This manuscript quantifies sensitivities of surface energy balance variables to phenological shifts (SOS/EOS) across northern lands using multiple datasets for 2001–2021. The latent heat results are the clearest contribution and are largely consistent across datasets. However, several methodological concerns need to be addressed before the paper is ready for publication.

Major Comments:

1. SOS and EOS are defined using a single threshold of 30% of annual maximum LAI across all vegetation types and latitudes (L78–80). For tundra or deciduous forests with strong seasonal amplitude this may work reasonably well, but for evergreen needleleaf forests — where intra-annual LAI variation is small — the same threshold likely captures noise rather than a real phenological transition. PFT-specific biases in phenological detection could plausibly explain part of the spatial heterogeneity the authors observe in energy balance sensitivities, but this possibility is never examined. A sensitivity test with alternative thresholds (e.g., 20% and 40%) would help, or at minimum the authors should acknowledge this as a source of spatially-structured uncertainty.

2. The partial correlation analysis uses ERA5 surface temperature as a covariate to control for confounding (L125–128). When ERA5 LE and H are the response variables, this is problematic because ERA5 skin temperature and ERA5 turbulent fluxes are products of the same land surface model, where energy balance closure is imposed — they are not independent. The issue is most acute in Section 3.4 where ERA5 LST serves as both the response variable (Figure 6) and the partial correlation covariate. The authors should acknowledge this explicitly and note that ERA5-based partial correlations need to be interpreted carefully. Results from FLUXCOM and GLEAM provide a more independent basis for inference here.

General comment

The manuscript is interesting and addresses the relevant scientific question of how observed phenological changes may feed back to the climate system. The methodology and analyses are generally sound, and the manuscript fits the focus area of the journal well. The conclusions seem well substantiated and balanced. Some clarifications and corrections remain for the manuscript to be ready for publication.

Specific comments

Abstract: The last, concluding, sentence appears somewhat disconnected or even contradictory given the second-to-last sentence and the preceding text. While the LE maps are very consistent, also the EF maps contain areas that are spatially coherent, lending some support to the indication of a negative feedback loop. I suggest expanding the abstract somewhat to express the motivation for the conclusion of a negative feedback more clearly.

Please clarify if detrending is done on all variables including satellite-derived LAI (as stated on L 62)? This is unclear since detrending is explicitly described for the surface balance variables (L 116-117) but not explicitly for the LAI data.

Please provide motivations of the technical choices of a 20-day window and phenology thresholds of 30%.

The conclusion (L254-255) refers to global modelling studies for supporting the conclusion of negative feedback but only provides a reference to the companion paper (?) by Winkler. Please provide examples of other independent studies.

A substantial part of the results and conclusions of the paper are based on interpretation of the maps. However, the limited size and the chosen colour scales limits their interpretability. At least, the size of the maps can be expanded to fit the full text width. The colours are sometimes insufficiently graded, e.g. Fig A1, where it is difficult to match the colours with the actual values in the legend. A separate colour scale for each map would improve legibility. In Fig 2, the colours are saturated with a maximum of 0.002 while values in the box plots below go up to 0.006, resulting in one single colour for the whole northern belt.

The results are presented per climate dataset. While this provides a good structure, the results section could benefit from a final paragraph where the effects of all these variables are summarised per land cover class.

Minor / typographical comments

L 200: Central America -> Central North America

Line 217: “Southern Canada and the central United States…” A bit difficult to see from the map if the region touches Canada; maybe rather “Central and south-eastern North America”?

Line 222, 223: two missing section numbers (see Section ) (Section )

L208-209: The two sentences appear contradictory. Please rephrase.

Fig2, Fig3, Fig 4 all captions: MF mixel forest -> MF mixed forest; CRO coplands -> CRO croplands

FigA2, caption: albendo -> albedo

Fig A4, title: Sensile heat -> Sensible heat