The paper  Globally- and Hemispherically-Integrated Joule heating rates during the 17 March 2015 geomagnetic storm, according to Physics-based and Empirical Models”compares Joule heating (JH) estimates by the GITM and TIEGCM models, as well as by several empirical relations, for the storm of March 2015. While JH has various impacts, both fundamental and practical, its quantitative assessment is far from being settled. Numerical exercises like the one in the paper are useful to understand better the various models and eventually how to reach convergence. I shall be happy to recommend publication, once the issues listed below are clarified.

1. A major problem with results like those in the paper is the missing ‘ground truth’. How to decide what result is better, which model to regard as more trustful? The problem is less critical when the various estimates are more or less similar, but this is hardly the case during storms - which are also of highest importance. Obviously, this major problem cannot be solved, but perhaps the authors can elaborate a bit, including hints to one or another more trustful JH proxy, if / when the case.

2. The Weimer driving of GITM and TIEGCM emphasizes the importance of the precipitation model, as aptly discussed and concluded (L359-360, 417, 430-431). This is not surprising, since the precipitation model drives the (height-integrated) conductivity, which is an essential contributor to JH. It would be nice to comment a bit on the matter, including, if possible, a quantitative perspective (to what extent the differences in the results are indeed correlated to differences in conductivity).

3. The differences between the AMIE results of GITM and TIEGCM are much smaller, but occasionally they are significant, like around 22 UT on March 17 (Figure 1 g). What could be the reason? As detailed in Section 2.3, the two codes use similar formulas for the Joule heating, the electric field is the same, and the precipitation, therefore the conductivity, is the same. Is the difference because of different boundary conditions? Different initializations? Different spatial / temporal resolutions, perhaps including different ways of addressing sub-grid? Some mix? Something else?

4. Any hint on why the GCMs provide systematically higher JH as compared to the empirical formulas? (e.g., L369, 427)

5. I do not fully understand panels f and g of Figure 1. The hemispheric powers in panel f, labeled hpower and expressed in GW, do not add up to the global power in panel g, labelled Joule heating and expressed in GWatts (why not GW?). Probably I am missing something. Please clarify, in the caption of the figure and in the para at L283-291. Further on, panels a and b of Figure 2, seem to (rightly) add up to panel g of Figure 1.

6. Others

L10-11: With AMIE, TIEGCM and GITM are rather similar, with TIEGCM occasionally higher (see also point 3).

L66: I think AMIE has a considerably broader scope, not just to mitigate the discrepancies between JH estimates.

L128: Please explain briefly ‘Photoelectron heating is based on a streamlined connection’.

L137: Please describe TIEGCM briefly, similar to GITM one line above.

L138: ‘the equivalence of the two methodologies is derived,’ => Please re-phrase. What is derived / demonstrated is the equivalence of the formulas used to compute JH. The methodologies are not really equivalent, as shown by the different results.

L141: ‘by assuming a quasi-steady state’ => Does this fit with a storm event?

L308: ‘the time of maximum percentage difference’ => Is this indicated by dotted line C?

L314 (and 446-447): ‘shows notable similarities’ => Except for the middle plots of a) and b).

Caption of Figure 1: Include explanation of dotted lines A, B, C, D (e.g., see text?)

L340-341, 367-368, 435: Please re-phrase (replace ‘driven by’ with ‘related to’?). I understand that SME is one of the parameters that drive the simulations, but this does not mean that SME / the electrojet drives JH (in particular, the electrojet is typically dominated by Hall current). Both JH and SME are (mostly) driven by magnetospheric dynamics via M-I-T coupling.

7. Typos and alike

L19, 22, 25, 29, 30, 39, 41, 42, 56-57, 76, 112, 115, 244, 325, 338, 401: References indicated by \citet instead of \citep. L23: Space-X => Starlink (?); L32: found in => related to (?); L34: Delete the bracket after 2023a; L34-35: Move ‘due to the large drag’ at the end of the sentence (it only affects the satellites, not the balloons); L35: Delete ‘current’ (?); L67: Explain LDFF; L78 and 600-601: Duplicate of 598-599; L97: makeS; L99: Eddy => eddy; L118: comma before and after j; L137: outlined in => by; L143: Move ‘component of the electric field’ before ‘parallel’ one line above; L151: Ohm’s LAW; L152: among => along; L181: based ON; Eq. 17: ‘-1’ should be aligned with ‘k’ => z_{k-1}; L185: analysis => techniques? calculations?; L245: (g) => (e); L253: n/cc => cm^{-3}; L257: are => is; L275 and 281: Resolve the question marks; L281: thRough; L305: is shown as a polar plots => are shown as polar plots; L310: Figure 4 => Figure 4 c, d; L312: (c) => (b) at, (b) => (c); Caption of Figure 1, third line: (d) => (f); L330: addresses => addressed.

The paper presents calculations for Joule heating during the well-studied 17 March 2015 St Patrick’s Day storm, using 2 physics-based models – TIEGCM and GITM – driven by 2 different electric field models – Weimer 2005 and AMIE – as well as a range of empirical models. The results reveal a large range of estimates for Joule heating, with empirical models predicting lower Joule heating rates vs the physics-based models. The results also reveal that Joule heating simulation outputs and asymmetries are highly dependent on the auroral precipitation model and electric field model used.

General comments:

The study is generally well-written and well-organised, with a clearly defined methodology and the figures/results supporting the conclusions. The scientific novelty comes from both a quantification of the uncertainty of the Joule heating estimates across various models, as well as the analysis of the causes of the different observed simulation results as a function of model drivers. Joule heating, as argued in the paper, is difficult to directly measure as it occurs in the region of space too high for balloons but too low for satellites. Therefore, simulations and indirect empirical inferences are often used, therefore characterizing and quantifying the uncertainties in these calculations is vital.

I would recommend the paper for publication after the following minor revisions.

Specific comments:

Line 275 “The datasets used in this study are readily available at ?.” – please replace the question mark with the dataset source.

Line 281 “datasets and the code are available though ?” – likewise, please replace the question mark with the datasets/code source.