pith. machine review for the scientific record. sign in

arxiv: 2604.19119 · v1 · submitted 2026-04-21 · 🌌 astro-ph.SR · astro-ph.GA· astro-ph.IM

Recognition: unknown

A General Framework for Radial Velocity Calibration in Low-Resolution Spectroscopic Surveys: Correcting Wavelength-Dependent and Global Systematics with Application to LAMOST DR9

Jinming Zhang , Haibo Yuan , Zhijia Tian
Authors on Pith no claims yet

Pith reviewed 2026-05-10 02:27 UTC · model grok-4.3

classification 🌌 astro-ph.SR astro-ph.GAastro-ph.IM
keywords radial velocityLAMOSTlow-resolution spectroscopysystematics correctionwavelength calibrationzero-point offsetstellar kinematicsspectroscopic survey
0
0 comments X

The pith

A calibration framework fixes wavelength-dependent shifts and global offsets in LAMOST radial velocities, halving repeat scatter to 1.8 km/s.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a method to address two main sources of error in radial velocity measurements from low-resolution spectra: relative shifts between different wavelength segments of the same observation and uniform zero-point shifts across entire spectra. Spectra are divided into eight segments of roughly 500 Angstroms each, with offsets measured relative to the full spectrum and modeled using low-order polynomials at both spectrograph and fiber levels. Global zero-points receive a two-stage correction that first minimizes a joint chi-squared statistic at the spectrograph level using repeat observations and external cross-matches, then averages seasonal fiber-level offsets. After these steps the scatter in high signal-to-noise repeat measurements falls by a factor of two, reaching 1.8 km/s standard deviation and implying roughly 1.3 km/s single-measurement precision, while comparisons to independent surveys confirm dispersions near 2 km/s. The cleaned velocities support more reliable kinematic mapping of large stellar samples and the same procedure can be applied to other low-resolution surveys.

Core claim

The authors show that LAMOST low-resolution spectra exhibit both wavelength-dependent RV inconsistencies between spectral segments and global zero-point offsets. They correct the wavelength dependence by segmenting each spectrum into eight parts, measuring segment-wise offsets relative to the full spectrum at spectrograph and fiber levels, and fitting low-order polynomials to those offsets. Zero-point corrections are applied hierarchically: a joint chi-squared minimization at the spectrograph level constrained by repeat observations and cross-matches to APOGEE and Gaia RVS, followed by fiber-level averaging of seasonal offsets. The resulting catalog for 5.7 million spectra achieves a factor-

What carries the argument

Division of each spectrum into eight wavelength segments, measurement of segment-wise RV offsets relative to the full spectrum, low-order polynomial fits at spectrograph and fiber levels for wavelength dependence, and hierarchical chi-squared minimization plus seasonal averaging for global zero-point corrections.

If this is right

  • Cross-night repeat RV differences at high signal-to-noise drop from 3.6 km/s to 1.8 km/s standard deviation, implying 1.3 km/s single-measurement precision.
  • Dispersions against APOGEE and Gaia data fall from about 4.0 km/s to 2.0 km/s.
  • A homogeneous value-added catalog of corrected radial velocities for approximately 5.7 million spectra is released.
  • The same segment-polynomial and hierarchical zero-point procedure applies directly to radial-velocity calibration in other large-scale low-resolution spectroscopic surveys.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The cleaned LAMOST sample could support finer mapping of stellar motions across the Milky Way disk and halo than was previously possible with the uncorrected data.
  • Adopting the framework in upcoming surveys with similar resolution and fiber-fed designs may allow their velocity catalogs to be combined with LAMOST without introducing large systematic mismatches.
  • If low-order polynomials leave higher-frequency residual structure in some fibers, adding targeted higher-order terms or per-fiber spline corrections could be tested on repeat observations to check for further gains.
  • The released catalog enables direct tests of whether the achieved precision is limited by photon noise or by uncorrected instrument effects at the current resolution.

Load-bearing premise

Reference radial velocities from external surveys are free of their own wavelength-dependent or zero-point systematics that could be absorbed into the LAMOST corrections.

What would settle it

A comparison of the corrected LAMOST velocities against radial velocities from a high-resolution survey that was not used in any part of the calibration process, checking whether the dispersion remains near 2 km/s.

Figures

Figures reproduced from arXiv: 2604.19119 by Haibo Yuan, Jinming Zhang, Zhijia Tian.

