| 14 Nov 2025
Identifying sea breezes from atmospheric model output (sea_breeze v1.1)
Abstract. The sea breeze is a mesoscale atmospheric circulation that has implications for human activity and wind energy availability in coastal areas. Sea breezes have been studied in many regions throughout the world, with analyses usually identifying them at individual coastal sites based on local characteristics. Therefore, there is currently a lack of robust and generalizable identification methods, resulting in difficulties analyzing sea breeze characteristics over large regions. Here, software is developed that applies three, physically-based diagnostics for sea breeze identification to atmospheric model datasets. The diagnostics are tested across the coastline of Australia over a 6-month period. These diagnostics identify sea breezes based on either a front or circulation, with additional filtering applied to each diagnostic to reduce mis-classifications. The diagnostics are tested on four different model datasets, ranging from 2.2 km to around 25 km horizontal grid spacing, to explore the impact of spatial resolution on sea breeze identification methods. Based on a range of individual cases, as well as statistics of sea breeze occurrences, we suggest that a method based on moisture frontogenesis may potentially be suitable for sea breeze identification from model data. However, results for individual sea breeze cases indicate that there are difficulties associated with separating the sea breeze from other coastal fronts and circulations. These results have applications for quantifying the effect of sea breezes on human activities, such as for coastal wind energy and the modulation of the urban heat island.
The manuscript compares and contrasts three complementary methods of objectively identifying sea breezes in Australia, as applied to different analysis products of varying resolution.  Model sensitivities are described and evaluated through physical argument, and satellite and surface weather station observations. Â
In general, the manuscript is well written with a clear experimental design and evaluation of the results. The authors also do a good job of critiquing the different methods and pointing out their shortcomings. The distance-dependent skill of different metrics (frontogenetic better near the coast and at early times; H less satisfactory in reproducing propagation) are interesting and helpful. Some of the results are not too surprising (analyses with resolution 10 km or finer are needed to properly capture these circulations, although even then apparently smoothing filters are needed due to model noise; topography and convection complicate results). Observational comparison is also limited, primarily based on satellite imagery for a few case studies; station-based evaluation is given in the supplement, but only for overall period frequencies.
I would recommend accepting the manuscript with only minor revisions as described below:
Specific Comments:Â Â
Line 18: I think ‘onshore’ generally means from sea to land. So I would rephrase this sentence to ‘…thermally-direct circulation with onshore flow and ascent over the land with descent over the sea.’
Line 62:Â AUS2200 is a variant of the Australian Community Climate and Earth System Simulator -- for those not familiar with it, could you give more details on what it is, and what it means that it is a 'Simulator'?
Line 94: Why are specific humidity and air temperature given at 1.5 m instead of 2 m? Does AUS2200 have any surface fields, or only atmospheric level fields?
Line 104: This seems to be the only place where you state what specifically the six-month period is. To assist the reader, maybe you could state the period (i.e., Jan-Feb 2013, 2016, 2018) at the end of the introduction?
Â
Line 160: Maybe this is a digression from the main theme of the manuscript which is objective method sensitivity, but I am wondering why the only non-satellite based observational evaluation (the application of H to surface observations) is placed in the supplement, and also why it is used to only evaluate overall frequencies of events over the whole six-month period. I know observational verification of these phenomena is difficult, but could H be applied to the observations for at least the six selected case studies, to get a more fine-grained view of at least the accuracy of H?
Line 196:Â State that the divergence here is the 2D divergence.
Line 273:Â You say that you do not constrain the filter by time of day, but by using a filter that only allows land warmer than ocean, you are inherently only looking at sea breezes vs. land breezes, daytime vs. nighttime, and features over land vs. features over water, are you not?
Line 357:Â Are there any references or data to generally support the geographical distribution of sea breeze events found in the analyses?
Supplement S1: The method to calculate coastline angle seems quite elaborate. But it appears to basically be a filter using length scales that are a function of the resolution of the analysis, at least at sufficient distances from the coastline. But the length scale of a natural phenomenon should be physically based. What would happen if R1 were chosen the same for all analyses? (Maybe ERA5 would be too noisy this way.)
Supplement S3: I'm wondering if it would be possible to do sensitivity tests of this nature on H. Since it is a fuzzy logic algorithm, could you look at the impact of just including temporal jumps in just air temperature alone, or just humidity alone, for example?
Â