the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Sep 2026
First spatially continuous fog maps for the central Namib Desert reveal seasonally shifting fog belt
Abstract. Fog represents the major hydrological input in the Namib Desert, yet its spatial distribution remains poorly constrained because ground-based observations are sparse, and there is no direct way to distinguish fog from elevated low clouds in satellite imagery. Here we present the first spatially continuous fog maps for the central Namib Desert derived by integrating satellite–based and ground-based observations. The method combines satellite-based fog/low cloud (FLC) detection from Meteosat SEVIRI with cloud-base height (CBH) estimates derived from observations in the FogNet station network. The regionally derived CBH is then used to separate fog from elevated low clouds. This enables spatiotemporally coherent fog mapping, which we investigate at fog event level and at monthly scales. Validation against independent station-based fog observations shows robust performance across the Namibian FogNet network, with a mean classification accuracy of about 85 % and a Heidke Skill Score of ≈0.6. The CBH constraint substantially reduces the systematic overestimation obtained by using satellite-based FLC detection alone, although some overestimation persists at inland stations during summer months when the cloud base seems to follow the topography. The resulting dataset provides the first observation-based fog maps of inland-reaching advective fog in the central Namib Desert and enables quantitative characterization of the regional fog belt. The maps reveal a fog belt extending ≈10–80 km inland from the coastline and between roughly 100 and 700 m above mean sea level (>25 fog days yr−1), within which a narrower core fog zone occurs between about 20 and 50 km inland and 150–450 m elevation (>90 fog days yr−1). The fog belt shows a pronounced seasonal shift due to the seasonality of the cloud base height. These spatially explicit fog maps provide a basis for identifying ecologically relevant fog zones and for future assessments of non-rainfall water inputs and fog-transported chemical depositions.
Competing interests: At least one of the (co-)authors serves as editor for the special issue to which this paper belongs.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(22127 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-2693', Anonymous Referee #1, 11 Oct 2026
-
RC2: 'Comment on egusphere-2026-2693', Anonymous Referee #2, 16 Oct 2026
This paper utilizes two existing methods identify fog and low clouds from satellite imagery, and then distinguish between fog and low clouds using a CBH estimation methodology. The authors validate this methodology using ground-based measurements, and then provide a climatology of fog frequency in the central Namib. This method provides better agreement between fog identified at the station and satellite derived fog classification, however the false alarm rate is high, particularly for stations at higher elevation. The methodology works best during the winter, when the uniform CBH assumption is least likely to be violated. This paper fills an important gap by providing a spatially continuous climatology of fog along the central Namib desert. It is well-written and the technique shows a clear improvement in agreement between satellite-derived cloud measurements and ground-based fog measurements. My primary comment is that the conclusion section felt a little confused (see comments on conclusions below). Additionally, the paper would benefit from a comparison of the results to prior climatologies that did not distinguish between fog and low clouds.
Comments on conclusions and outlook:
It would be helpful to have more of a discussion on how this climatology differs from other satellite based climatologies that combine fog and low clouds
Is there a bias to the overall accuracies and F1 scores because so many are along a similar longitude/elevation?
Ln 365: “which points towards interesting seasonal changes in boundary layer dynamics”, I understood that seasonal changes in boundary layer dynamics is the explanation the authors posited for why CBH may not be as uniform, as opposed to the over-estimation in fog occurrence demonstrating that the boundary layer is not uniform. None of the figures in this manuscript show CBH variance nor boundary layer dynamics. As such I would re-phrase, or if you are referring to other work please cite.
Ln 367:368 “The characteristic fog zone that emerges from the climatology is a broader fog belt”, I thought that these definitions of a broader fog belt and narrower core fog zone were from ecologically relevant thresholds, as opposed to emerging from the climatology. Could the authors clarify?
Specific Comments:
Please define acronyms, for ex. DOGMA (ln 65) and AMSL (ln 252). Additionally, other acronyms are defined twice, like FAR.
Ln 139 why is fog and no fog only identified during nighttime? Is this an instrument/methods limitation? What % of total fog occurs at night?
Ln 140 “Boundary layer well-mixed at night”, could you elaborate a bit more? Is it well mixed all throughout the night?
Ln 179 What has prior literature shown for how many fog events are identified by ceilometer data, at the stations with ceilometers, versus by fog precipitation measurements? Does this differ between stations based on topography / distance from the coast? I wonder how differences in wind speed with and liquid water content vertically throughout the cloud may impact performance measurements, and if there has been any prior research comparing these fog precipitation measurements to other fog detection methods.
Ln 223 what is the variation of topography within the grid cells for the ground stations? Authors mention they use 3km topographic spacing to match satellite spatial resolution but I’d imagine there is a higher resolution topographic map that can be used to quantify this. This may quantitatively explain performance differences between stations.
