Abstract

Fog is a high-impact weather phenomenon affecting human activity, including aviation, transport, and health. Its prediction is a longstanding issue for weather forecast models. The success of a forecast depends on complex interactions among various meteorological and topographical parameters; even very small changes in some of these can determine the difference between thick fog and good visibility. This makes prediction of fog one of the most challenging goals for numerical weather prediction. The Local and Nonlocal Fog Experiment (LANFEX) is an attempt to improve our understanding of radiation fog formation through a combined field and numerical study. The 18-month field trial was deployed in the United Kingdom with an extensive range of equipment, including some novel measurements (e.g., dew measurement and thermal imaging). In a hilly area we instrumented flux towers in four adjacent valleys to observe the evolution of similar, but crucially different, meteorological conditions at the different sites. We correlated these with the formation and evolution of fog. The results indicate new quantitative insight into the subtle turbulent conditions required for the formation of radiation fog within a stable boundary layer. Modeling studies have also been conducted, concentrating on high-resolution forecast models and research models from 1.5-km to 100-m resolution. Early results show that models with a resolution of around 100 m are capable of reproducing the local-scale variability that can lead to the onset and development of radiation fog, and also have identified deficiencies in aerosol activation, turbulence, and cloud micro- and macrophysics, in model parameterizations.

A collaborative field and modeling study used a small system of valleys as a natural laboratory to study the formation and evolution of fog.

Atmospheric fog can have a high impact on human activity, particularly transport (Gultepe et al. 2007). Delays due to poor visibility can be extensive and costly. Agarwal et al. (2005) estimated that fog causes a decrease in vehicle speed of 6%–12% and a reduction in traffic capacity of 10%–12% on freeways in Iowa. Figures for the impact of fog on aviation presented by Robinson (1989) indicate that the cost to an airline of a single day of heavy fog at a single airport was about $92,000 (U.S. dollars; about $200,000 when adjusted for inflation). It is likely that current costs are much higher because of the expansion of the aviation industry. To some extent the disruption from fog can be mitigated if events are correctly forecast by weather service providers, allowing suitable preparations to be put in place. According to Vautard et al. (2009), between 1980 and 2005 Europe experienced approximately 40 days per year with visibility <5 km. Given that their analysis was based on data taken at four times per day only (0300, 0900, 1500, and 2100 UTC), we expect this figure to be an underestimate. Thus, the frequency of occurrence is sufficient for fog to be a high forecast priority.

Despite a long history of fog research [Taylor (1917) is an early example], numerical weather prediction (NWP) models still require improvement to more accurately forecast fog (Tudor 2010; van der Velde et al. 2010; Steeneveld et al. 2015; Boutle et al. 2016). Other techniques including machine learning (Bartoková et al. 2015; Herman and Schumacher 2016), use of selected local observations (Haeffelin et al. 2016), use of stochastic physics (McCabe et al. 2016), and statistical methods (Román-Cascón et al. 2016; Menut et al. 2014) may be used to supplement deterministic NWP forecasts.

Notwithstanding the difficulties in modeling fog, there remain gaps in our understanding of this phenomenon. For example, the initial formation of fog has been discussed previously, including by Rodhe (1962), Roach et al. (1976), Duynkerke (1999), Bergot (2013), Nakanishi (2000), and Zhou and Ferrier (2008). Some of these studies proposed different formation mechanisms. While most of these studies agree that very low levels of turbulence are required for radiation fog to form, Rodhe (1962) and Duynkerke (1999) proposed that increased levels of turbulence mixing nearly saturated masses of air at different temperatures are responsible.

Radiation fog can be defined as fog that forms principally as a result of radiative cooling at the surface and the air immediately above it. It therefore normally forms in a nocturnal stable boundary layer (SBL) and is initially thermally stable. In the United Kingdom it is a very common type of fog. The normal processes present in SBLs, such as drainage flows, are expected to influence the development of the radiation fog, but radiation is the root cause of the fog.

Once radiation fog has formed, its subsequent evolution is also of significant interest. For a site in southeast England, Price (2011) identified that when radiation fog occurred, it developed into deep, optically thick fog (defined here as being opaque to thermal radiation in the 8–12-µm range) in approximately 50% of cases. The other 50% remained shallow, optically thin (i.e., transparent to thermal radiation in the 8–12-µm range), and often inhomogeneous. The latter often remained less than 50 m deep, whereas the former usually reached 100 m deep or more. While shallow, optically thin fog remained thermally stable, the deeper, optically thick fog normally developed a saturated adiabatic temperature profile (due to surface warming and fog-top cooling). This allowed more turbulence and weak convection to readily form. In this manuscript we define the term “shallow fog” to refer to shallow radiation fog that is optically thin and thermally stable and “deep fog” to refer to deeper optically thick fog with a saturated adiabatic temperature profile. Thus, shallow and deep fog are quite distinct. Shallow fog is usually not long lived (generally less than 8–10 h) and often dissipates during the morning (even in winter). Deep fog is normally more persistent and can last 24 h or more, with therefore much greater potential to cause disruption to human activity. Identifying whether fog is likely to remain shallow or become deep is clearly an important forecasting goal.

