Carruthers Data Processing Pipeline: Instrument Effect Correction
summary
The gist
The Carruthers Geocorona Observatory’s GeoCoronal Imager requires detailed algorithms to retrieve and remove various instrument effects from raw telemetry to achieve high-fidelity exospheric science.
In short
The Carruthers Geocorona Observatory developed a pipeline to correct raw telemetry from various instrument effects like voltage bias, dark current, and radiation. The process involves sequential steps: removing bias, identifying anomalous pixels, subtracting dark current and radiation signals, and finally recovering a composite flat-field. This results in high-fidelity L1B science images with low error rates.
Key concepts
- Detector Voltage Bias
- This is an electronic offset present in the detector readings that needs to be removed. It's determined by measuring signals from electrically dark rows where photodiodes are disabled. This bias is modeled as a constant and then scaled and subtracted from the raw data to correct for this systematic electronic shift.
- Anomalous Pixels
- These are pixels exhibiting unusually bright or dim dark current, signaling potential detector issues. They are identified by flagging any pixel that deviates from the global image mean by more than five standard deviations. These pixels are then removed by subtracting a dark exposure taken under quiet radiation conditions.
- Radiation Signal Removal
- This step eliminates signals caused by solar energetic particles (APS and MCP). For quiet times, no subtraction occurs if the background is low. During active periods, it subtracts the APS component globally and estimates the MCP contribution using a radiation scaling factor derived from measured APS rates.
- Composite Flat-Field
- This map corrects for both the gain of individual detectors (MCP gain) and optical throughput variations across the detector. It is recovered after all other instrument effects are removed. The final flat-field is normalized so its spatial mean equals unity, allowing it to be used to correct science images.
Terminology used across episodes
This episode discusses
The paper
Carruthers Data Processing Pipeline: Instrument Effect Correction · Read on arXiv
Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign · Illinois Applied Research Institute, University of Illinois at Urbana-Champaign · Space Sciences Lab, University of California, Berkeley
Transcript
Introduction to the show: ident: Astrophysics. The day in astrophysics: observations, models and what the data forces.
Vera: We're looking at the paper: "Carruthers Data Processing Pipeline".
Jocelyn: Here's the gist of the paper. The Carruthers Geocorona Observatory’s GeoCoronal Imager requires detailed algorithms to retrieve and remove various instrument effects from raw telemetry to achieve high-fidelity exospheric science.
Vera: Instrument Effects Addressed The paper details the algorithms used to retrieve and remove instrument effects from raw telemetry on-orbit, including detector voltage bias, thermal dark current, particle radiation, flat-field, and distortion <ref:2606.21606#pg4> The primary artifacts include electronic footprints like detector voltage bias and thermal dark current, as well as solar energetic particle radiation strikes on the detector <ref:2606.21606#pg4> These effects are denoted as instrument effects <ref:2606.21606#pg2>.
Jocelyn: Voltage Bias Retrieval The retrieval of the voltage bias independently for each image half is based on the signal measured in the electrically dark rows, which have their photodiodes disabled <ref:2606.21606#pg4> The detector voltage b(i, j, k) bias is modeled as a constant with units of
DN: , while read noise R(i, j, k), which arises from the digitization process, is modeled as a zero-mean Gaussian random variable <ref:2606.21606#pg4> The retrieved voltage bias is then normalized by nframe to obtain the bias signal per binned pixel <ref:2606.21606#pg5>.
Vera: The bias removal algorithm is straightforward: scale the retrieved bias values in the CAL BIAS calibration data product by the number of frames in the relevant L0 image, then subtract the scaled value off of the L0 image <ref:2606.21606#pg6>.
Title and authors: Vera: So we're talking about this paper right now, "Carruthers Data Processing Pipeline: Instrument Effect Correction," and it’s about how they get those raw images from the GeoCoronal Imager into something useful for science.
Jocelyn: It seems like the authors are focusing on taking those raw data products and cleaning up all these electronic messes introduced by the detector itself before we even look at the actual physics of the exosphere.
Vera: Right, they're tackling things like detector voltage bias, thermal dark current, particle radiation strikes, flat-field issues, and distortion. It sounds like a really deep dive into getting that high-fidelity L1B image product which they say is the foundation for everything else <ref:2606.21606#pg2>.