Figure 1
Figure 1. Figure 1: Consistency of radial velocity measurements from LAMOST repeat observations. The sample is divided into S/N bins of width 5 from 5 to 250. For each repeat pair, the lower S/N sets the bin. In each bin, the RV difference distribution is fitted with a Gaussian; its standard deviation (σ) is plotted as the vertical value. The green curve shows same-night repeats, and the red curve shows cross-night re￾peats. … view at source ↗
Figure 2
Figure 2. Figure 2: Flowchart of this work 3.2. Internal inconsistency of LAMOST LRS wavelength calibration In wavelength calibration, any deviation in the pixel–wavelength relation introduces a systematic error in the spectrum at the corresponding wavelength. Ac- [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: This figure illustrates the segmentation strategy applied to a representative low-resolution spectrum from LAMOST LRS DR9. The entire spectrum is divided into eight working segments, each used for independent radial velocity measurements. For clarity, the four segments in the red arm are highlighted with red shading, while the four segments in the blue arm are shaded in light blue. The overlapping region b… view at source ↗
Figure 4
Figure 4. Figure 4: This figure shows segmented velocity offsets for 138 low-resolution spectra taken with spectrograph No. 9 in LAMOST plate GACII008N38B1 on December 1, 2019. The horizontal axis is fiber ID, and the vertical axis shows eight spectral segments (3900–9000 ˚A). The color indicates the dif￾ference between the radial velocity from each segment and that from the full spectrum. Gray dashed lines separate the segme… view at source ↗
Figure 5
Figure 5. Figure 5: Segmented velocity offsets for all spectrographs in LAMOST plate GACII008N38B1 on December 1, 2019. The horizontal axis is spectrograph ID, and the vertical axis shows eight spectral segments (3900–9000 ˚A), separated by gray dashed lines. For each spectrograph and segment, the offset is the µ of the segmental radial velocity differences relative to the full-spectrum velocity, derived from Gaussian fits. C… view at source ↗
Figure 6
Figure 6. Figure 6: Segmental velocity offsets of Spectrograph 9 by observing year. The offset is the difference between each segment’s radial velocity and that from the full spectrum, using the per-exposure µ offsets from a Gaussian fit to all spectra in that exposure. Each grid cell shows the offset for one segment in a single exposure. Panels show different years, with subtitles listing the year range and number of observi… view at source ↗
Figure 8
Figure 8. Figure 8: Monthly median residual velocity offsets for fiber 217 of spectrograph 3 after removing spectrograph-level offsets. Each observing year is shown in a separate panel, with months on the horizontal axis, spectral segments on the vertical axis, and color indicating residual velocity offset, enabling direct comparison of systematic deviations across years and spectral ranges [PITH_FULL_IMAGE:figures/full_fig_… view at source ↗
Figure 9
Figure 9. Figure 9: Radial velocity offsets per fiber for 2017–2018, after subtracting the overall spectrograph-level offset. For each fiber, RV offsets in eight spectral segments over the year were collected, and the offset was the µ obtained from a Gaussian fit. Each grid cell shows the segment-wise RV offset for one fiber. The horizontal axis is fiber ID; the vertical axis is the eight spectral segments. White vertical ban… view at source ↗
Figure 10
Figure 10. Figure 10: Example correction curve for Fiber 3 of Spec￾trograph 9 in the 2019–2020 observing year. The horizontal axis is wavelength and the vertical axis is RV offset. Blue and red points with error bars show the µ and σ of RV offsets in the blue- and red-arm segments, plotted at each segment’s central wavelength. Solid blue and red curves are cubic poly￾nomial fits to the four segments in each arm, showing how RV… view at source ↗
Figure 11
Figure 11. Figure 11: Comparison of radial velocity consistency between two plates with about 1700 common sources: plate HIP29425K201 observed on 2016-11-23 and plate HIP29425K2 on 2017-11-18. The horizontal axis shows spectrograph ID, and the vertical axis shows radial velocity differences (∆RV ) of repeated sources. Each spectrograph has about 100 repeats, whose velocity differences are fitted with a Gaussian; the fitted µ a… view at source ↗
Figure 12
Figure 12. Figure 12: Zero-point corrections for individual spectrographs. Each vertical strip represents a plate, and the sixteen horizontal cells within each strip represent its sixteen spectrographs. Cell color indicates the zero-point correction for that spectrograph, and the white cells indicate spectrographs without correction. Plates are ordered by observing time along the horizontal axis; the axis is labeled by month, … view at source ↗
Figure 13
Figure 13. Figure 13: Comparison of radial velocity consistency before and after the zero-point correction. Panel (a) shows the standard deviation of radial velocity differences (σ∆RV ) between repeated observations of the same sources in different LAMOST spectrographs, as a function of the number of repeated sources per spectrograph pair. Note that larger number of repeated sources provides a more reliable standard-deviation … view at source ↗
Figure 14
Figure 14. Figure 14: Distribution of velocity differences between LAMOST and Gaia versus observing time for the three fibers with the most common sources: SP3–Fiber144 (blue; 1162 sources), SP7–Fiber109 (red; 1142), and SP5–Fiber190 (green; 1140). The vertical axis shows ∆RVGaia = RVLAMOST − RVGaia after applying the wavelength-dependent correction and spectrograph-level zero-point calibration. The horizontal axis gives the o… view at source ↗
Figure 15
Figure 15. Figure 15: Radial velocity zero-points of individual fibers after spectrograph-level corrections. For each fiber and observing year, the zero-point is the weighted median RV difference between LAMOST and external references (Gaia, APOGEE), with weights proportional to the number of common sources. The horizontal axis shows fiber ID; the vertical axis lists observing years. Vertical black lines separate groups of 25 … view at source ↗
Figure 16
Figure 16. Figure 16: Comparison of radial-velocity consistency before and after the fiber level correction. Panel (a) shows the velocity scatter (σ∆RV ) of internal repeat observations within fiber-years as a function of the number of repeated pairs. Panels (b) and (c) show the velocity scatter of common sources between LAMOST and Gaia, and between LAMOST and APOGEE, respectively, also as a function of the number of matched p… view at source ↗
Figure 17
Figure 17. Figure 17: Fiber-level radial velocity zero-point corrections. Each panel corresponds to one spectrograph, with fiber ID (1–250) along the horizontal axis and observing years (2011–2012 to 2020–2021) arranged from top to bottom in chronological order. The color of each pixel represents the zero-point correction for the corresponding fiber in a given year (in km s−1 ), as indicated by the color bar on the right. Gray… view at source ↗
Figure 18
Figure 18. Figure 18: Comparison of RV precision and zero-point offsets before and after corrections. Panels (a) and (b) show internal consistency from LAMOST repeat observations: (a) same-night and (b) cross-night repeats. Each point is the Gaussian-fitted standard deviation σ∆RV of RV differences in S/N bins (bin size = 5). Panels (c) and (d) show external validation: (c) LAM￾OST–Gaia and (d) LAMOST–APOGEE common sources, wi… view at source ↗
Figure 19
Figure 19. Figure 19: Comparison of RV precision and zero-point offsets between our measurements and those from DESI DR1 and APOGEE DR19 for common stars. Panels (a) and (b) show the standard deviation σ∆RV of RV differences versus S/N. For DESI DR1 (panels a), we divided stars into 20 S/N bins from 0 to 200 and fit a Gaussian to the RV differences in each bin to obtain σ and µ. For APOGEE DR19 (panels b), we used 25 S/N bins … view at source ↗
Figure 20
Figure 20. Figure 20: Distribution of RV differences between DESI DR1 and APOGEE DR17 for their 14,449 common stars, where ∆RV = vDESI − vAPOGEE. The solid black line shows the µ of offset, and the dashed black lines show the ±1σ dispersion. This work greatly improves RV precision for LAM￾OST low-resolution spectra and provides a general cali￾bration framework for other large spectroscopic surveys, laying the groundwork for fu… view at source ↗
read the original abstract

