Large Ensemble Research Articles

GARD-LENS: A downscaled large ensemble dataset for understanding future climate and its uncertainties

This article introduces the Generalized Analog Regression Downscaling method Large Ensemble (GARD-LENS) dataset, comprised of daily precipitation, mean temperature, and temperature range over the Contiguous U.S., Alaska, and Hawaii at 12-km, 4-km, and 1-km resolutions, respectively. GARD-LENS statistically downscales three CMIP6 global climate model large ensembles, CESM2, CanESM5, and EC-Earth3, totaling 200 ensemble members. GARD-LENS is the first downscaled SMILE (single model initial-condition large ensemble), providing information about the role of internal climate variability at high resolutions. The 150-year record of this large ensemble dataset provides ample data for assessing trends and extremes and allows users to robustly assess internal variability, forced climate signals, and time of emergence at high resolutions. As the need for high resolution, robust climate datasets continues to grow, GARD-LENS will be a valuable tool for scientists and practitioners who wish to account for internal variability in their future climate analyses and adaptation plans.

Enhancing generalizability of data-driven urban flood models by incorporating contextual information

Abstract. Fast urban pluvial flood models are necessary for a range of applications, such as near real-time flood nowcasting or processing large rainfall ensembles for uncertainty analysis. Data-driven models can help overcome the long computational time of traditional flood simulation models, and the state-of-the-art models have shown promising accuracy. Yet the lack of generalizability of data-driven urban pluvial flood models to both unseen rainfall and distinctively different terrain, at the fine resolution required for urban flood mapping, still limits their application. These models usually adopt a patch-based framework to overcome multiple bottlenecks, such as data availability and computational and memory constraints. However, this approach does not incorporate contextual information of the terrain surrounding the small image patch (typically 256 m×256 m). We propose a new deep-learning model that maintains the high-resolution information of the local patch and incorporates a larger surrounding area to increase the visual field of the model with the aim of enhancing the generalizability of data-driven urban pluvial flood models. We trained and tested the model in the city of Zurich (Switzerland), at a spatial resolution of 1 m, for 1 h rainfall events at 5 min temporal resolution. We demonstrate that our model can faithfully represent flood depths for a wide range of rainfall events, with peak rainfall intensities ranging from 42.5 to 161.4 mm h−1. Then, we assessed the model's terrain generalizability in distinct urban settings, namely, Lucerne (Switzerland) and Singapore. The model accurately identifies locations of water accumulation, which constitutes an improvement compared to other deep-learning models. Using transfer learning, the model was successfully retrained in the new cities, requiring only a single rainfall event to adapt the model to new terrains while preserving adaptability across diverse rainfall conditions. Our results indicate that by incorporating contextual terrain information into the local patches, our proposed model effectively simulates high-resolution urban pluvial flood maps, demonstrating applicability across varied terrains and rainfall events.

The role of the Indian Ocean Dipole in modulating the austral spring ENSO teleconnection to the Southern Hemisphere

Abstract. The combined influence of the El Niño–Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) on the extratropical circulation in the Southern Hemisphere (SH) during austral spring is examined. Reanalyses and the large ensemble of National Centers for Environmental Prediction (NCEP) Climate Forecast System version 2 (CFSv2) model outputs were used to compute composites and linear regressions for relevant variables. The results show that a positive IOD can reinforce the El Niño-induced circulation by merging the Indian Ocean wave train with the Pacific South American (PSA) pattern over the Pacific Ocean. In addition, the results obtained with the CFSv2 model output shows that strong positive IODs can contribute to enhancing the circulation signal of the El Niño anomalies and the Indian Ocean wave train. On the other hand, negative IODs in combination with La Niña do not have that combined circulation response. While there is a moderate intensification of the circulation anomalies associated with La Niña, accompanied by some changes in the location of their main action centers, results vary considerably between linear regression, the observed composites, and model composites. Regarding the influence of the IOD activity (independent of ENSO), reanalysis-based results show that the IOD positive phase has a significant impact over the entire SH, while the negative phase is associated with weaker anomalies and less consistent atmospheric response.