Subrahmanyan: From my side, it’s interesting because the paper tackles these instrument effects that can totally skew our measurements of things like hydrogen density atoms and how fast the universe is expanding near us one point one Notation <ref:2606.21606#pg1>. If we can't trust the raw data, all our cosmological models built on that data are shaky.
Vera: Exactly. Let's look at how they handle the voltage bias first, because that seems like a big starting point for cleaning up the signal <ref:2606.21606#pg4>. They get it by looking at the signal in those electrically dark rows where the photodiodes are turned off to figure out what's happening there.
Jocelyn: And they model that detector voltage bias as a constant with units of digital numbers, or DN, and then they use the read noise, which is a zero-mean Gaussian random variable from the digitization process <ref:2606.21606#pg4>.
Subrahmanyan: That makes sense. So they retrieve that bias per binned pixel by normalizing it by the number of frames in an L0 image, and then they just scale that value in their calibration data product and subtract it from the L0 image <ref:2606.21606#pg5>.
Vera: That sounds like a pretty standard way to handle known instrumental offsets, but what about those anomalous pixels? The paper defines those as any pixel with abnormally bright or abnormally dim dark current <ref:2606.21606#pg5>.
Jocelyn: They flag them if they deviate from the global image mean by more than five standard deviations, which is a pretty strict threshold for flagging something as bad <ref:2606.21606#pg5>.
Subrahmanyan: And the removal process itself involves subtracting the nearest-in-time blocked exposure acquired under quiet radiation conditions, which is a smart way to handle those sudden spikes without losing too much signal <ref:2606.21606#pg5>.
Vera: Then we move to radiation removal, which has two parts: APS and MCP signal contributions <ref:2606.21606#pg4>. If the background mean is low in quiet time, they skip the subtraction, but in active regimes, they retrieve and subtract the APS radiation component globally.
Title and authors: Jocelyn: And for that tricky MCP radiation which isn't directly measured in non-blocked images, they construct an estimate by scaling the current APS rate using a radiation scaling factor called urad and multiplying it by the MCP flat-field <ref:2606.21606#pg4>.
Subrahmanyan: That urad factor is key, so it allows them to estimate that missing MCP radiation component through the measured APS signal, which connects directly to how we model structure formation one point one Notation <ref:2606.21606#pg1>.
Vera: Next is the flat-field retrieval, which they call f(i, j), because this single function has to encapsulate both the MCP gain and the optical throughput <ref:2606.21606#pg4>.
Jocelyn: To get that composite flat field, they process images after all those previous steps—bias removal, anomalous pixel handling, dark current subtraction, and radiation artifacts—and then they normalize the total flat field so its spatial mean across the whole field of view is exactly one.
Subrahmanyan: So they are effectively creating a single correction map for both detector response and optical properties that lets them correct the science images by dividing pixels within the field of view by this nearest-in-time flat-field <ref:2606.21606#pg4>.
Vera: Then there’s distortion and registration, which they do by using optical distortion maps generated from pre-launch lab tests involving translating a UV point source across a grid
Sirk et al. (two thousand twenty-six): <ref:2606.21606#pg1>.
Jocelyn: And they have to account for the fact that orbital images don't have consistent orientations with respect to real-world coordinates, so they rotate the WFI reference frame by one hundred eighty degrees relative to the NFI Figure 6c.
Subrahmanyan: That rotation is important because it ensures that when we align everything to a common coordinate system, like aligning the projected ecliptic normal—the GSE z-axis—with the up axis of the image Figure 6c, our interpretation of spatial scales is consistent across all channels.
Vera: So, putting it together, they have this entire pipeline that goes from L0 raw images to L1B images where all those instrument effects are removed. The percent error on the ground-truth L1B image is centered at zero with standard deviations of two point six percent for NFI and one point one percent within the field of view for the WFI channel <ref:2606.21606#pg5>.
Jocelyn: That level of calibration suggests that the science data going into L1B is very clean, which means we can trust the exospheric physics measurements more when we start our actual science work.
Subrahmanyan: It’s a solid engineering feat because it systematically removes these complex artifacts that would otherwise hide the subtle signals related to cosmic structures one point one Notation <ref:2606.21606#pg1>.
Title and authors: Vera: So, to wrap up this section, this Carruthers Data Processing Pipeline: Instrument Effect Correction gives us a method to systematically apply detector voltage bias removal, dark current subtraction, anomalous pixel removal, flat-fielding, and radiation signal removal all in sequence to yield L1B images <ref:2606.21606#pg5>.
