the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 22 Jul 2024
Enabling process science with CubeSat intersections: An orbit resampling study inspired by PREFIRE
Abstract. The Polar Radiant Energy in the Far-InfraRed Experiment (PREFIRE) will use two 6U CubeSats to continuously measure spectral far-infrared (FIR) emissions for the first time in the modern satellite era. By strategically operating two CubeSats in separate sun-synchronous orbits, PREFIRE will achieve frequent orbit resampling ("intersections") that afford insights into the underlying polar processes that modulate FIR emissions. These orbit intersections are integral to PREFIRE science and will likely feature prominently in future CubeSat missions, motivating methods to characterize resampling distributions. In this study, we develop new methods to locate orbit intersections and extract co-located pixels within crossovers. Such methods are applied to simulated PREFIRE orbits to characterize the spatial and temporal distribution of hypothetical PREFIRE intersections and identify the subset with short revisit times that can be used to inter-calibrate the PREFIRE sensors. The analysis confirms that hundreds of intersections are anticipated each day, with the majority (> 75 %) occurring poleward of 66.5°. Inter-calibration intersections are concentrated between 72° and 78° N/S and will be used to monitor changes in spectral differences between PREFIRE sensors.
Generalizing the analysis to pairs of polar orbiting CubeSats with different equatorial crossing times, we conclude that the second CubeSat nearly quadruples the number of total intersections available for polar process studies. Spatial and temporal resampling coverage is clearly enhanced when more than one CubeSat is deployed, securing greater latitudinal representation and more diverse time differences between crossovers. The spatio-temporal profile of intersections between CubeSats varies with the relative offset in their equatorial crossing times. Further, when two CubeSats are placed at different altitudes, we find that their intersections exhibit time-varying, cyclic coverage, significantly increasing latitudinal coverage relative to two CubeSats placed at identical altitudes that resample each other in constant latitude bands. This study ultimately illustrates some factors to consider when designing future CubeSat science missions and outlines methods for conducting the associated trade studies.
- Preprint
(5065 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-2040', Anonymous Referee #1, 27 Sep 2024
This manuscript outlines a strategy to identify collocations between the swaths of one or more push-broom sensors. The method is applied to two months of simulated orbits for the PREFIRE constellation. The spatial distribution of collocations is displayed, highlighting that, for two sensors at the same altitude (or for a sensor with its previous orbits), collocations fall in bands of constant latitude while sensors at different altitudes achieve collocations at all latitudes. These distributions can be used in planning validation efforts for future satellite missions.
I recommend that this paper is rejected for publication because I cannot think of anyone that would benefit from reading it. The manuscript is well drafted with clear explanations and logic. With improved figures, it would make an excellent technical report. However, the algorithm described is largely a brute force search with efficiencies that rely on the geometry of PREFIRE’s swath. The authors claim that the methods will be useful to future CubeSat missions but they make no attempt to generalise their algorithm: referring to the “fourth scene” rather than the swath centre or sub-orbital track (S2.1.2); dedicating several plots to analysis of spectral differences that are fairly uncommon in satellite constellations (Fig. 14/15); asserting that intersections will be “relatively compact diamond shapes” when this will not be the case for sensors with a wide swath or point-and-scan strategies (S2.1.4); and providing no guidance on how to select a sensible altitude separation to achieve a desired distribution of collocations (Fig. 11).
I do not mean to denigrate the effort that went into this work. It is undeniably important to the PREFIRE team. Beyond them, far too many of us have spent far to long searching for collocations between satellite datasets. (A web search immediately reveals half-a-dozen “tools” to find them.) It would be beneficial to the community to agree best practice in this area, and this paper has a role to play in that conversation. But it adds only one team’s experience and followed too narrow a path to inspire greater action.
I add a few minor thoughts that the authors may wish to consider if the paper is revised, particularly if my overall judgement is found unduly harsh by the other reviewers.
- I repeatedly desired a diagram that plainly illustrated the footprint of PREFIRE. In particular, I found it difficult to visualise how every sixth footprint is partially collocated. Was that in the across or along-track direction? Is the effect identical in each “scene” or does it vary? How does the pixel size relate to the spacing between them? Perhaps I’ve spent too long worrying about the MODIS bow-tie effect and overlaps between that sensor’s ten-line scans, but I struggled to create a mental image of what this paper described.
- I found many of the figures unhelpful. To explain a few:
- Fig. 2 looks like a flow chart but doesn’t summarise any of the logic applied. It appears to only specify every case that the example made in the text applies to.
- Figs. 3 and 4 are explained as showing that the interpolation scheme introduces negligible differences between orbits but shows absolute values rather than differences. This makes the magnitude of difference impossible to assess, either because the colour images are identical to the human eye or because the points overlap.
- Tab. 2 provides standard deviations for quantities that are obviously not normally distributed.