Revisiting the “East African Paradox”: CMIP6 Models Also Struggle to Reproduce Strong Observed Long Rain Drying Trends

Abstract Societies in much of the Horn of Africa are affected by variability in two distinct rainy seasons: the March–May (MAM) “long” rains and the October–December (OND) “short” rains. A recent five-season, La Niña–forced drought has renewed concerns about possible anthropogenic drying trends in the long rains, which had partially recovered after a multidecadal drying trend in the 1980s through the 2000s. Despite observed drying, previous generations of global climate models (GCMs) have consistently projected long-term wetting due to increased greenhouse gas concentrations, an East African “paradox” which complicates the interpretation of East African rainfall projections. We investigate the paradox in new phase 6 of the Coupled Model Intercomparison Project (CMIP6) and seasonal forecast models, leveraging an improved observational record and large ensembles to better differentiate internal and forced trends. We find observed drying trends are at the limits of the GCM spread during the peak paradox period, though the recent recovery is comfortably within the model spread. We find that the apparent paradox is largely removed by prescribing sea surface temperatures (SSTs) and is likely caused by the GCM difficulties in simulating observed tropical Pacific SST trends in recent decades. In line with arguments that these SST trends are at least partially forced anthropogenically, we recommend users of future rainfall projections in East Africa consider the possibility of long-term MAM drying despite GCM wetting and call for future model simulations that better sample the expected spread of SSTs. Significance Statement Societies in the Horn of Africa depend on the March–May “long” rains and the September–December “short” rains to support agricultural and pastoral practices on rain-fed lands. Recent major droughts have raised worries about possible anthropogenically forced drying trends in the long rains through global climate models (GCMs) project wetting. This East African “paradox” complicates long-term climate adaptation planning. We find the paradox continues in the newest generation of GCMs and seasonal forecast models; though a recent recovery in rainfall is well-captured, the strongest observed drying trends are rare in simulations. The paradox likely arises from known GCM biases in the Pacific Ocean interacting with natural variability. We recommend that researchers and policymakers consider possible long-term drying despite GCM projections of wetting.

Linking projected changes in seasonal climate predictability and ENSO amplitude

Abstract Recent studies have shown that potential predictability and actual forecast skill have varied throughout the historical record, primarily due to natural decadal variability. In this study, we explore whether and how potential predictability are projected to change in the future as a distinct response to anthropogenic climate change. We estimate the potential predictability of the El Niño-Southern Oscillation (ENSO) as well as global surface temperature, precipitation, and upper atmospheric circulation anomalies from 1921-2100, within a perfect model framework, using five coupled model large ensembles. We find that historical and projected ENSO amplitude changes generate global scale shifts in climate predictability via ENSO-driven changes in the signal-to-noise ratio of seasonal forecasts, with a 10% change in Nino3.4 standard deviation leading to a 14% change in globally averaged forecast skill at 12-month lead. This relationship suggests that potential predictability changes across much of the globe in the coming decades could be linked to anthropogenic climate change of ENSO. However, since current models substantially disagree on the sign and intensity of projected ENSO change, the trajectory of future global predictability changes cannot yet be determined. This problem is demonstrated by widely varying predictability changes seen across the five large ensembles, with models exhibiting a robust increase, robust decrease, or no significant change in predictability, depending upon their respective projected ENSO amplitude trends. Our results highlight the need for climate model development aimed at better capturing past forced and unforced changes to ENSO variability, which is necessary (if not sufficient) to constrain projected changes to climate predictability worldwide.