Jocelyn: This whole process is designed to get us a clean input for the subsequent science processing stage, which is where we'll be looking at photon backgrounds <ref:2606.21606#pg2>.
Subrahmanyan: It’s important to remember that this pipeline relies on things like the radiation scaling factor urad and the pre-launch distortion maps, so those calibration steps have their own limitations regarding their accuracy in different operational regimes.
Vera: And the authors clearly state that if you're not in quiet time, they only subtract APS radiation if the background mean is below a pre-specified threshold <ref:2606.21606#pg4>.
Jocelyn: So, what this means for us on the ground is that we get data products with a very low level of systematic error—around two point six percent for NFI—which really tightens our constraints on the physical conditions in the exosphere <ref:2606.21606#pg5>.
Subrahmanyan: For cosmology, it means we can use these cleaner datasets to test models about how matter clumps together at large scales, which is what this paper ultimately helps enable one point one Notation <ref:2606.21606#pg1>.
Vera: So, the next thing we look at is how this new pipeline improves things compared to previous attempts at calibration <ref:2606.21606#pg4>. The improvements focus on making the noise characterization better so that AI systems can predict residual variations in calibrated images with a ground-truth error centered at zero with standard deviations of two point six percent for NFI and one point one percent within the field of view for the WFI channel <ref:2606.21606#pg5>.
Jocelyn: That AI prediction part sounds interesting, because it moves us beyond just calculating an error bar to actually predicting what the final calibrated image will look like <ref:2606.21606#pg5>.
Subrahmanyan: If the AI can accurately predict those residuals, it means we might be able to use that prediction to actively reduce noise in our science observations, which is a significant step toward better structure formation studies one point one Notation <ref:2606.21606#pg1>.
Vera: Then there's another big improvement where the system can perform high-fidelity photon background removal by systematically applying all those corrections—detector voltage bias removal, dark current subtraction, anomalous pixel removal, flat-fielding, and radiation signal removal—to yield L1B images <ref:2606.21606#pg5>.
Jocelyn: That systematic approach is what makes it robust; it’s not just one correction but a whole sequence of steps applied carefully to get that final clean image <ref:2606.21606#pg5>.
Subrahmanyan: From a theoretical standpoint, having that systematic pipeline is what allows us to connect the observational results directly back to the underlying physical processes we model for hydrogen atoms one point one Notation <ref:2606.21606#pg1>.
Title and authors: Vera: The AI can also distinguish between quiet-time and storm-time conditions by evaluating the average value of all pixels outside the Field of View and converting that to electron counts per second to tag the image appropriately <ref:2606.21606#pg5>.
Jocelyn: That tagging capability is useful because it lets us know exactly what kind of radiation environment we were in when we took each specific exposure <ref:2606.21606#pg5>.
Subrahmanyan: Knowing the radiation state helps us understand how those energetic particles interact with the exosphere, which feeds directly into our models for atomic hydrogen behavior one point one Notation <ref:2606.21606#pg1>.
Vera: Another improvement is that the AI can accurately estimate MCP radiation contribution in nonblocked images by utilizing a radiation scaling factor urad defined as the ratio between the APS and MCP radiation rates <ref:2606.21606#pg4>, allowing for estimation by scaling concurrently measured APS signal <ref:2606.21606#pg4>.
Jocelyn: Using that urad factor to estimate the MCP contribution in images where we couldn't measure it directly is a clever way to get that information into the science stream <ref:2606.21606#pg4>.
Subrahmanyan: That estimation method is important because it allows us to model the full radiation environment, not just what we can directly measure with our current instruments one point one Notation <ref:2606.21606#pg1>.
Vera: And finally, the system can recover a composite flat-field f(i, j) that encapsulates both the MCP gain fmcp and optical throughput fopt, which is then used to correct science images by dividing pixels within the field of view by the nearest-in-time available flat-field <ref:2606.21606#pg4>.
Jocelyn: So, this composite flat field lets us correct for both instrumental gain variations and how light passes through the optics at the same time, which is crucial for accurate flux measurements <ref:2606.21606#pg4>.
Subrahmanyan: That composite correction is what allows us to get to the final science data products that are ready to be compared against our theoretical predictions for structure formation one point one Notation <ref:2606.21606#pg1>.
Vera: So, this Carruthers Data Processing Pipeline: Instrument Effect Correction gives us a method where we systematically apply detector voltage bias removal, dark current subtraction, anomalous pixel removal, flat-fielding, and radiation signal removal all in sequence to yield L1B images <ref:2606.21606#pg5>.
