---
title: "Read a velocity–delay map as an inference from changing spectra"
description: "Open STORM Paper IX at its pinned version. Follow measured spectra into response maps and inspect the assumptions behind a disk interpretation."
canonical_url: "https://sandboxphysics.com/stories/agn-reverberation/reading-velocity-resolved-reverberation/"
language: "en"
page_type: "story"
alternate_languages:
  zh-CN: "https://sandboxphysics.com/zh/stories/agn-reverberation/reading-velocity-resolved-reverberation/"
last_modified: "2026-11-15"
---

L4 · Method guide

# Read a velocity–delay map as an inference from changing spectra

Sandbox Physics · 2026-11-15

Open Horne and collaborators’ STORM Paper IX, arXiv:2003.01448v2, beside this guide. Follow the measured light curves into the fitted response maps, then use an ideal ring to test the proposed interpretation. The useful habit is to ask what each axis means before trying to recognize a shape.

## Start with the observations and their reference light

The selected source is the version dated 27 November 2020, later published in 2021. The reading here covers Sections III–V and Figures 4, 6, 7, 9 and 10. It is a paper-reading exercise, not a rerun of the authors’ spectral calibration or MEMEcho reconstruction.

In Figure 4, locate the driving light curve at the bottom. Read upward to a line response and leftward to its delay distribution. Identify the observed points and their error bars before following the model curves. Then locate the separate slowly varying background. The same distinction matters when reading the velocity-resolved examples in Figure 9. [Open the selected paper](https://arxiv.org/html/2003.01448v2).

Make a small table with three columns: quantity measured, function fitted, and assumption needed. A spectrum gives flux at wavelengths on an observing date. The response at a particular delay is inferred by relating many such dates to earlier continuum changes. No telescope has directly photographed a delay coordinate.

## Read the response equation before the colours

Section III.1 defines changes around reference continuum and line levels. In the course notation, a wavelength-resolved response has the form

$$
\Delta L(v,t)=\int_0^\infty \Psi(v,\tau)\,\Delta C(t-\tau)\,d\tau.
$$

For fixed velocity $v$, the function $\Psi(v,\tau)$ tells how earlier central changes contribute at different delays. At fixed delay, it gives the spectrum of that response. Integrating over velocity recovers a one-dimensional delay distribution. Integrating over delay gives a response-weighted line profile, with units that follow the adopted normalization.

Write those two projections on paper. If you cannot explain one of them in words, return to the two axes. The height of a feature represents response strength; it is not automatically the number or density of clouds.

## Why the smoothness choice belongs beside the map

Sparse noisy time series permit many candidate response functions. Regularization chooses among them by favouring specified properties, such as smoothness. It supplies information about preferred solutions that the data alone do not determine.

Section III.2 explains the paper’s maximum-entropy regularization. The reading task is to find what is smoothed, then ask whether a feature persists when flexibility changes. In any inverse problem, a very smooth solution can erase structure, and a very flexible solution can follow noise. A small residual by itself does not choose the correct level of detail.

This is different from drawing a forward model with known clouds. In the Lab, geometry is supplied and the response is calculated. In the paper, the spectrum time series constrains a response, after which geometric models are used to interpret it.

## Check an apparent ring with four predictions

Figures 6 and 7 offer the main map-and-projection views. Before reading a geometric interpretation, write four predictions for a uniform circular ring: earliest response, latest response, maximum projected speed and the delay at that maximum. The preceding chapters provide all four.

For radius $R$ and inclination $i$, the predicted endpoints are $R(1-\sin i)/c$ and $R(1+\sin i)/c$. The largest projected speeds occur at delay $R/c$. Their magnitude is $\sqrt{GM/R}\sin i$. This is an independently derived ideal comparison, not a fit to the paper.

Check both sides of an apparent ellipse. Does the response have comparable near-side and far-side strength? Do different lines show the same structure? Are line blends marked? An apparent missing half can reflect unequal response rather than missing gas. These questions are more informative than deciding that a coloured patch “looks like a disk.”