Detailed Calculations of the Efficiency of Planetesimal Accretion in the Core-accretion Model. III. The Contribution of Planetesimals beyond Saturn

Continuing our initiative on advancing the calculations of planetesimal accretion in the core-accretion model, we present here the results of our recent study of the contributions of planetesimals around and beyond the orbit of Saturn. In our first two papers, where our focus was on the effects of the Sun and Saturn, the initial distribution of planetesimals was limited to the regions around the accretion zone of a growing Jupiter. In this paper, we expanded that distribution to regions beyond the accretion zone of Saturn. We integrated the orbits of a large ensemble of planetesimals and studied the rate of their capture by the growing proto-Jupiter. In order to be consistent with our previous studies, we did not consider the effect of the nebular gas. Results demonstrated that the exterior planetesimals, especially those beyond 8 au, have only slight contributions to the growth and metallicity of the growing Jupiter. The final mass and composition of this planet is mainly due to the planetesimals inside and around its accretion zone. Our study shows that although the rate of capture varies slightly by the size and composition of planetesimals, in general, the final results are independent of the size and material of these bodies. Results also pointed to a new finding: the rate of accretion follows the same trend as that of the evolution of Jupiter’s envelope, with the largest accretion occurring during the envelope’s collapse. We present details of our analysis and discuss the implications of its results.

Future Projection of Extremely Heavy Rainfall in the Tohoku District of Japan with Large Ensemble Simulations Using the 5 km Regional Climate Model

Future Projection of Extremely Heavy Rainfall in the Tohoku District of Japan with Large Ensemble Simulations Using the 5 km Regional Climate Model

Fast generation of high-dimensional spatial extremes

Widespread extreme climate events cause many fatalities, economic losses and have a huge impact on critical infrastructure. It is therefore of utmost importance to estimate the frequency and associated consequences of spatially concurrent extremes. Impact studies of climate extremes are severely hampered by the lack of extreme observations, and even large ensembles of climate simulations often do not include enough extreme or record-breaking climate events for robust analysis. On the other hand, weather generators specifically fitted to extreme observations can quickly generate many physically or statistically plausible extreme events, even with intensities that have never been observed before. We propose a Fourier-based algorithm for generating high-resolution synthetic datasets of rare events, using essential concepts of classical modelling of (spatial) extremes. Here, the key feature is that the stochastically generated datasets have the same spatial dependence as the observed extreme events. Using high-resolution gridded precipitation and temperature datasets, we show that the new algorithm produces realistic spatial patterns, and is particularly attractive compared to other existing methods for spatial extremes. It is exceptionally fast, easy to implement, scalable to high dimensions and, in principle, applicable for any spatial resolution. We generated datasets with 10000 gridpoints, a number that can be increased without difficulty. Since current impact models often require high-resolution climate inputs, the new algorithm is particularly useful for improved impact and vulnerability assessment.

Orbital dynamics in galactic potentials under mass transfer

Context. Time-dependent potentials are common in galactic systems that undergo significant evolution, interactions, or encounters with other galaxies, or when there are dynamic processes such as star formation and merging events. Recent studies show that an ensemble approach along with the so-called snapshot framework in the theory of dynamical systems provide a powerful tool to analyze the time-dependent dynamics. Aims. In this work, we aim to explore and quantify the phase space structure and dynamical complexity in time-dependent galactic potentials consisting of multiple components. Methods. We applied the classical method of Poincaré surface of sections to analyze the phase space structure in a chaotic Hamiltonian system subjected to parameter drift. This, however, makes sense only when the evolution of a large ensemble of initial conditions is followed. Numerical simulations explore the phase space structure of such ensembles while the system undergoes a continuous parameter change. The pair-wise average distance of ensemble members allowed us to define a generalized Lyapunov exponent, which might also be time-dependent, to describe the system stability. Results. We provide a comprehensive dynamical analysis of the system under circumstances where linear mass transfer occurs between the disk and bulge components of the model.

