Guide to RHESSI Image Archive

Last Update 29-Jan-2023, Kim Tolbert (kim.tolbert at
This version uses the updated flare catalog created in September 2022 based on the updated aspect software and database.

Images and related products were generated for >75,000 RHESSI flares using the procedures described in RHESSI Image Archive Strategy. Quick access to the flare images is provided through the RHESSI Flare Image Archive monthly listings. To view the available information for a given flare, first select the year and month via the pulldown options, and click "Load". Then scan the line-per-flare listing for the flare of interest and select the link to the full web page for that flare, or to a quick look at various plots. The figure below shows the beginning of the list for July 2002.

flare list with links

Some things to note regarding the list:


Some of the images in this archive are invalid. Usually this is obvious from the lack of coherent structure in a single image, or by inconsistency between successive times or energies. There are several reasons for this.

If you encounter problems with images in the archive, please let me know (kim.tolbert at

Below, we discuss the items on each flare web page, explain how to use the FITS files in the archive, and describe each type of plot included in the image archive. We used the flare of 10-Apr-2002 18:59 - 19:21 (flare number 2041055) as an example.

Flare Web Pages
Using the FITS files from the Image Archive
Aspect Solution Plots
Time Profile Plots
Image Panel Display
Visibility Comparison Plots
Profile Comparison Plots

Flare Web Pages

flare page sample Each single-flare web page shows some general information about the flare (start, peak, and end times, peak rate, and more from the RHESSI event list) and two plots - the aspect solution plot during the full flare interval and GOES and RHESSI time profile plots showing the time/energy binning selected for this flare (both plots are discussed in detail below). Links to the following items are located to the right of these plots:

Below this general area, the images, image movies, and other plots generated for each image reconstruction algorithm (Back Projection, Clean, Clean_59, MEM_GE, VIS_CS, and VIS_FWDFIT) are displayed or links are provided.

Using the FITS files from the Image Archive

You can download the imagecube, visibility, and eventlist FITS files for a particular flare by right-clicking the link on the flare web page, or by going directly to the archive (links below). In the commands below, we assume the relevant file name is in the variable 'fitsfilename'.

  1. Imagecube FITS files (archive)
  2. Visibility FITS files (archive)
  3. Eventlist FITS files (archive)

Aspect Solution Plots

aspect plot

The direction of the imaging axis is determined on sub-second time scales with sub-arcsecond accuracy from the data provided by the RHESSI Aspect System made up of the Solar Aspect System and the Roll Angle System. The aspect solution plot shows this direction of the imaging axis on the solar disk (solar limb shown in dark blue) as it moves in circles (black) about the spin axis (red +), during the full flare time interval. Also shown is the flare location (green X).

Each black circle in the plot represents one rotation of the spacecraft (4 seconds). The spin axis is plotted simply at the center of each 4s circle. The diameter of the circle changes with time as the image axis precesses about the spin axis. Three main factors control this precession - (1) the contraction/expansion of the spacecraft components with changes in temperature resulting from the day/night cycle, (2) the physical motion of the attenuators moving into and out of the detector fields of view, and (3) the actions of the spacecraft attitude control system as the magnetic torquers keep the spin axis in approximately the same location on the solar disk (near disk center in the southwest quadrant). Once a month or so, the torquers are used to increase the spin rate back to the nominal 15 rpm.

The aspect solution plot on the flare web page covers the entire flare time interval. Movies showing the aspect solution during each image time bin are available through the 'Movie of Aspect Solution in Image Time Bins' link.

bad aspect plot Occasionally, errors in the aspect solution show up as wild departures from the circular motion of the imaging axis. An example is shown in the figure on the right for 06-Mar-2002 00:02 - 00:22. A bad aspect solution usually makes it impossible to make an image for those time intervals. Looking at the aspect solution plot movies reveals when the aspect solution failed. In the case on the right, all image time intervals had a good aspect solution except for the first one as can be seen in this movie.