231:234 I see the value in Appendix B to show different conceptual CBH representations, but I dont know why it is being discussed in context of the false alarm rate since, as the authors write, “The CBH separation step only partitions already detected FLC scenes into fog or low cloud without altering the underlying set of predicted fog events relative to the ground observations, and therefore does not change the number of false alarms”.
Figure 2: I would say that low clouds are color coded as grey, not black.
Figure 4: could you report the r or r2 for the fits shown in b-e? Also could you clarify what the color bar represents, is it the number of months or number of monthly occurrences of fog/low clouds?
271: Climatologically, what is the average cloud top height of these advected marine stratus clouds? Is cloud top lower or higher in summer or winter? If cloud top does not always extend above the highest elevation points, and if this changes seasonally, this could also explain the breakdown in agreement between NaSa-Fog and station measured fog.
Fig 5: It is a little confusing to switch between low / high fog seasons, representing 8/12 months in figure 5, to showing all 12 months in figure 6. Is there a reason for this?
Fig 6: what is the latitudes that are being averaged over in this figure? Those shown in Fig 5?
318:333 - the discussion of Figure 6, and the implications of the shifting core fog zone and how it compares to prior literature could be expanded. Are the red and black lines in figure 6 from Mitchel et al 2020?
Citation: https://doi.org/10.5194/egusphere-2026-2693-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 340 | 171 | 56 | 567 | 66 | 63 |
- HTML: 340
- PDF: 171
- XML: 56
- Total: 567
- BibTeX: 66
- EndNote: 63
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review: AMT manuscript egusphere-2026-2693
First spatially continuous fog maps for the central Namib Desert reveal seasonally shifting fog belt
This study presents the first spatially continuous fog maps for the central Namib Desert. The authors combine a satellite-derived fog and low cloud (FLC) product with cloud base height (CBH) estimates derived from the FogNet station network, using CBH to distinguish fog from elevated low clouds. The resulting fog classification shows a better agreement with fog observations at both the event and monthly scales. The maps are then used to characterize the regional fog belt and its core zone, revealing a seasonal shift in the position of the fog belt. The paper is well-written, and the contribution is really important as the separation is a well-known problem to understand the actual contribution to the coastal ecosystems. My main comments concern the spatial coherence of the CBH assumption, the lack of an explicit definition of the cloud/fog top, and several points of terminology and thresholds that would benefit from clarification. These are mostly requests for additional support or clarity rather than concerns about the overall approach. General and specific comments are provided below.
General Comments:
The manuscript focuses on distinguishing fog from low clouds using cloud-base height (CBH); however, cloud-top height (CTH) could also provide valuable information on the vertical structure and depth of the fog layer. In the current methodology, the fog–low cloud separation (Step 2) relies on the comparison between the assigned CBH and the terrain elevation, and although the fog belt is defined as an altitude band, a cloud top is not explicitly identified for the cloud/fog layer itself. Although the authors briefly discuss seasonal variations in fog occurrence, it would be valuable to examine whether CTH exhibits similar seasonal variations to CBH and how these variations affect the vertical structure of the cloud layer. For example, Figure 2 could be enhanced by including the temporal variability of CTH alongside CBH.
The authors could consider different approaches to estimate or constrain CTH. For instance, reanalysis products (e.g., ERA5) could be used to derive the height of the temperature inversion or the boundary-layer top, which typically caps the fog and stratocumulus layer in coastal regions. Alternatively, an approximation based on observations from existing field campaigns in the region, such as the NaFoLiCA experiment (Spirig et al., 2019), could be developed and used to validate or constrain the cloud-top estimates. Such an analysis would also help assess the potential of the dataset to distinguish fog from elevated low clouds. If a full CTH analysis is not feasible within this study, I encourage the authors to clarify whether any definition of the upper limit of the fog layer is used or implied in the methodology (e.g., in the definition of the fog belt altitude band), and to discuss explicitly the limitations of not estimating CTH and how it could be addressed in future work.
Abstract:
Line 13-16: This section presents one of the main findings of the study, but the relationship between the fog belt and the core fog zone is not entirely clear (in fog days at least). The fog belt is described as having >25 fog days yr-1, whereas the core fog zone has >90 fog days yr-1. Could the authors clarify whether these values are descriptive characteristics of the two zones or thresholds used to identify them in the maps? Are these thresholds applied to each pixel or do they represent an aggregated statistic for each zone? Furthermore, since the algorithm identifies fog during nighttime, would it be more appropriate to refer to “fog nights” rather than “fog days”? I suggest clarifying these points and reviewing the terminology consistently throughout the manuscript (and figures) to ensure accuracy and avoid ambiguity in the interpretation of the results.
Introduction:
Line 37-39: The fog belt is defined here as an area with 25 fog days annually, but it is unclear whether this threshold is adopted from previous studies or is an operational criterion used in the present study to delineate the fog belt. Could the authors clarify this distinction? If this threshold is used to define the fog belt in the present study, it would be helpful to explain the reason for its selection and how it is applied in the Methods section, ensuring consistent use of the term throughout the manuscript.