Radial velocity (RV) is crucial for stellar kinematics and Galactic archaeology. The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) has obtained over ten million low-resolution spectra ($R \sim 1800$), yielding RVs for millions of stars, but these suffer from (1) wavelength-dependent inconsistencies (relative shifts between spectral segments) and (2) global zero-point offsets (uniform shifts of entire spectra). In this work, we comprehensively characterize and correct both. Each spectrum is first divided into eight segments of about 500 Angstrom. We organize the data at the spectrograph and fiber levels, measure segment-wise RV offsets relative to the full spectrum at each level, and then fit these offsets with low-order polynomials to correct wavelength-dependent systematics. We then correct zero-points hierarchically: at the spectrograph level by minimizing a joint chi-squared constrained by repeat observations and cross-matches with APOGEE and Gaia RVS, and at the fiber level by averaging seasonal offsets. After correction, RV precision improves significantly: for cross-night repeats, the standard deviation of RV differences at high signal-to-noise ratios drops by a factor of two from about 3.6 to about 1.8 km s$^{-1}$, implying a single-measurement precision of about 1.3 km s$^{-1}$. External checks with APOGEE and Gaia show dispersions drop from about 4.0 to about 2.0 km s$^{-1}$. The precision approaches, though slightly below, the theoretical limit at $R \sim 1800$. We release a value-added RV catalog with corrected velocities for about 5.7 million spectra, providing a homogeneous and systematically corrected dataset. The framework established in this work is also applicable to RV calibration in other large-scale spectroscopic surveys.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper presents a general framework for correcting wavelength-dependent and global zero-point systematics in radial velocities from low-resolution LAMOST spectra. Spectra are divided into eight ~500 Å segments; segment-wise RV offsets are measured and fit with low-order polynomials at the spectrograph and fiber levels to address wavelength dependence. Global zero-points are then corrected hierarchically: spectrograph-level offsets via joint χ² minimization constrained by repeat observations and APOGEE/Gaia cross-matches, followed by fiber-level seasonal averaging. Post-correction, cross-night repeat RV difference SD at high S/N drops from ~3.6 to ~1.8 km s⁻¹ (implying ~1.3 km s⁻¹ single-measurement precision), and external dispersions with APOGEE/Gaia drop from ~4.0 to ~2.0 km s⁻¹. A value-added catalog of corrected RVs for ~5.7 million spectra is released.