How could 50 °C be reached in Paris: Analyzing the CMIP6 ensemble to design storylines for adaptation

Reaching a surface temperature of 50 °C in a heavily populated region, like Paris, would have devastating effects. Although such a high value seems far from the present-day record of 42.6 °C, its occurrence cannot be dismissed by the end of the 21st century, due to the continuous increase of global mean temperature. In this paper, we address two questions that were asked by the City of Paris to a group of scientists: When does this event start to be likely? What are the prevailing meteorological conditions? We base our study on the CMIP6 simulation ensemble. Many of the CMIP6 yield biases in temperature. Rather than using methods of bias correction, which are not necessarily adapted to high extremes, we propose a pragmatic approach of model selection in order to seek such high temperature events that are deemed realistic. We analyze the meteorological conditions leading to first occurrences of such hot events and their common atmospheric patterns. This paper describes a simple data mining approach (on a large ensemble of climate model simulations) which could be adapted to other regions of the world, in order to help decision makers anticipating and adapting to such devastating meteorological events.

Aerosol-Induced Changes in Atmospheric and Oceanic Heat Transports in the CESM2 Large Ensemble

Abstract The total poleward energy transport (PET) is set by the top of atmosphere radiation flux and is therefore sensitive to any process which can alter those fluxes, particularly in the shortwave. One example is the direct and indirect effects of anthropogenic aerosols, which increase the local reflection of solar radiation back into space. The historic emission of sulfur dioxide, which peaked in the northern midlatitudes during the 1980s, has been proposed as a primary contributor to historic anomalies in cross-equatorial energy transport, as well as related processes such as a shift in the tropical rainband. In this study, we analyze simulations from the Community Earth System Model, version 2 (CESM2), large ensemble and single-forcing projects to better understand the forced response of PET to historical forcings. First, analysis of the single-forcing project reveals that the position of the intertropical convergence zone (ITCZ) responds in a nonlinear manner to greenhouse gas forcing in CESM2. This type of nonlinearity has been found previously in the context of the aerosol-only simulations in the CESM2 single-forcing project but may be the first identification of a similar effect in the greenhouse gas–only simulations. Second, through analysis of the full CESM2 large ensemble simulations, we find that anomalous heat transport occurred in both the atmosphere (through the mean meridional circulation and atmospheric eddies) and the oceans (through the Atlantic and Indo-Pacific sectors) due to a variety of related processes including the Hadley cells, the midlatitude storm tracks, the Atlantic meridional overturning circulation (AMOC), and the Pacific wind-driven subtropical gyre. Significance Statement In this study, we investigate how the Earth system changed the transport of energy from the tropics to the poles in response to historic pollution. We analyze a large number of climate model simulations of the recent past (1850 to present) and find that historic emissions of sulfur dioxide caused the model to transport more energy in the form of stronger ocean currents, stronger storms, and stronger prevailing winds. This is because the modeled currents in the Atlantic were too sensitive to historic pollution and transported too much warm water northward.

A novel conditional generative model for efficient ensemble forecasts of state variables in large-scale geological carbon storage

Integrating monitoring data to efficiently update reservoir pressure and CO2 plume distribution forecasts presents a significant challenge in geological carbon storage (GCS) applications. Inverse modeling techniques are commonly used to fuse observational data and refine reservoir model parameters, thereby improving state variable forecasts. However, these techniques often rely on linear or Gaussian assumptions, which can limit their effectiveness in accurately predicting state variables. Moreover, simulating large-scale three-dimensional (3D) GCS problems is computationally expensive, making iterative runs in inverse problems prohibitive. To address these challenges, we propose a conditional generative model utilizing the score-based diffusion method for real-time 3D pressure and saturation field distribution predictions. Our approach involves solving the score function with a mini-batch-based Monte Carlo estimator to generate labeled data. This data is subsequently employed to train a fully connected neural network, enabling it to learn the conditional sample generator within a supervised learning framework. This method enables the rapid generation of a large ensemble of predictions, facilitating comprehensive uncertainty quantification of state variables. We applied our method to forecast the dynamic 3D distributions of pressure and saturation fields over a 30-year injection period. The statistical assessment with low root mean square error (RMSE) values demonstrates that our method can accurately predict the spatiotemporal distributions of both pressure and saturation fields. Moreover, the developed conditional generative model shows high computational efficiency by generating 100 ensemble forecasts of 3D state variables in less than 10 min. The consistency between ensemble averages and ground truth values further illustrates the model’s capability to capture state variable dynamics during the CO2 plume injection process. Notably, the ground truth values fall within the ensemble forecasts, indicating that our uncertainty quantification effectively captures variability and potential noise in the observations. Thus, the developed conditional generative model proves to be a more efficient, accurate, and practical tool for GCS applications, facilitating timely risk analysis and informed decision-making.