RHESSI image reconstruction relies on modulation by the grids, but note that there is no modulation at all for a source located on the spin axis if the image axis is also coincident with the spin axis. Thus, image reconstruction becomes more problematic if the source is close to the spin axis, especially if the diameter of imaging axis circles is small. In those cases, there may not be a full modulation period for the coarser grids and only compact sources can be imaged from the modulation with the finer grids.

Time Profile Plots

time profile plot The time profile plot has 3 panels.

  • The top panel shows the profile of the two GOES channels (1.0-8.0 Å in red, 0.5-4.0 Å in blue) in watts / m2. The time interval is expanded (RHESSI flare time +- 1 hour) for context. The gray shading (and the dashed lines angling down to the next panel) indicates the time range of the two lower panels.
  • The middle panel shows the time and energy bins selected for making images. The colors correspond to the energy band colors in the bottom panel, i.e.
    3 - 6 keV - black
    6 - 12 keV - magenta
    12 - 25 keV - green
    25 - 50 keV - cyan
    50 - 100 keV - yellow
    100 - 300 keV - red
    Note: these are the potential bins - if the signal-to-noise ratio is below the selected limit, an image will not be generated for that time/energy bin.
  • The bottom panel shows the count rate time profile (counts s-1 detector-1) in six energy bands for the flare. These profiles are generated from the quicklook data in the 'full_rate' observing summary files. They are different from the observing summary plots shown in the RHESSI Browser because they show only the rates from the detector front segments, while the Browser plots use the combined front plus rear segment rates.

    The colored bars across the top of the plot indicate the flare (red), night (cyan), SAA (orange), and attenuator state (magenta). The rates are not corrected for changes in attenuator state or decimation level. The attenuator changes are seen as large steps in the count rate as the shutters moved in and out of the detector fields of view. In this example, RHESSI was in attenuator state 1 (thin attenuator) for the entire time interval.

Image Panel Display

For each algorithm, all of the images that were generated are displayed in a grid of time (horizontal) and energy (vertical) bins. Some of the cells are blank if the signal-to-noise (SNR) ratio was too low in that time/energy bin. The SNR requirement was 4.0 for MEM_GE and 2.0 for the other algorithms. It was determined empirically with the goal of producing only scientifically meaningful images (unfortunately some nonsense images are still made). There are two panel display plots:

  • Scaled - The scaled panel display is shown by default in the web page. The color scheme for each image in one energy band is scaled to the maximum value of ALL the images in that energy band. At the top of each image, two numbers are displayed: 'SNR' and 'Rel Flux'. The SNR is the signal-to-noise ratio for that time/energy bin. The Rel Flux, relative flux, is the ratio of the maximum flux of that image to the maximum of the peak image in that energy band. The dashed line around an image indicates which image the others are scaled to.
  • Unscaled - In the unscaled panel display, the color scheme of each image is scaled to the maximum flux in that image, without regard to the other images in the panel. The SNR for each image is displayed at the top of the image.

Here (and similarly in the flare image web page), clicking the plot opens the plot file, and clicking again in the plot enlarges the plot, centered on the location you clicked, with a scroll bar along the bottom if needed.

The example shown here is for the CLEAN algorithm.

Scaled Panel Display
Scaled panel display

Unscaled Panel Display
Unscaled panel display


For each image reconstruction algorithm, the image frame at the peak time interval for each energy band that could be imaged is displayed and can be enlarged by clicking the image. Below the image is a figure showing the corresponding visibility comparison plot. Above each image are links for the following:

  • "Image Movie" - movie of the images for each time interval in that energy band.
  • "Image Movie with Contours" - movie of the images for each time interval in that band overlaid by contours at levels relative to the peak flux dependent on the SNR of the image - SNR > 10: 5,10,30,50,70,90%, SNR = 5 to 10: 10,30,50,70,90%, and SNR < 5: 30,50,70,90%.
  • "Visibility Comparison Movie" - movie of the visibility comparison plots for each time interval in that energy band.
  • "Profile Comparison Movie" - movie of the count rate profile comparison plots for each time interval in that energy band (available only for CLEAN and BACK PROJECTION).