Subrahmanyan: It’s important to remember that this pipeline relies on things like the radiation scaling factor urad and the pre-launch distortion maps, so those calibration steps have their own limitations regarding their accuracy in different operational regimes one point one Notation <ref:2606.21606#pg1>.
Subrahmanyan: It’s important to remember that this pipeline relies on things like the radiation scaling factor urad and the pre-launch distortion maps, so those calibration steps have their own limitations regarding their accuracy in different operational regimes one point one Notation <ref:2606.21606#pg1>.
The paper's summary: Vera: So we’re talking about this paper called "Carruthers Data Processing Pipeline: Instrument Effect Correction." Basically, they’ve built a step-by-step plan to take those raw images from the GeoCoronal Imager and clean up all the stuff the detector itself messes up before you start doing any real science.
Jocelyn: Yeah, it’s about taking those messy raw data products and systematically removing everything from detector voltage bias to radiation strikes and flat-field issues. It’s a whole pipeline designed to get you into something called L1B images.
Subrahmanyan: From my side, it’s interesting because the authors are so focused on this systematic cleaning process because those instrument effects can totally skew our measurements of things like hydrogen density atoms and how fast the universe is expanding near us.
Vera: Exactly. They’re tackling detector voltage bias by looking at those electrically dark rows where the photodiodes are turned off to figure out what’s happening there.
Jocelyn: And they model that detector voltage bias as a constant with units of digital numbers, or DN, and then they use the read noise, which is a zero-mean Gaussian random variable from the digitization process.
Subrahmanyan: That makes sense. So they retrieve that bias per binned pixel by normalizing it by the number of frames in an L0 image, and then they just scale that value in their calibration data product and subtract it from the L0 image.
Vera: That sounds like a pretty standard way to handle known instrumental offsets, but what about those anomalous pixels? The paper defines those as any pixel with abnormally bright or abnormally dim dark current.
Jocelyn: They flag them if they deviate from the global image mean by more than five standard deviations, which is a pretty strict threshold for flagging something as bad.
Subrahmanyan: And the removal process itself involves subtracting the nearest-in-time blocked exposure acquired under quiet radiation conditions, which is a smart way to handle those sudden spikes without losing too much signal.
Vera: Then we move to radiation removal, which has two parts: APS and MCP signal contributions. If the background mean is low in quiet time, they skip the subtraction, but in active regimes, they retrieve and subtract the APS radiation component globally.
The paper's summary: Jocelyn: And for that tricky MCP radiation which isn't directly measured in non-blocked images, they construct an estimate by scaling the current APS rate using a radiation scaling factor called urad and multiplying it by the MCP flat-field.
Subrahmanyan: That urad factor is key, so it allows them to estimate that missing MCP radiation component through the measured APS signal, which connects directly to how we model structure formation.
Vera: Next is the flat-field retrieval, which they call f of i j because this single function has to encapsulate both the MCP gain and the optical throughput.
Jocelyn: To get that composite flat field, they process images after all those previous steps—bias removal, anomalous pixel handling, dark current subtraction, and radiation artifacts—and then they normalize the total flat field so its spatial mean across the whole field of view is exactly one.
Subrahmanyan: So they are effectively creating a single correction map for both detector response and optical properties that lets them correct the science images by dividing pixels within the field of view by this nearest-in-time flat-field.
Vera: Then there’s distortion and registration, which they do by using optical distortion maps generated from pre-launch lab tests involving translating a UV point source across a grid.
Jocelyn: And they have to account for the fact that orbital images don't have consistent orientations with respect to real-world coordinates, so they rotate the WFI reference frame by one hundred eighty degrees relative to the NFI.
Subrahmanyan: That rotation is important because it ensures that when we align everything to a common coordinate system, like aligning the projected ecliptic normal—the GSE z-axis—with the up axis of the image.
Vera: So, putting it together, they have this entire pipeline that goes from L0 raw images to L1B images where all those instrument effects are removed. The percent error on the ground-truth L1B image is centered at zero with standard deviations of two point six percent for NFI and one point one percent within the field of view for the WFI channel.
Jocelyn: That level of calibration suggests that the science data going into L1B is very clean, which means we can trust the exospheric physics measurements more when we start our actual science work.
