers and that maintain the connection between the ECI and ECEF frames; and (2) it analyzes the error of the Sagnac correction for orbits of different radius. The analysis shows that Sagnac corrections introduce range errors less than \(7.57\times 10^{-4}\) meters for GNSS satellites at medium Earth orbit.

We provide an overview of the isotopic signatures of presolar supernova grains, specifically focusing on 44Ti-containing grains with robustly inferred supernova origins and their implications for nucleosynthesis and mixing mechanisms in supernovae. Recent technique advancements have enabled the differentiation between radiogenic (from 44Ti decay) and nonradiogenic 44Ca excesses in presolar grains, made possible by enhanced spatial resolution of Ca-Ti isotope analyses with the Cameca NanoSIMS (Nano-scale Secondary Ion Mass Spectrometer) instrument. Within the context of presolar supernova grain data, we discuss (i) the production of 44Ti in supernovae and the impact of interstellar medium heterogeneities on the galactic chemical evolution of 44Ca/40Ca, (ii) the nucleosynthesis processes of neutron bursts and explosive H-burning in Type II supernovae, and (iii) challenges in identifying the progenitor supernovae for 54Cr-rich presolar nanospinel grains. Drawing on constraints and insights derived from presolar supernova grain data, we also provide an overview of our current understanding of the roles played by various supernova types – including Type II, Type Ia, and electron capture supernovae – in accounting for the diverse array of nucleosynthetic isotopic variations identified in bulk meteorites and meteoritic components. We briefly overview the potential mechanisms that have been proposed to explain these nucleosynthetic variations by describing the transport and distribution of presolar dust carriers in the protoplanetary disk. We highlight existing controversies in the interpretation of presolar grain data and meteoritic nucleosynthetic isotopic variations, while also outlining potential directions for future research.

Strong gravitational lensing at the galaxy scale is a valuable tool for various applications in astrophysics and cosmology. Some of the primary uses of galaxy-scale lensing are to study elliptical galaxies’ mass structure and evolution, constrain the stellar initial mass function, and measure cosmological parameters. Since the discovery of the first galaxy-scale lens in the 1980s, this field has made significant advancements in data quality and modeling techniques. In this review, we describe the most common methods for modeling lensing observables, especially imaging data, as they are the most accessible and informative source of lensing observables. We then summarize the primary findings from the literature on the astrophysical and cosmological applications of galaxy-scale lenses. We also discuss the current limitations of the data and methodologies and provide an outlook on the expected improvements in both areas in the near future.

Satellite remote sensing serves as a crucial means to acquire cloud physical parameters. However, existing official cloud products from the advanced geostationary radiation imager (AGRI) onboard the Fengyun-4A geostationary satellite lack spatiotemporal continuity and important micro-physical properties. In this study, an image-based transfer learning ResUnet (TL-ResUnet) model was applied to realize all-day and high-precision retrieval of cloud physical parameters from AGRI thermal infrared measurements. Combining the observation advantages of geostationary and polar-orbiting satellites, the TL-ResUnet model was pre-trained with official cloud products from advanced Himawari imager (AHI) and transfer-trained with official cloud products from moderate resolution imaging spectroradiometer (MODIS), respectively. For comparison, a pixel-based transfer learning random forest (TL-RF) model was trained using the equally distributed data sets. Taking MODIS official products as the benchmarks, the TL-ResUnet model achieved an overall accuracy of 79.82% for identifying cloud phase and root mean squared errors of 1.99 km, 7.11 μm, and 12.87 for estimating cloud top height, cloud effective radius, and cloud optical thickness, outperforming the precision of AGRI and AHI official products. Compared to the TL-RF model, the TL-ResUnet model utilized the spatial information of clouds to significantly improve the retrieval performance and achieve more than a 6-fold increase in speed for single full-disk retrieval. Moreover, AGRI TL-ResUnet products with spatiotemporal continuity and high precision were used to accurately describe the spatial distribution characteristics of cloud fractions and cloud properties over the Tibetan Plateau, and provide the diurnal variation of cloud cover and cloud properties across different seasons for the first time.