Spatially resolved emulated annual temperature projections for overshoot pathways

Due to insufficient climate action over the past decade, it is increasingly likely that 1.5 °C of global warming will be exceeded – at least temporarily – in the 21st century. Such a temporary temperature overshoot carries additional climate risks which are poorly understood. Earth System Model climate projections are only available for a very limited number of overshoot pathways, thereby preventing comprehensive analysis of their impacts. Here, we address this issue by presenting a novel dataset of spatially resolved emulated annual temperature projections for different overshoot pathways. The dataset was created using the FaIR and MESMER emulators. First, FaIR was employed to translate ten different emission scenarios, including seven that are characterised by overshoot, into a large ensemble of forced global mean temperatures. These global mean temperatures were then converted into stochastic ensembles of local annual temperature fields using MESMER. To ensure an optimal tradeoff between accurate characterization of the ensemble spread and storage requirements for large ensembles, this procedure was accompanied by testing the sensitivity of sample quantiles to different ensemble sizes. The resulting dataset offers the unique opportunity to study local and regional climate change impacts of a range of overshoot scenarios, including the timing and magnitude of temperature thresholds exceedance.

The predictability of the effectiveness of chains of classifiers in the out-of-domain detection

Models in machine learning (ML) can produce erroneous results when exposed to the out-of-domain (OOD) data. Thus the problem of correctly incorporating the detection of OOD samples should be considered for all production systems that use the ML component. The production system can incorporate the filtering of OOD in three ways: pre-filtering data, filtering it in-model, and posteriori filtering, with the option to chain them into larger ensembles. Considering the number of possible types of filters, as well the length of the ensemble, it is important to be able to predict the upper and the lower bound of the performance of such ensembles on the basis of the performance of individual filters. The paper analyses challenges associated with the construction of the model, and then proposes the model where the performance of any chain of filters can be predicted, without actually assembling and testing it. An experiment is used to demonstrate the feasibility of the model, using the simple digit classification task, and F1 as the metric, resulting in the prediction error well below 0.1.

Robust handling of extremes in quantile mapping – “Murder your darlings”

Abstract. Quantile mapping is a method often used for the bias adjustment of climate model data toward a reference, i.e. to construct a transformation of the model's distribution to that of the reference. The main moments of the distributions are typically well transformed by quantile mapping, but statistical uncertainty increases towards the extreme tails, making robust transformations challenging. Because of the limited data at the extreme tails, an empirical quantile mapping also needs to make some estimation or fit a parameterised function for data beyond the calibration data range. Here, the MIdAS bias adjustment platform is employed to explore different methods for handling the extreme tail; these approaches are evaluated using an indicator of extreme precipitation – the maximum daily precipitation amount per year. Different methodologies are evaluated for a large ensemble of regional climate model projections over Scandinavia. The sensitivity of the empirical quantile mapping to the tails of the distribution is demonstrated, and it is found that the behaviour is significantly different within and outside of the calibration period, causing severe issues with the temporal consistency of the time series. The sensitivity is identified to be due to differences in the activated features of the bias adjustment within the calibration period (where the empirical transfer function is applied) and outside of that period (where the extrapolation method is likely applied). This means that the bias adjustment method is, in a sense, different between different time periods. Furthermore, finding a robust parameterisation for the tail is not straightforward. We identify a two-step solution that works well for this problem: We refer to the first step as “Murder your darlings”. By excluding data from the tail data in the calibration period, the extrapolation feature is activated for all time periods, even the calibration period. In the second step, applying an outlier-insensitive method for linear regression works well for finding an extrapolation parameterisation for the tail.