Significance. If the corrections are robust and independently validated, the work delivers a substantially more homogeneous and precise RV dataset for Galactic archaeology and stellar kinematics, approaching the theoretical limit at R~1800. The hierarchical, multi-level approach and release of the catalog are practical strengths; the framework is also positioned as reusable for other surveys.

major comments (2)
  1. [§3.3] §3.3 (zero-point correction): The spectrograph-level zero-point offsets are obtained by minimizing a joint χ² explicitly constrained by the same repeat observations and APOGEE/Gaia cross-matches that are later used to quantify the post-correction scatter reductions (3.6→1.8 km s⁻¹ for repeats; 4.0→2.0 km s⁻¹ externally). This creates a risk that the headline precision gains are partly by construction. A held-out validation subset, temporal split, or explicit cross-validation procedure must be demonstrated to confirm that the reported improvements reflect removal of systematics rather than fitting to the validation data.
  2. [§3.2] §3.2 (wavelength-dependent correction): The selection of exactly eight segments and the specific polynomial orders are presented without quantitative justification or residual-structure diagnostics. It is unclear how these choices were optimized to avoid under- or over-fitting while fully capturing wavelength-dependent shifts; additional tests (e.g., segment-number sensitivity or post-fit residual RV maps) are needed to support that the polynomial model is sufficient and generalizable.
minor comments (2)
  1. [Abstract] The abstract and methods should explicitly state the exact wavelength boundaries of the eight segments and confirm that the ~500 Å figure is uniform across the LAMOST range.
  2. Figure captions for before/after RV comparison plots should include the exact S/N cuts and number of objects used in each panel to allow direct reproduction of the quoted SD values.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments. We address each major point below and will revise the manuscript accordingly to strengthen the validation and justification of our methods.

read point-by-point responses

  1. Referee: [§3.3] §3.3 (zero-point correction): The spectrograph-level zero-point offsets are obtained by minimizing a joint χ² explicitly constrained by the same repeat observations and APOGEE/Gaia cross-matches that are later used to quantify the post-correction scatter reductions (3.6→1.8 km s⁻¹ for repeats; 4.0→2.0 km s⁻¹ externally). This creates a risk that the headline precision gains are partly by construction. A held-out validation subset, temporal split, or explicit cross-validation procedure must be demonstrated to confirm that the reported improvements reflect removal of systematics rather than fitting to the validation data.

