the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Jul 2026
Diurnal cycles of deep convective processes: Biases of ERA5 and DYAMOND km-scale models revealed by satellite observations of frozen water path and precipitation
Abstract. Tropical convective processes show pronounced diurnal variability with strong land and ocean contrasts. Here, we compare their diurnal cycles from satellite observations, ERA5 and km-scale models. As observational base, we use the Chalmers Cloud Ice Climatology (CCIC), a machine-learning satellite product, for frozen water path (FWP) and high cloud cover, and IMERG for precipitation. CCIC captures the full diurnal cycle of FWP in contrast to older satellite FWP products. The satellite observations show late-night or early-morning peaks over the oceans, and afternoon peaks over tropical land areas for FWP and precipitation. The diurnal cycle of tropical FWP and precipitation has a general synchronicity in satellite observations. High cloud cover over tropical land lags behind FWP and precipitation by several hours and peaks during the night in the warm, wet season. ERA5 shows a smaller diurnal amplitude of FWP over tropical land with local phase shifts of more than ± 6 hours compared to CCIC. In contrast, the diurnal amplitude of ERA5 precipitation exceeds that of IMERG over land areas and lacks night-time peaks related to mesoscale convective systems and topography. Km-scale models have smaller diurnal phase shifts; however, regional biases exist over the South Pacific Convergence Zone, as well as systematic differences in diurnal amplitudes over tropical land areas, which can be quantified with satellite observations like CCIC. Our analysis illustrates the power of combining different satellite products to better understand the diurnal cycles of tropical convective processes when evaluating reanalyses or models.
- Preprint
(3284 KB) - Metadata XML
-
Supplement
(7406 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-3869', Anonymous Referee #1, 20 Aug 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-3869/egusphere-2026-3869-RC1-supplement.pdfCitation: https://doi.org/
10.5194/egusphere-2026-3869-RC1 -
RC2: 'Comment on egusphere-2026-3869', Anonymous Referee #2, 21 Sep 2026
Review of manuscript egusphere-2026-3869 titled “Diurnal cycles of deep convective processes: Biases of ERA5 and DYAMOND km-scale models revealed by satellite observations of frozen water path and precipitation”
By Leko et al., submitted to ACP
Overview: This study leverages a machine learning-derived algorithm for ice cloud, IMERG precipitation, ERA5, and DYAMOND global km-scale simulations to document the diurnal cycle and examine biases in models and reanalysis. The ML-based ice cloud product provides estimates of frozen water path (FWP) and high cloud cover (HCC), with the algorithm trained on CloudSat and CALIPSO data, carried out by prior studies. This ML-based algorithm provides a novel opportunity to describe the (near-)global diurnal cycle from satellite estimates, albeit with significant error bars. In short, the algorithm exploits active remote sensing measurements to constrain ice behavior, which is sparsely sampled, but scaling this up via geostationary satellite observations. Their results tell a story of substantial biases in ERA5, especially over land, with greatly underestimated diurnal amplitude of FWP despite overestimated rainfall amplitude. While they demonstrate some improvements relative to ERA5 in the km-scale models, they surprisingly perpetuate some issues: both ERA5 and the km-scale models suffer issues representing the diurnal cycle in the continental regime downstream of major mountain chains (east of the Rockies and Andes). One might have otherwise assumed that such issues would be addressed by circumventing cumulus parameterizations, alas, no. This is an important result.
Reviewer recommendation: accept with minor revision. The study is overall quite clear, compelling, and easy to follow. There are some areas where clarification would be valuable, as outlined in my comments below. Since these comments primarily emphasize clarifications, I do feel that they can be addressed quite readily without much additional work. Aside from those issues, I recommend publication.
Comments:
The intro would benefit from an additional paragraph that succinctly describes/summarizes the purpose of the paper, potentially outlining science questions that the study will address, hypotheses that will tested…as appropriate. That should be followed by a succinct description of how you will achieve those goals. I personally do not find end-of-intro outline paragraphs to be necessary or helpful. I skip them. Always. An option you might consider is replacing that outline paragraph with what I am suggesting, if you agree. It would provide an ideal segue into the methods.
CCIC: Something I am unclear on is what species and what clouds we should assume are dominant in this product, including both FWP and HCC, and especially if the answer to that differs between these. Namely, is FWP expected to be dominated by large precipitating ice during times of deep convection? If so, does that potentially imply that deep convection should overwhelm any diurnal signal of smaller suspended (more “cirrus-ey” type) ice? Elaborating on how we should interpret these products would substantially help clarify things here.
Very much related to the above: I am quite baffled by how out-of-phase the diurnal cycle of HCC can be from FWP. Can you explain this? I saw some references to other papers when this was discussed, but short of fully reading those other papers I could not find clear explanations. It does not seem to be explained merely by a lingering of high clouds following peak rainfall (e.g., the lingering of anvils from deep convection). The core of the issue here links to the above comment: I do not know what HCC physically represents. Let me walk through a hypothetical thought process to perhaps underscore my confusion. Perhaps this will provide you a case study from which to identify where clarity will help save future readers. Let’s assume there is a diurnal maximum in FWP at a given LST, say 15:00, closely corresponding with a peak in rainfall as you show to often be the case. If this is also the time of minimum HCC, this tells us that there is a maximum in the total integrated ice in the sample grid column, but concentrated over small patches of cloud, which indeed have minimum coverage in that same grid column at that time. Is this a correct interpretation of these two products? I am not arguing that you have to reconcile every puzzle: some of the observations your study documents may require more time to understand. But more discussion of how we should interpret these things, and/or emphasizing when something indeed seems to conflict with intuition, will help readers feel like they are not crazy and will make some of the areas for future work better stand out.
Have you tried a simpler composite approach to compare against the Fourier transform? The diurnal cycle does not have to and indeed does not always conform nicely to sinusoidal structure, so it would be worth at least testing more a general compositing approach to convince yourself that features are not being lost by the Fourier fit.
I found a number of quite minor grammatical errors throughout the manuscript that can be readily addressed by carefully combing through.
Specific comments:
L21: While I agree that it is not surprising that it is imperfect, I do not feel that the specific phase error is expected or intuitive.
L164: is this a continuous time adjustment or using even 15deg increments?
L460-464: For clarity I think it’s important to note that it has both the observed peak captured relatively well and an additional peak in the northern hemisphere, repeating the well-known double ITCZ problem down at km-scale.
Fig. 1: the lines for CCIC and 2C-ICE are difficult to distinguish based on their colors. Please consider other ways to help these stand apart.
Fig. 2b: it might be preferable to use a color map with white separating red and blue to help distinguish from gray regions of excluded data.
Citation: https://doi.org/10.5194/egusphere-2026-3869-RC2
Data sets
Data for: Diurnal Cycles of Deep Convective Processes in Satellite Measurements, the ERA5 Reanalysis and DYAMOND km-scale Models Gunnar Behrens et al. https://doi.org/10.5281/zenodo.20842368
Interactive computing environment
Code for: Diurnal Cycles of Deep Convective Processes in Satellite Measurements, the ERA5 Reanalysis and DYAMOND km-scale Models Lara Leko et al. https://doi.org/10.5281/zenodo.21041035
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 266 | 120 | 26 | 412 | 68 | 28 | 32 |
- HTML: 266
- PDF: 120
- XML: 26
- Total: 412
- Supplement: 68
- BibTeX: 28
- EndNote: 32
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