The relative importance of forced and unforced temperature patterns in driving the time variation of low-cloud feedback

Abstract Atmospheric models forced with observed sea-surface temperatures (SSTs) suggest a trend toward a more-stabilizing cloud feedback in recent decades, partly due to the surface cooling trend in the eastern Pacific (EP) and the warming trend in the western Pacific (WP). Here we show model evidence that the low-cloud feedback has contributions from both forced and unforced feedback components, and that its time variation arises in large part through changes in the relative importance of the two over time, rather than through variations in forced or unforced feedbacks themselves. Initial-condition large ensembles (LEs) suggest that the SST patterns are dominated by unforced variations for 30-year windows ending prior to the 1980s. In general, unforced SSTs are representative of an ENSO-like pattern, which corresponds to weak low-level stability in the tropics and less-stabilizing low-cloud feedback. Since the 1980s, the forced signals have become stronger, outweighing the unforced signals for the 30-year windows ending after the 2010s. Forced SSTs are characterized by relatively uniform warming with an enhancement in the WP, corresponding to a more-stabilizing low-cloud feedback in most cases. The time-evolving SST pattern due to this increasing importance of forced signals is the dominant contributor to the recent stabilizing shift of low-cloud feedback in the LEs. Using single-forcing LEs, we further find that if only greenhouse gases evolve with time, the transition to the domination of forced signals occurs 10-20 years earlier compared to the LEs with full forcings, which can be understood through the compensating effect between aerosols and greenhouse gases.

Subradiance and superradiant long-range excitation transport among quantum emitter ensembles in a waveguide

In contrast to free space, in waveguides the dispersive and dissipative dipole–dipole interactions among quantum emitters exhibit a periodic behavior over remarkably long distances. We propose a novel setup, to our knowledge, exploiting this long-range periodicity in order to create highly excited subradiant states and facilitate fast controlled collective energy transport among far-apart ensembles coupled to a waveguide. For sufficiently large ensembles, collective superradiant emission into the fiber modes dominates over its free space counterpart. We show that, for a large number of emitters, a fast transverse coherent pulse can create almost perfect subradiant states with up to 50% excitation. On the other hand, for a coherent excitation of one sub-ensemble above an overall excitation fraction of 50% we find a nearly lossless and fast energy transfer to the ground state sub-ensemble. This transport can be enhanced or suppressed by controlling the positions of the ensembles relative to each other, while it can also be realized with a random position distribution. In the optimally enhanced case this fast transfer appears as superradiant emission with subsequent superabsorption, yet, without a superradiant decay after the absorption. The highly excited subradiant states, as well as the superradiant excitation transfer, appear as suitable building blocks in applications such as active atomic clocks, quantum batteries, quantum information protocols, and quantum metrology procedures such as fiber-based Ramsey schemes.

Modulation of Northern Europe near-term anthropogenic warming and wettening assessed through internal variability storylines