    Authors: We acknowledge the validity of this concern about potential circularity. The joint χ² minimization determines the offsets that best reconcile the repeat observations and external cross-matches, after which the scatter reductions are measured on the corrected values. While this is a common approach for deriving global corrections, we agree that independent validation is required to confirm the gains arise from systematic removal. In the revised manuscript, we will add a held-out validation analysis in §3.3: the repeat observations will be partitioned (e.g., 70/30 training/validation split or temporal split for cross-night data), the minimization performed only on the training subset, and the scatter reduction then evaluated on the held-out data. Analogous checks will be shown for the APOGEE/Gaia cross-matches. Results will be presented quantitatively to demonstrate that the improvements persist. revision: yes

  2. Referee: [§3.2] §3.2 (wavelength-dependent correction): The selection of exactly eight segments and the specific polynomial orders are presented without quantitative justification or residual-structure diagnostics. It is unclear how these choices were optimized to avoid under- or over-fitting while fully capturing wavelength-dependent shifts; additional tests (e.g., segment-number sensitivity or post-fit residual RV maps) are needed to support that the polynomial model is sufficient and generalizable.

    Authors: We agree that the manuscript would benefit from explicit justification and diagnostics for these choices. The eight ~500 Å segments were selected to partition the LAMOST wavelength range while ensuring sufficient spectral features and S/N per segment for stable RV measurements; low-order polynomials were adopted after inspecting the observed offset trends to capture smooth wavelength dependence. To address the referee's request, we will expand §3.2 with: (i) a sensitivity study varying segment number (e.g., 4, 8, 16) and reporting effects on final RV precision and residual structure; (ii) post-fit residual RV maps or binned diagnostics showing the reduction of wavelength-dependent trends; and (iii) a brief description of how polynomial orders were chosen (e.g., via residual minimization on a subset). These additions will support the robustness and generalizability of the model. revision: yes

Circularity Check

1 steps flagged

Zero-point corrections fitted via chi-squared to repeats and APOGEE/Gaia matches, with precision gains then reported on the same data

specific steps
  1. fitted input called prediction [Abstract]
    "We then correct zero-points hierarchically: at the spectrograph level by minimizing a joint chi-squared constrained by repeat observations and cross-matches with APOGEE and Gaia RVS, and at the fiber level by averaging seasonal offsets. After correction, RV precision improves significantly: for cross-night repeats, the standard deviation of RV differences at high signal-to-noise ratios drops by a factor of two from about 3.6 to about 1.8 km s^{-1}, implying a single-measurement precision of about 1.3 km s^{-1}. External checks with APOGEE and Gaia show dispersions drop from about 4.0 to about "

    The zero-point corrections are determined by minimizing chi-squared that incorporates the repeat observations and APOGEE/Gaia cross-matches as constraints. The reported post-correction reductions in RV scatter for cross-night repeats and dispersions versus APOGEE/Gaia are measured on these identical datasets, so the improvement is statistically enforced by the fitting process rather than an independent test.

full rationale

The wavelength-dependent polynomial corrections are derived from internal segment-to-full-spectrum offsets at spectrograph and fiber levels and appear independent of the target RV values. However, the global zero-point corrections minimize a joint chi-squared explicitly using the repeat observations and APOGEE/Gaia cross-matches. The headline precision claims (cross-night repeat SD dropping 3.6→1.8 km s^{-1}; external dispersions 4.0→2.0 km s^{-1}) are then evaluated on these same fitted datasets without held-out validation. This reduces the reported improvements partly by construction, matching the fitted-input-called-prediction pattern for the central validation step. No self-citations or other circular patterns are present.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The framework rests on standard spectroscopic assumptions about the smoothness of wavelength-dependent offsets and the reliability of external reference catalogs; no new physical entities or ad-hoc constants are introduced beyond the choice of eight segments and low-order polynomials.

free parameters (2)
  • number of spectral segments
    Fixed at eight segments of ~500 Å each to enable per-segment offset measurement.
  • polynomial order
    Low-order polynomials chosen to model wavelength-dependent RV offsets at spectrograph and fiber levels.
axioms (2)
  • domain assumption Radial velocity offsets between spectral segments vary smoothly with wavelength and can be captured by low-order polynomials.
    Invoked when fitting segment-wise offsets to remove wavelength-dependent systematics.
  • domain assumption APOGEE and Gaia RVS radial velocities provide unbiased external anchors for zero-point calibration.
    Used in the joint chi-squared minimization at the spectrograph level.

pith-pipeline@v0.9.0 · 5656 in / 1559 out tokens · 44335 ms · 2026-05-10T02:27:00.244081+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

15 extracted references · 1 canonical work pages

  1. [1]

    2025, arXiv preprint arXiv:2503.14745 Bai, Z.-R., Zhang, H.-T., Yuan, H.-L., et al