- Figs. 6 and 7 provide absolute counts of collocations. This might have been a clever way of avoiding making a colour legend but I struggle to interpret “counts”. Perhaps omit entirely or plot the number of collocations relative to the number of pixels at that latitude?
- Figs. 14 and 15 use two whole pages to convey that the difference in spectral radiance between satellites largely does not depend on scene. There must be a more efficient way of doing that, given almost every frame looks the same.
- Though the introductory details of the PREFIRE mission were interesting, they aren’t particularly relevant to the algorithms and data presented. One or more pages could be saved by trimming these down to a brief introduction.
- The fact that there are a fixed number of self-overlaps for each orbit, and that they fall at fixed latitudes that vary with altitude, strongly hints at there being a functional way of deriving many of the results of this work. The mathematics to do so are almost certainly beyond me so there is obligation to work it out, but I’d have been much more interested in this paper if it had attempted to empirically derive that relationship from an ensemble of simulated orbits.
- Section 2 cites a number of mathematically apt methods of collocation but, as best I can tell, fails to mention the theory underpinning many of the “collocation tools” that NASA and ESA have funded over the last few decades. I’m familiar with the Community Intercomparison Suite but one of its many competitors would be better suited to your application.
- L138: Did you mean “skipping every sixth along-track segment”?
- L149: You omit pixel pairs located within 500 km. Was there any physical motivation for this choice, as it strikes me that near the poles there may be almost complete overlap between orbits so they could occur quite close together.
- L172: You have my sympathy. My methods usually fail at the International Date Line.
- S1.4: In would be curious to know more about why you omitted repeating ground tracks. These seem ideal for your purposes, albeit with a long temporal separation.
- S2.2: The first paragraph caught my attention. I’ve had great success using the Python package shapely to efficiently manipulate the overlap of orbit tracks. Getting the pixels in the right order to produce a valid polygon is annoying, but the method explained in the rest of this section strikes me an as ad hoc solution. I’m not saying its wrong, but fixes like this section are what inclined me to reject the paper as the logic herein would be difficult to adapt to another sensor.
- Something horrible has happened to page number 18.
Citation: https://doi.org/10.5194/egusphere-2024-2040-RC1 -
RC2: 'Comment on egusphere-2024-2040', Anonymous Referee #2, 06 Dec 2024
This manuscript describes a methodology for the mission planning of the PREFIRE cubesat constellation emphasizing why a constellation is important for climate and instrument science, and it attempts to generalize its methodology for other cubesat constellations. It identifies, under a variety of scenarios, where two polar-orbiting cubesats overlap on the ground within differing time scales and in doing so highlights important characteristics for orbital planning of constellation missions.
I’m recommending this paper for publication after some major structural adjustments that clarify the methods and improve the applicability of the results to the broader community. I have little prior knowledge on PREFIRE or the Thermal-IR science community but, from my experience with Cubesats and vis/near-IR missions, it is important for details such as those included in this manuscript to undergo peer review and open discussion, and therefore be worthy of publication. This includes: the details on the orbital characteristics of the mission and the science rationale behind why those choices were made; the details on the instrument characteristics and the interactions between them, the orbit, and the science targets; and the current instrument science planning to maintain and utilize the data (i.e. intercalibration planning). This last point is especially important to undergo public discussion as the mission is ongoing and changes to those plans can be made through results of discussions on a publication such as this or responding publications. Therefore, it will surely help maximize PREFIRE’s science effectiveness by publishing and inviting further community input/iterations. It is for this reason that I think greater emphasis needs to be placed on the “why” of the colocation datasets rather than the how, though review of the specific algorithms used in more technical detail is also useful for people that may need to perform this analysis themselves when data is released, or those that may have improvements, but the current presentation of them is a bit convoluted and needs clarification.
General statements by section:
Section 1:
Overall, the background appears sufficient to properly place the scientific context of the mission and project goals within the broader TIR community, but there are some things which need to be rephrased or expanded on to allow the reader to fully follow the later discussions. For example:
L45: “… supercooled detectors (…) and long integration times (…) to narrowly resolve outgoing FIR” Is this statement trying to tell me that long integration times allow us to more narrowly resolve FIR? If so, how? What does resolve mean in this context? Spatially I know a longer integration time will reduce my ability to resolve ground phenomena, but spectrally it would improve signal to noise and improve that kind of resolution. Please clarify this.
L65: A reference (Miller et al. 2023) is provided for further details on instrument characteristics, but I think this paper would benefit from a better short form summary of the instruments’ characteristics. Why is it that the two instruments have different SRF’s? What is the instrument total FOV and why is it that the footprints overlap in the along-track? Is this a desired function or the result of inevitable trade-offs in the instrument design? You say by L73 that the instruments are capable of cover seasonal variations of the polar spectra. Are there references giving me an idea of the spectral range and radiance range of these variations that confirm that statement?
L76: “Guided by the premise that PREFIRE intersections provide insights into the dynamic local conditions … “ Are there references that inform this premise? What makes you certain the colocation delta T values you selected are relevant to the phenomena posed in the first few paragraphs? If the information is contained in the current references already, reiterating which ones here would be helpful.