Internal variability arising from the inherently chaotic nature of the climate system has amplified or obscured human-caused changes, especially at regional scales in the extratropics, where its contribution to climate variability is the largest. It is virtually certain that this will continue in the near-term. We here focus on the Northern Europe region, whose variability is largely controlled by the North Atlantic Oscillation (NAO) and the Atlantic Meridional Overturning Circulation (AMOC) through remote dynamical and thermodynamic processes, and introduce the concept of internal variability storylines (IVS) to explore, understand, and quantify the role of the two combined drivers of internal variability in the modulation of the anthropogenic warming by 2040 in winter. Based on a large ensemble of historical-scenario simulations, we show that the high-impact IVS, characterised by weak AMOC decline and a decadal shift of the NAO toward dominant positive phase, leads faster to warmer-wetter conditions independently of actual and future greenhouse gases emissions. By contrast, amplified AMOC reduction and more recurrent negative NAO can considerably damp both warming and wettening at near-term. In the latter IVS, we provide evidence that winter-severe conditions similar to those in 2010, that had been responsible for widespread socio-economic disruptions, remain almost as likely to occur by 2040. Reframing the uncertain climate outcomes into the physical science space in a conditional form through the prism of IVS makes climate information relevant for accurate risk assessments and adaptation planning.

Exploring a Data‐Driven Approach to Identify Regions of Change Associated With Future Climate Scenarios

AbstractA key consideration for evaluating climate projections is uncertainty in future radiative forcing scenarios. Although it is straightforward to monitor greenhouse gas concentrations and compare observations with specified climate scenarios, it remains less obvious how to detect and attribute regional pattern changes with plausible future mitigation scenarios. Here we introduce a machine learning approach for linking patterns of climate change with radiative forcing scenarios and use a feature attribution method to understand how these linkages are made. We train a neural network using output from the SPEAR Large Ensemble to classify whether temperature or precipitation maps are most likely to originate from one of several potential radiative forcing scenarios. Despite substantial atmospheric internal variability, the neural network learns to identify “fingerprint” patterns, including significant localized regions of change, that associate specific patterns of climate change with radiative forcing scenarios in each year of the simulations. We illustrate this using output from additional ensembles with sharp reductions in future greenhouse gases and highlight specific regions (in this example, the subpolar North Atlantic and Central Africa) that are critical for associating the new simulations with changes in radiative forcing scenarios. Overall, this framework suggests that explainable machine learning could provide one strategy for detecting a regional climate response to future mitigation efforts.

Encoding manifolds constructed from grating responses organize responses to natural scenes across mouse cortical visual areas.

We have created "encoding manifolds" to reveal the overall responses of a brain area to a variety of stimuli. Encoding manifolds organize response properties globally: each point on an encoding manifold is a neuron, and nearby neurons respond similarly to the stimulus ensemble in time. We previously found, using a large stimulus ensemble including optic flows, that encoding manifolds for the retina were highly clustered, with each cluster corresponding to a different ganglion cell type. In contrast, the topology of the V1 manifold was continuous. Now, using responses of individual neurons from the Allen Institute Visual Coding-Neuropixels dataset in the mouse, we infer encoding manifolds for V1 and for five higher cortical visual areas (VISam, VISal, VISpm, VISlm, and VISrl). We show here that the encoding manifold topology computed only from responses to various grating stimuli is also continuous, not only for V1 but also for the higher visual areas, with smooth coordinates spanning it that include, among others, orientation selectivity and firing-rate magnitude. Surprisingly, the encoding manifold for gratings also provides information about natural scene responses. To investigate whether neurons respond more strongly to gratings or natural scenes, we plot the log ratio of natural scene responses to grating responses (mean firing rates) on the encoding manifold. This reveals a global coordinate axis organizing neurons' preferences between these two stimuli. This coordinate is orthogonal (i.e., uncorrelated) to that organizing firing rate magnitudes in VISp. Analyzing layer responses, a preference for gratings is concentrated in layer 6, whereas preference for natural scenes tends to be higher in layers 2/3 and 4. We also find that preference for natural scenes dominates the responses of neurons that prefer low (0.02 cpd) and high (0.32 cpd) spatial frequencies, rather than intermediate ones (0.04 to 0.16 cpd). Conclusion: while gratings seem limited and natural scenes unconstrained, machine learning algorithms can reveal subtle relationships between them beyond linear techniques.