    Abdul-Karim, M., Adame, A., Aguado, D., et al. 2025, arXiv preprint arXiv:2503.14745 Bai, Z.-R., Zhang, H.-T., Yuan, H.-L., et al. 2017, Research in Astronomy and Astrophysics, 17, 091 Blomme, R., Fr´ emat, Y., Sartoretti, P., et al. 2023, A&A, 674, A7, doi: 10.1051/0004-6361/202243685 Bohlin, R. C., M´ esz´ aros, S., Fleming, S. W., et al. 2017, The Astr...

    work page doi:10.1051/0004-6361/202243685 2025

  2. [2]

    A3, but for Spectrograph

    26 1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=358) 9 11 1 3 5 2012–2013 (N=796) 9 11 1 3 5 2013–2014 (N=704) 9 11 1 3 5 2014–2015 (N=707) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 20...

    2011

  3. [3]

    A3, but for Spectrograph

    1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=339) 9 11 1 3 5 2012–2013 (N=795) 9 11 1 3 5 2013–2014 (N=694) 9 11 1 3 5 2014–2015 (N=723) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2015–...

    2011

  4. [4]

    A3, but for Spectrograph

    27 1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=341) 9 11 1 3 5 2012–2013 (N=786) 9 11 1 3 5 2013–2014 (N=714) 9 11 1 3 5 2014–2015 (N=723) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 20...

    2011

  5. [5]

    A3, but for Spectrograph

    1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=338) 9 11 1 3 5 2012–2013 (N=776) 9 11 1 3 5 2013–2014 (N=698) 9 11 1 3 5 2014–2015 (N=702) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2015–...

    2011

  6. [6]

    A3, but for Spectrograph

    28 1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=340) 9 11 1 3 5 2012–2013 (N=786) 9 11 1 3 5 2013–2014 (N=701) 9 11 1 3 5 2014–2015 (N=703) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 20...

    2011

  7. [7]

    A3, but for Spectrograph

    1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=333) 9 11 1 3 5 2012–2013 (N=784) 9 11 1 3 5 2013–2014 (N=706) 9 11 1 3 5 2014–2015 (N=708) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2015–...

    2011

  8. [8]

    A3, but for Spectrograph

    29 1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=349) 9 11 1 3 5 2012–2013 (N=777) 9 11 1 3 5 2013–2014 (N=705) 9 11 1 3 5 2014–2015 (N=709) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 20...

    2011

  9. [9]

    A3, but for Spectrograph

    1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=336) 9 11 1 3 5 2012–2013 (N=768) 9 11 1 3 5 2013–2014 (N=688) 9 11 1 3 5 2014–2015 (N=686) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2015–...

    2011

  10. [10]

    A3, but for Spectrograph

    30 1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=299) 9 11 1 3 5 2012–2013 (N=773) 9 11 1 3 5 2013–2014 (N=702) 9 11 1 3 5 2014–2015 (N=687) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 20...

    2011

  11. [11]

    A3, but for Spectrograph

    1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=354) 9 11 1 3 5 2012–2013 (N=715) 9 11 1 3 5 2013–2014 (N=702) 9 11 1 3 5 2014–2015 (N=696) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2015–...

    2011

  12. [12]

    A3, but for Spectrograph

    31 1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=368) 9 11 1 3 5 2012–2013 (N=764) 9 11 1 3 5 2013–2014 (N=692) 9 11 1 3 5 2014–2015 (N=693) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 20...

    2011

  13. [13]

    A3, but for Spectrograph

    1012 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=359) 9 11 1 3 5 2012–2013 (N=788) 9 11 1 3 5 2013–2014 (N=712) 9 11 1 3 5 2014–2015 (N=716) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2015–...

    2011

  14. [14]

    A3, but for Spectrograph

    32 10 12 2 46 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2011–2012 (N=355) 9 11 1 3 5 2012–2013 (N=763) 9 11 1 3 5 2013–2014 (N=702) 9 11 1 3 5 2014–2015 (N=681) 9 11 1 3 5 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments 2...

    2011

  15. [15]

    50 100 150 200 250 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–4400 ÅSpectral Segments Spectrograph 1 50 100 150 200 250 Spectrograph 2 50 100 150 200 250 Spectrograph 3 50 100 150 200 250 Spectrograph 4 50 100 150 200 250 8500–9000 Å 8000–8500 Å 6500–7000 Å 6000–6500 Å 5300–5800 Å 4900–5400 Å 4400–4900 Å 3900–...

    2011