![An ideal rotating ring traces an ellipse, with near-side response at ten days and far-side response at thirty days.](https://sandboxphysics.com/astronomy/ac52/figures/velocity-delay.en.svg)

The course’s analytic ring gives a reference pattern to compare with a published reconstruction. Its mass, radius and inclination are chosen teaching inputs. No pixel values from the STORM maps enter this drawing.

## Check the model where it struggles

The slowly varying backgrounds in Figure 10 and the discussion of the BLR Holiday identify changes that the simplest fixed response does not absorb. Use them as a reason to inspect the input–response assumptions, including the choice of observable continuum as a proxy for the unobserved ionizing radiation.

The uncertainty discussion in Section V.1 also deserves a separate note. A geometric reading of a smooth reconstructed map is different from a posterior distribution produced by fitting a specified dynamical model. Pancoast, Brewer and Treu’s [2014 modelling paper](https://arxiv.org/html/1407.2941v2), Sections 2.2–2.7, gives the latter framework: parameters, prior assumptions, a probability model for the observations and a search through allowed models.

The distinction explains what a reported mass uncertainty is conditional on. Better temporal sampling helps the data constrain rapid changes. A richer physical model tests additional interpretations. Both are useful, and they solve different parts of the problem.

## A checklist for the next paper

1. Record the source version, campaign dates, reference continuum and frame used for time and wavelength.
2. Separate measured spectra, fitted light curves, recovered responses and illustrative physical models.
3. Read both map projections and identify blends, gaps and background components.
4. Find the regularization or prior assumptions and the tests for sensitivity to them.
5. State whether parameter uncertainty came from a fitted model, a heuristic map comparison or an external calibration.

## Sources

[Horne and collaborators, STORM IX, arXiv:2003.01448v2](https://arxiv.org/html/2003.01448v2), Sections III–V and the figure numbers named above. The selected sections and captions were read; the complete observational reduction and reconstruction were not rerun.

[Pancoast, Brewer and Treu, arXiv:1407.2941v2](https://arxiv.org/html/1407.2941v2), Sections 2.1–2.7, for the structure and assumptions of direct geometric and dynamical modelling.

[Peterson and Horne, arXiv:astro-ph/0407538v1](https://arxiv.org/abs/astro-ph/0407538v1), Section 2 and Figure 1, for a simple forward ring comparison.

## Try the prediction

### [Compare a forward ring with a recovered map](https://sandboxphysics.com/labs/agn-reverberation/)

1. Predict the four ring landmarks before opening the paper’s map.
2. Use zero radial spread to inspect the model’s response ellipse, then compare both halves with the paper’s recovered response.
3. Identify which map comes from supplied geometry and which is inferred from spectral observations.

## Keep reading

Measure a black hole with light echoes · Reading path

### Measure a black hole with light echoes

1. [Article **How a black hole can be measured with an echo**](https://sandboxphysics.com/stories/agn-reverberation/why-gas-echoes-around-black-holes/)
2. [Article **Find the extra distance travelled by an echo**](https://sandboxphysics.com/stories/agn-reverberation/light-travel-time-and-delay-surfaces/)
3. [Article **Turn an orbiting ring into a velocity–delay map**](https://sandboxphysics.com/stories/agn-reverberation/velocity-delay-maps-and-keplerian-motion/)
4. [Article **Measure a lag before turning it into a mass**](https://sandboxphysics.com/stories/agn-reverberation/measuring-lags-and-virial-products/)
5. [Article **Recompute an observed echo in NGC 5548**](https://sandboxphysics.com/stories/agn-reverberation/recomputing-an-observed-reverberation-lag/)
6. [You are here **Read a velocity–delay map as an inference from changing spectra**](https://sandboxphysics.com/stories/agn-reverberation/reading-velocity-resolved-reverberation/)