However, uncertainties remain regarding the mechanisms and subtle interactions responsible for fog formation, growth, and development. For example, the relative importance of in situ development versus nonlocal advective development, along with the conditions that might favor either of these, is not yet clear. Advective effects have been observed to be important for fog development in a number of observational studies, including Guedalia and Bergot (1994), Ye et al. (2015), and Price et al. (2015), who observed the rapid spread of a fog layer caused by a gravity current at its edge.

Previous field studies investigating fog include Roach et al. (1976), who discuss early observations at Cardington, United Kingdom. Fitzjarrald and Lala (1989) discuss an experiment in the Hudson valley, Fuzzi et al. (1992) give an overview of the Po valley fog experiment in Italy, Gultepe et al. (2016) examined ice fog occurrence in a mountain valley environment, and Cuxart and Jiménez (2012) show results from a study in the Ebro valley, Spain. While all of these studies provided valuable insight into the behavior of fog, precise details that govern its initial formation, and subsequent development of stable radiation fog into deeper adiabatic fog, remain elusive in the literature. The Local and Nonlocal Fog Experiment (LANFEX) field campaign was designed to address some of the outstanding questions relating to the formation, development, and dissipation of radiation fog, as discussed above, in both an observational and modeling context. The broad objectives are to

  1. better understand the sensitivity of radiation fog formation to turbulence, humidity, and dew deposition;

  2. better understand the factors affecting the vertical growth of radiation fogs and their potential to transition from stable shallow fog to deeper fog with a saturated adiabatic temperature profile;

  3. better understand the relative importance of local and nonlocal processes (such as drainage currents) on radiation fog;

  4. assess the current performance of both forecast and research models using bespoke quality observations; and

  5. develop improved model parameterizations leading to more accurate forecasting of fog.

The experiment ran over an 18-month period, from September 2014 to March 2016, with over 300 instruments deployed at 18 sites in Shropshire and Bedfordshire, United Kingdom. A particular emphasis was to deploy a network of sites over a relatively small region (∼20–30 km) to assess high-resolution models. The campaign was a collaboration among the Met Office, Météo-France, and the Universities of East Anglia, Leeds, Manchester, and Hertfordshire. LANFEX is a development from the Cold-Air Pooling Experiment (COLPEX) campaign (see Price et al. 2011), which examined the dynamics of nocturnal cold pools in the same Shropshire region.

DESCRIPTION OF LANFEX SITES AND INSTRUMENTATION.

The field campaign was based in two regions of the United Kingdom: Bedfordshire in southeast England and Shropshire in western England. The former was centered at the Met Office’s U.K. research site at Cardington. Figure 1 shows both locations. Cardington (52°6ʹN, 0°25.5ʹW) is located in a wide, shallow valley characterized by a patchwork of mostly arable fields with low hedges (field size about 50 ha). The valley is approximately 10 km wide at this location and essentially flat across its width. The ground at the valley sides rises approximately 40 and 30 m (see Fig. 1), and the down-valley gradient is 1:375 or 0.15°. The rationale of using this site was that it represents a reasonably homogeneous landscape, representative of the region with relatively shallow orography, and it offers the opportunity to study fog that may be dominated by in situ development rather than advection.

The Shropshire region (centered on 52°25.2ʹN, 3°6ʹW) was chosen for its network of valleys and small hills. These range in valley-to-hilltop heights of approximately 100–150 m and valley widths of 1–4 km. Land use is mostly pasture with low hedges (field size about 10–20 ha) and some forestry. The rationale of choosing this area was to use the valley network as a natural laboratory. During overnight SBL conditions, each valley was expected to experience subtly different meteorological conditions that may have induced formation of fog at certain locations but not others. When studied over a suitable time scale (in this case, 18 months), a picture can be built wherein the more important meteorological processes leading to or affecting fog formation can be identified and quantified.

Figure 1 also shows the orography in the two regions and the locations of instrumented sites. Note that most of these sites are located in valley bottoms, as radiation fog usually forms in the SBLs that can develop there on calm, clear evenings. However, a few sites were located on hilltops, including one extensively instrumented site (Springhill), and these were intended to measure conditions over the valley tops as well as detecting whether fog development reached the hilltops. A full list of the approximately 300 sensors deployed is not given here, but we provide an overview of the site instrument deployments in Table 1. The characteristics of the main measurements are given in Table 2. The deployment included a suite of remote sensing equipment, the most useful of which to date have been the infrared (IR) camera (fog dynamics and temperature), microwave radiometer (integrated liquid water path), and Doppler lidar (SBL dynamics). Example data from the IR camera are presented below.