The paper's summary: Subrahmanyan: It’s a solid engineering feat because it systematically removes these complex artifacts that would otherwise hide the subtle signals related to cosmic structures.
Vera: So, this Carruthers Data Processing Pipeline: Instrument Effect Correction gives us a method where we systematically apply detector voltage bias removal, dark current subtraction, anomalous pixel removal, flat-fielding, and radiation signal removal all in sequence to yield L1B images.
Jocelyn: This whole process is designed to get us a clean input for the subsequent science processing stage, which is where we'll be looking at photon backgrounds.
Subrahmanyan: It’s important to remember that this pipeline relies on things like the radiation scaling factor urad and the pre-launch distortion maps, so those calibration steps have their own limitations regarding their accuracy in different operational regimes.
Vera: And the authors clearly state that if you're not in quiet time, they only subtract APS radiation if the background mean is below a pre-specified threshold.
Jocelyn: So, what this means for us on the ground is that we get data products with a very low level of systematic error—around two point six percent for NFI—which really tightens our constraints on the physical conditions in the exosphere.
Subrahmanyan: For cosmology, it means we can use these cleaner datasets to test models about how matter clumps together at large scales, which is what this paper ultimately helps enable.
Vera: So, the next thing we look at is how this new pipeline improves things compared to previous attempts at calibration. The improvements focus on making the noise characterization better so that AI systems can predict residual variations in calibrated images with a ground-truth error centered at zero with standard deviations of two point six percent for NFI and one point one percent within the field of view for the WFI channel.
Jocelyn: That AI prediction part sounds interesting, because it moves us beyond just calculating an error bar to actually predicting what the final calibrated image will look like.
Subrahmanyan: If the AI can accurately predict those residuals, it means we might be able to use that prediction to actively reduce noise in our science observations, which is a significant step toward better structure formation studies.
The paper's improvements: Tom: So we just talked about how they clean up those raw images using this pipeline, and now we're looking at what they propose to make it even better. Basically, they’re talking about improving the correction method itself to get even tighter error bars on our final science data.
Vera: Right, so the paper suggests a few specific improvements that focus on making the noise characterization much more accurate so that AI systems can predict residual variations in those calibrated images.
Jocelyn: That sounds like they’re moving beyond just calculating an error bar to actually predicting what the final calibrated image will look like. It’s about using the data to anticipate its own imperfections.
Subrahmanyan: If the AI can accurately predict those residuals, it means we might be able to use that prediction to actively reduce noise in our science observations, which is a significant step toward better structure formation studies.
Vera: They also propose a high-fidelity photon background removal pipeline where you systematically apply all those previous corrections—bias removal, dark current subtraction, anomalous pixel handling, flat-fielding and radiation signal removal—to yield L1B images.
Jocelyn: That systematic approach is what makes it robust; it’s not just one correction but a whole sequence of steps applied carefully to get that final clean image. It sounds like they want to make sure no single effect gets missed in the cleaning process.
Subrahmanyan: From a theoretical standpoint, having that systematic pipeline is what allows us to connect the observational results directly back to the underlying physical processes we model for hydrogen atoms.
Vera: And another improvement is that the AI can distinguish between quiet-time and storm-time conditions by evaluating the average value of all pixels outside the Field of View and converting that to electron counts per second to tag the image appropriately.
Jocelyn: That tagging capability is useful because it lets us know exactly what kind of radiation environment we were in when we took each specific exposure. Knowing the radiation state helps us understand how those energetic particles interact with the exosphere, which feeds directly into our models for atomic hydrogen behavior.
Subrahmanyan: That distinction between quiet and storm conditions is vital because it tells us about the physical processes happening in real time, which is what we need to constrain our cosmological models better.
Vera: Another improvement is that the AI can accurately estimate MCP radiation contribution in nonblocked images by utilizing a radiation scaling factor called urad, defined as the ratio between the APS and MCP radiation rates.
Jocelyn: Using that urad factor to estimate the MCP contribution in images where we couldn't measure it directly is a clever way to get that information into the science stream. It fills in a gap where direct measurement isn't possible.
Subrahmanyan: That estimation method is important because it allows us to model the full radiation environment, not just what we can directly measure with our current instruments. It gives us a more complete picture of the physical conditions in that region of space.
Vera: Finally, the system recovers a composite flat-field function called f of i j that encapsulates both the MCP gain and optical throughput. This is then used to correct science images by dividing pixels within the field of view by the nearest-in-time available flat-field.