Section 2:
L117: I understand SGP4 to lose accuracy the longer you run the simulation. You’ve included drag, which I’m sure reduces this, but do you know how accurate you really are going out to the 2-month mark? I can’t tell if 2 months was an arbitrary choice or if it was selected specifically to stop before the uncertainty becomes unwieldly. Now that the mission has launched and been flying for a number of months you could do an empirical comparison to get this without getting too deep in the math. If you took an actual TLE for June 28th, ran it through SGP4 with your drag model until August 28th how close is the prediction with the actual TLE from August 28th? Is whatever differences you find in doing so significant in any way as to influence the long term trends you observed in your simulations? My guess is probably not as the patterns are very cyclic and the orbit is relatively high for a cubesat (and therefore stable), but it’d improve the value of the publication to present that information for others that may want to use the same methods on similar cubesat missions, especially since one of the goals of the paper is to generalize for other cubesat missions.
The remainder of section two is very hard to follow as it relies on the authors’ expertise with this particular set of instruments. Things which are clear to them are difficult for me to follow. Partly this will be improved by my comments for Section 1 that request a better summary of the instrument characteristics. A clear bulleted list, or italics around instrument specific terms and their definitions would be helpful. Things like “scene”, “footprint”, “granule”, “along-track segment” and so on are defined explicitly in the text, but they’re spread out and hard to refer to when they become relevant later. I also think a clear, annotated figure showing the ground track footprints with the appropriate labels of these terms would be helpful. I’d also like clarification on the spectral dimensions (i.e. I’m still unclear if each scene has its own subset of spectral bands or contains all of them and knowledge of this is important to understand the later intercalibration project). Something like Figure 1, but for a single instrument ground-track rather than the vertices of an intersection.
L270: I believe this is the first use of the acronym “TCWV” so it should be written out in full here.
Figure 2: This flow chart is not very helpful. Overall, I’d like to see the entirety of the colocation algorithm clearly defined in a much more detailed flow chart, perhaps with annotations or highlighted sections which the body text of Section 2 can refer to directly. I’d prefer seeing a detailed flow chart and shorter Section 2 body text, as opposed to a large amount of body text and a simplistic flow chart. The algorithm appears very specific to the characteristics of the PREFIRE instruments and perhaps it would be better to have an in-depth, separate ATBD with math and code snippets, so this publication can focus on the generalizing the important concepts for cubesat constellations of all kinds.
Figure 4: The scatter plot points overlap greatly, as noted in the text, but it’s so much so that the figure really has no new information. The figure can either be omitted and summarized with statistical statement in text (i.e. differences were good within X%), or perhaps you can plot them as differences rather than overlaps.
Section 3:
For the section introduction, it’d be useful to see the references which validate what time intervals are relevant to the TIR analysis you hope to do. The plots and analysis are for |delta T| < 12 hours, and it should be confirmed from the literature why that interval is important.
Figure 5 shows that SAT1-SAT2 intersections over the 2-month test period which are within 48 hours of one another number over 100,000. Table 2 tells me the mean daily number of intersections was 474. A basic check of 474 times 60 days is well less than 100,000 even accounting for the standard deviation. Am I misunderstanding the relationship between these two datasets? If not, then this is a strong indication that the statistical objects listed here are not sufficient to actually describe the full dataset. Please clarify.
L370: Typo “… fdecrease …”
Section 4:
Figure 13: Typo L460 you’re missing a space between “(incomplete)” and “Planck functions”
Overall, the methodology for the intercalibration is the strongest part of this paper as it is very relevant to the future data quality of PREFIRE and readers may want to iterate on what already exists to improve the methodology as the mission progresses. Therefore, I’d like to see this section clearly outline a step-by-step algorithm for how the intercalibration process is done and what thresholds the instrument team will be flagging for corrections. The example is a 10% shift, but does the instrument pre-launch calibration uncertainty indicate that a 1% shift would warrant a similar correction? This section will also benefit from my earlier comments to help me better understand how scenes and spectral data are organized in the image data.
Section 5:
This section is very useful for the generalization of the work done in this manuscript to other missions. It provides useful information on how another Cubesat mission may want to target MLTAN differences according to their latitude coverage needs. Primarily, I’d like to see a proper x-axis label for Figure 17.
Section 6:
Overall, this section does a good job of succinctly clarifying and summarizing the results of prior sections. In fact, it was only in reading these conclusions that I fully understood some earlier parts of the manuscript. I’d like to see more direct references to the relevant sections. For example, on L550, you specify your conclusions come from the results of Section 5. Provide a similar clarification for other results for the remaining sections.
Citation: https://doi.org/10.5194/egusphere-2024-2040-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 176 | 67 | 104 | 347 | 10 | 13 |
- HTML: 176
- PDF: 67
- XML: 104
- Total: 347
- BibTeX: 10
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1