Data and Methods
2.1 Satellite FLC product
The analysis period is restricted to January 2015–December 2023 to ensure temporal overlap with the FogNet ground-based observations. Could the authors clarify the temporal completeness of the FogNet records during this period? Were there any data gaps or periods of missing observations at individual stations, and, if so, how were these handled in the analysis? Providing this information would help readers assess the robustness of the ground-based validation.
2.3 Fog-Low Cloud Separation Framework
The entire study relies on the assumption that CBH is spatially coherent across the study domain (Fig. 1a). Could the authors use the available observations to provide further support for this assumption, for example, by comparing CBH estimates from the coastal stations to assess whether they show similar values? Although this comparison would not fully validate the assumption across the entire west–east transect, it could provide additional evidence of regional consistency and help justify the use of the broader domain currently shown in the figures. If the results do not support this assumption beyond the specified domain (22.5–24°S and 14–15.5°E), I suggest limiting all figures to this extent to avoid implying that the assumption also holds outside it.
Could the authors clarify whether the use of a fixed nighttime window (20:00–07:00 UTC) introduces any potential effects on FLC detection or on the relationship between CBH and near-surface relative humidity, given seasonal and spatial variations in sunrise and sunset times across the study region? Could these variations affect the reliability of the estimates at different stations or times of year? In addition, the time window is inconsistent across the manuscript: 20:00–07:00 UTC in the Methods and Conclusions (line 352), but 22:00–07:00 UTC in Step 3 of Figure 1. Could this discrepancy be clarified and the time window used consistently throughout the manuscript?
Line 108 and Line 135: Could the authors clarify the calculation in Step 2? Given that the bias is described as positive, should the correction not involve subtracting rather than adding this value?
Line 142: Could the authors clarify whether the four 15-min time steps must be consecutive or whether non-consecutive time steps can also be counted towards the 1-hour threshold?
2.5 Limitations of fog precipitation data
The authors acknowledge several limitations of using fog precipitation measurements to validate fog occurrence. Could they provide references supporting the use of these measurements for this purpose and further discuss how these limitations affect the interpretation of the validation results? I also suggest reiterating this limitation in the Conclusions.
Results
I suggest defining a consistent nomenclature for seasons or month groups, as the manuscript currently uses terms such as “summer,” “winter,” “spring–summer,” and “ONDJFM.” Could the authors establish a consistent terminology and use it throughout the text?
In Figure 5, the inland boundary of the fog frequency maps (annual, low-season and high-season) appears almost straight and shows little spatial variability, whereas more structure is visible along the coast. Since the terrain contours in the same area are clearly sinuous, it is not obvious whether this pattern reflects the actual fog distribution or is an effect of the methodology (e.g., the spatially uniform CBH, the DEM resolution, the domain boundary, or the detection of the upper cloud limit in the FLC mask). Could the authors clarify this and discuss how it affects the interpretation of the fog belt extent? This is also related to my general comment on the lack of an explicit cloud-top definition. Also, if the fog belt is defined as areas with more than 25 fog days per year, could the authors clarify why areas with fewer than 25 fog days appear to be included within its boundaries?
Conclusions and Outlook
Line 369-370: Please check the consistency of these values throughout the manuscript. The ranges reported here differ from those in the Abstract (20–50 km and 150–450 m AMSL) and should be harmonized.
Could the authors briefly discuss the potential to extend the classification to daytime fog events as a direction for future research? Although the current approach focuses on nighttime conditions, fog can also occur during the day.
Minor comments:
Line 32: Could the authors clarify whether “purely coastal fog” refers specifically to fog or to both fog and low clouds? Also, since this statement concerns the spatial scope of the study and the limitations imposed by the station network, would it be more appropriate to include it in the Methods section rather than the Introduction?
Line 65-71. Could the authors define the acronyms DOGMA and AMARU at their first occurrence? Since these refer to specific algorithms or frameworks, providing their full names would improve clarity for readers unfamiliar with these approaches. This may be less necessary for CALIPSO/CALIOP, which are established satellite mission and instrument names, respectively.
Line 96: Could the authors specify which algorithm is being referred to here?
Line 108-109: The authors attribute this bias to the ceilometer design. Although this explanation has been provided, it would be helpful to include a brief justification in the manuscript, supported by an appropriate reference if available.
Line 147: The phrase “to account for a possibly occurring topography following boundary layer structure” is unclear and would benefit from rephrasing. Could the authors clarify what they mean by this expression?
Line 255-256: Could the authors correct the phrase “in within the panel” in the text?
Line 272-273: I suggest moving this description (using density…) to the caption of Figure 4.
Line 283: NaSa-FLC or NaSa-Fog?