For some flares, the left-most column contains an image and link for a movie of the low-energy images overlaid by contours at 50,70,90% of the high-energy images. These are available when images were made for more than one energy band and there is a high-energy band at 12-25 keV or above with an SNR value of > 5.

image plot

Each movie frame (an example of the 12-25 keV CLEAN image at peak time is on the left) shows the following:

  • The quicklook time profile in six energy bands for the entire flare, with the time interval of the image identified by a box. The color of the energy traces and the box are the same as for the Time Profile Plot described above. The attenuator state changes are shown in the magenta bar along the top.
  • The image for the time/energy band in units of photons cm-2 s-1 asec-2 (except for back projection images which are unitless) using IDL color table 5. The dotted lines are the latitude/longitude grid every five degrees. The solid line is the solar limb (not shown in example since this flare is not near the limb).
  • The legend in the image shows
    • the time interval and energy band used,
    • the detectors used (F means front segment),
    • the attenuator state during this time interval,
    • the reconstruction algorithm used,
    • the Signal-to-Noise Ratio (SNR),
    • and the total number of counts in the time/energy bin used to make the image. You may wonder why the total number of counts is smaller in the visibility-based images (MEM_GE, VIS_CS, and VIS_FWDFIT). That is normal because some visibilities cannot be made because there is incomplete phase coverage at some rotation angles.

Visibility Comparison Plots

visibility comparison plot For each image generated (each time/energy/algorithm), we have generated a plot like the one to the right, comparing the observed visibilities to the visibilities predicted by the image. The X-axis tick values are the detector number plus the position roll angle (spatial direction of each grid response referenced to solar north) fraction, so e.g. the points between ticks 3 and 4 show the data for Detector 3 as a function of roll angle. The red outlines show which detectors were used to make the image. The Y-axis is the visibility amplitude in photons cm-2 s-1.

The legend includes the time/energy/algorithm used and the reduced chi-square values computed from the measured and predicted visibility vectors weighted by the statistical uncertainties for

  • the detectors used to make the image,
  • all detectors, and
  • each detector separately.
The plot symbols are
  • white Xs - measured visibility amplitudes,
  • red diamonds - predicted visibility amplitudes from the reconstructed image,
  • blue vertical lines - error bars on the measured visibilities, and
  • green triangles - vector difference between the measured and predicted visibilities.

The closer the predicted amplitudes are to the measured values, quantified by the chi-square value, the more reliable the image. In this example, measured and predicted visibilities for detectors 5, 6, and 7 are well matched but significant deviations exist for detectors 3 and 4 reflecting the limited ability of the clean algorithm to accurately reproduce the finer source structure. The deviations for detectors 8 and 9 are an indication of the systematic uncertainties in the relative sensitivities of the different detectors.

Profile Comparison Plots

profile comparison plot Visibility (described above) and profile comparison plots have been generated for all events to provide two convenient ways to quantitatively evaluate how well the reconstructed images would reproduce the measured count rates in each of the detectors.

For each image generated using the CLEAN algorithm, we have generated a plot like the one to the left, comparing the measured count rates to the count rates predicted by the reconstructed image. (These plots are not available for back projection because the image is unitless, or for visibility-based algorithms because the count data is no longer available in the imagecube-generating software after making the visibilities.)

The plots show the individual detector count rates plotted as a function of the sum of the regularized roll angle (spacecraft roll angle corrected for the offset between the spin axis and the mean subcollimator optical axis) and the phase of the modulation in each roll-angle bin. The measured rates are in red and the rates predicted from the reconstructed image are in white. The data displayed have been phase-stacked into rotation angle/phase bins for the duration of the image time interval. There are typically between 6 and 64 roll-angle bins and 12 phase bins within each roll-angle bin. The number of roll-angle bins is larger for the finer grids and is equal to the number of peaks present in the plot for each detector. Thus, the X-axis corresponds to the combination of roll-angle bins with 12 phase bins (0-2π) within each roll-angle bin. The scale is marked with continuous angles for convenience.

The agreement between the measured and predicted rates is an indication of how well the reconstructed image matches the data. This agreement is quantified with the Cash or C-statistic, given as a separate value for each detector and as an overall value for all detectors together.