The paper we are working on was accepted and we got a comment saying "Quality of Figures could be improved". All the figures that are given have been exported from matplotlib and have all the standard labels with different color schemes and have attached the sample as well. A sample of a graph used So what do I do? make it more pretty? or use a different package?

Note - This is just a sample, and the real one has a set of 6 plots big enough to fit in almost half a page

  • Not my area. Why does it look blurry?
    – VitaminE
    Apr 10 at 9:17
  • 5
    The "time" labels are inconsistent, and you probably need to increase the resolution.
    – Louic
    Apr 10 at 11:12
  • 10
    The figure aught to be pixelated at the level the underlying data is pixelated. Do not listen to anyone who tells you to smooth it. Apr 10 at 15:18
  • 5
    The axis labels should all be written out in full, with correct capitalization, and units. The tic marks should point out of the boxes so that they do not cover up the data. Apr 10 at 15:20
  • 3
    There is no reason to use a different package. Plenty of room to improve the figure using matplotlib.
    – Anyon
    Apr 10 at 17:58

2 Answers 2


The illustration you provide seems like a standard plot that can be seen in most journals. That said, it can certainly be improved; many journals do not seem to care what they publish. What is acceptable also varies from person to person. There are however some standards that should be followed and your figure fall short on a few points. Many of the comments made to your post outline some of the improvements and I will also reiterate them here.

  1. A figure should be understandable when shown without its figure caption. Your figure falls short of this. By understandable I do not mean the deep scientific significance of what is shown but to understand what is shown. This is a point made by Tufte (2001) and is a good guideline for finishing any figure.
  2. Axis labels should include units. You have latitude and longitude which would be in degrees. The color scale is more problematic because I do not understand what it should be showing, units would obviously help.
  3. Axis labels should start with capital letters so Lat/Lon or Latitude/Longitude.
  4. The heading of the subplots containg a time stamp should be consistent. The ISO standard for date/time is yyyy-mm-dd HH:MM:SS. So I would suggest using this consistently or (which might be better for a more general audience) use an unambiguous date and time format such as "dd month, yyyy, HHMM hrs" (e.g. 29 April, 2022, 1200 hrs). The problem with date and time is that almost every country has their own standard.
  5. The third diagram in the top row has no x-axis label and tick labels
  6. The number of tick labels could be reduced so that the axes are not overloaded with information. Every second label can be removed while keeping the tick marks
  7. The resolution of the coloured fields should reflect the resolution of the data shown in the fields. If the data is as coarse as it looks in the figure, then that is how it should be shown.
  8. Figures must be made to work in their final size in the publication. Submitting vector based formats avoids many problems. When exporting to a bitmap one should consider the necessary resolution (in dpi) to make the illustration sharp in its final size. A figure such as yours that include bitmap fields should follow the same guidelines but obviously the fields in your subfigures will not change in apparent resolution as was stated in the previous point.
  9. The data is latitude/longitude based running from c. 10 degrees N(?) to c. 80 degrees N(?). The width of the space between longitude lines will obviously decrease from S to N so that a square representation is not strictly correct. The plots are better representing reality if they are shown in some cartographic projection. This is clearly something many will ignore and maybe not even consider but it is one aspect to consider.
  10. Any graphing package or software is not perfect. It is always useful to learn to use for example the open source Inkscape to edit standard output from plotting packages or software to optimse your graphics before publication.

Considering the colour scale I cannot see anything wrong. Aspects to consider regarding choice of colour scale include what is standard (if any), what may make sense depending on what is shown and what can be read by someone with colour blindness.

So there are several aspects that can be improved or at least considered. At the same time providing a comment such as the one you appear to have received without any exemplification of what to do is quite useless on behalf of the reviewer/editor.


Tufte, E.R. 2001. The visual display of quantitative information. Second Edition. Cheshire, CT: Graphics Press.


The figures now are just a blob of colors with poor titles and bloated aspect. If you are an expert on the exact topic of your publication, you would be able to find the secret message hidden inside. So there is almost no advantage between having that figure and having a table showing the raw data: a figure serves exactly the opposite scope, make evident the messagge hidden behind data.

First) the colors. Why on Earth you have this mild but oversaturated range of colors if your scale is going symmetrically from +value to -value? these figures are figures that are hard to read, while figures should be popping out. There is simmetry, please have a strong contrast without saturation, no one sane of mind will look at your data in black and white, so you can use something like the RdBu scale (Red for positive, Blue for negative, white for zero)

Second) the colors. Why the colorscale is continuous, while the data seems not (yes, spatial discretization may play a role ... but what about resoultion of the measurments?)? and the scale truncated but not limited going from +32 to -32 ... please, use values readable from humans, like +30 to -30. Base 8 is nice, but leave it to your DAC sensors.

Third) the colors. They carry an overload of information, but what you want to show is "something". Please add some isolines in your plot (it is easy to do if the colorscale is a slightly desaturated RdBu instead of this very saturated colormap). And if the isolines have a label, plus dashed if negative and continuos if positive, you are good to go for the 1% of the world still printing papers and accessing papers only in black&white.

Fourth: the title must be in a standard format, whatever standard your peers are using, then please avoid writing "time": either it is obvious that it is a timestamp, or you have to find another format of timestamp. Do not strive making your best to be as concise as possible. Just be concise.

Fifth (bonus) if you use a lighter colormap, you can plot the grid with some dashed lines, to improve readability.

Example of colormap: enter image description here (taken from here, please note that all the text in this figure is bad, that the grid is not plotted but at least colormap and isolines are what would make much better your figures)

Not the answer you're looking for? Browse other questions tagged .