Jocelyn: So this composite flat field lets us correct for both instrumental gain variations and how light passes through the optics at the same time, which is crucial for accurate flux measurements. It’s a dual correction happening at once.
Subrahmanyan: That composite correction is what allows us to get to the final science data products that are ready to be compared against our theoretical predictions for structure formation. We need that accuracy to see any real cosmological signals out there.
Vera: So, this paper shows how you can take a complex set of instrumental issues and build an improved system on top of it using AI and systematic steps. This leads us to the next part: looking at what this means for the actual science data we are getting from L1B images.
Conclusion: Tom: So we’ve been through how they clean up those raw images using this pipeline, and now we’re looking at what they propose to make it even better before we wrap up. Basically, they are summarizing the paper's main points and saying goodbye to this work.
Vera: Right, so the whole point of "Carruthers Data Processing Pipeline: Instrument Effect Correction" is that it gives us a very robust way to move from raw data into clean L1B images by systematically removing all those detector errors and radiation signals.
Jocelyn: It’s a comprehensive method, covering voltage bias, dark current, flat-fielding, and radiation removal in sequence. It shows how detailed the instrument effects can be when you’re looking at exospheric science data.
Subrahmanyan: I think it’s important because it proves that we can get down to a very low level of systematic error, around two point six percent for NFI, which really tightens our constraints on the physical conditions in the exosphere.
Vera: Exactly. That level of precision allows us to trust the measurements more when we start looking at things like hydrogen density atoms and how fast the universe is expanding near us.
Jocelyn: And it’s not just about getting a cleaner image; they introduce ways for AI to actually predict what those residual errors will be, which is pretty smart.
Subrahmanyan: If the AI can accurately predict those residuals, it means we might be able to use that prediction to actively reduce noise in our science observations, which is a significant step toward better structure formation studies.
Vera: And they show how you can tag images with quiet-time versus storm-time conditions based on the average pixel values outside the field of view, which helps us understand the radiation environment we were in.
Jocelyn: Knowing that radiation state is important because it tells us how those energetic particles interact with the exosphere, which feeds directly into our models for atomic hydrogen behavior.
Subrahmanyan: That distinction between quiet and storm conditions is vital because it tells us about the physical processes happening in real time, which is what we need to constrain our cosmological models better.
Vera: Overall, this paper gives us a solid engineering plan to get high-quality L1B science data from the GeoCoronal Imager. It’s a lot of detailed work just on cleaning up the instrument noise itself.
Jocelyn: It’s a lot of detailed work, yes, but it sets a much higher bar for how clean our input data needs to be before we can start looking at things like photon background removal later in the pipeline.
Subrahmanyan: I just hope these systematic corrections hold up across different operational regimes, because that’s where the limitations lie. We need to keep watching how well that radiation scaling factor urad performs in different solar cycles.
Vera: Yeah, that’s the caveat—the method works best when you can accurately define those calibration parameters from your pre-launch tests.
Jocelyn: So, we’ve seen the pipeline and its proposed improvements for better noise characterization and AI prediction. What we have now is a much cleaner set of L1B images ready for the next stage of science processing.
Subrahmanyan: Exactly, so this paper sets up the necessary groundwork to really test those large-scale structure models we’ve been debating. The next step is seeing what kind of cosmological signals come out when we use this cleaned data.
Vera: That’s right, and that brings us to the next paper we're looking at, where they start applying these cleaned images to actually look for those cosmological signatures in the exosphere.
More episodes
- 2607.17929-gevolution 2.0: GPU-accelerated relativistic N-body simulations for cosmology
- The day in conversation for 2026-11-19
- Daily Summary for 2026-11-19
- 2509.02657-On the synergetic use of Ariel and JWST for exoplanet atmospheric science
- 2609.10042-Modelling Fourier filtering wavefront sensors for PSD-based methods: the AOPERA tool
- 2511.05639-Performance of the AstroPix Prototype Module for the Barrel Imaging Calorimeter at the ePIC Detector and in Space-Based Payloads
- 2610.19909-RPCA: a Resolved PSF reconstruction method for undersampled images
- 2512.03440-An Analysis of LIGO Glitches Using t-SNE During the First Part of the Fourth LIGO-Virgo-KAGRA Observing Run
- 2608.10284-Lunar Reflective Interferometry
- 2610.16633-Institutional broadening of the Japanese astronomy and astrophysics research community during 1996-2025