Two types of site were deployed: main sites and fog-monitor sites. Main sites were based around larger towers (either 10, 16, or 50 m), were extensively instrumented, and were capable of measuring the surface energy balance. Fog-monitor sites were smaller weather stations based around a single 2.5-m mast (see below). Some sites were additionally equipped to allow for the release of radiosondes or operation of a tethered balloon. Flux measurements were based around Gill HS50 sonic anemometers and LI-COR LI-7500 or Campbell Scientific Krypton hygrometers.

Figure 2 shows valley profiles for the five main Shropshire sites, illustrating the different valley geometries and heights above mean sea level (MSL). Note that the aspect ratio of this plot (vertical to horizontal) is approximately 13:1, which greatly exaggerates the apparent gradient of the slopes. Vosper and Brown (2008) and Price et al. (2011) illustrated the effect of valley geometry on the formation of SBLs and thus radiation fog. Narrow, deep valleys tend to decouple from the overlying flow (such that there is minimal momentum or heat exchange between the two) early during the evening and hence can cool radiatively from an earlier time compared to a wider, more open valley. This may influence the onset of fog. These results were taken into consideration when selecting sites for LANFEX.

Some novel instrumentation was deployed during LANFEX, including a set of dewmeters. Dewfall can remove excess water vapor and supersaturation from the air mass next to the ground and thus inhibit fog formation. Additionally, deposition of fog droplets is an important process that not only directly influences visibility but can determine the fate of fog (Mazoyer et al. 2017) and is thus important to quantify. Price and Clark (2014) described an instrument to measure dew and fog droplet deposition, and several of these instruments were deployed during LANFEX. The devices work by measuring the weight of water deposited onto a canopy, which may be made of real or artificial grass (identical artificial canopies were used for LANFEX to aid comparison between sites). A typical deployment is shown in Fig. 3.

An element of the LANFEX campaign was to build and deploy a number of small weather stations (at fog-monitor sites) with two important capabilities, namely, the ability to measure light winds (using a Gill 2D sonic anemometer) and also to measure a fog droplet spectrum using a newly developed (prototype), inexpensive optical particle counter (with 28 bins from 0.8 to 40 µm). These devices proved effective at unambiguously identifying the presence of fog, although it remains to be established how well they captured quantitative microphysical information. Winds can be measured down to a few centimeters per second, allowing light SBL winds to be monitored. The stations also measure basic mean meteorological parameters: temperature T, relative humidity (RH), and pressure P. The meteorological station was designed and built by the Met Office Cardington and the fog spectrometer by the University of Hertfordshire. A deployment of a fog-monitor station is shown in Fig. 4.

Infrared (IR) cameras were deployed during the COLPEX campaign (Price et al. 2011) and found to provide useful information on local variations in canopy (i.e., surface) temperature throughout the night. During LANFEX, IR cameras were deployed at Cardington and Skyborry, United Kingdom, to observe the development of the SBL and fog. These sites were chosen to contrast the significantly different orography at each location. The camera at Cardington operated continuously, while that at Skyborry was deployed during intensive observation periods (IOPs; see below). The cameras were manufactured by FLIR Systems and are sensitive in the 8–12-µm region. An example still image of a nighttime fog case is shown in Fig. 5, but the real power of these devices is to create overnight video sequences of fog evolution (we took one image per minute to do this). In this way, the evolution of the SBL and dynamics of the fog often become very apparent, with gravity waves and “sloshing” of the air within a valley clearly visible. Importantly, it is possible in some instances to see clearly whether a fog forms and develops in situ or whether advective effects dominate. An example video file of a night time fog at Skyborry is available as supplemental material to this paper (https://doi.org/10.1175/BAMS-D-16-0299.2).

Most measurements were collected continuously, and, where possible, equipment deployed in Shropshire was monitored from Cardington via satellite link to check correct operation. Regular visits were carried out (at least once per month) to perform a full equipment check and visual inspection at each site. At certain times throughout the campaign, IOPs were conducted, during which extra measurements were collected via a tethered balloon, radiosondes, and an infrared camera. In Shropshire, a small tethered balloon (21 m3) carried the same type of droplet probe fitted to the fog-monitor stations (attached to the tether cable). At Cardington, a much larger balloon (51 m3) was flown that carried a Droplet Measurement Technologies (DMT) cloud droplet probe (CDP). The IOP periods were chosen because fog was expected at these times. However, we also conducted some IOPs in clear SBL cases (i.e., no fog or significant cloud present) to make comparisons with the foggy cases.

DATA AVAILABILITY AND IOPS.

Instrument calibrations and intercomparisons were performed at Cardington before and after the field deployment to ensure that the quality of collected data could be verified. A record of data availability was kept that revealed that instruments produced good data for 75%–97% of possible instances, with most instruments producing 80%–90% of possible data. During the deployment, 19 IOPs were conducted (numbered 1 to 19). Of these, 12 cases experienced some fog (defined when measured visibility < 1 km for at least 30 min) and 7 were clear-sky cases. None experienced snow cover, but several experienced air frosts. IOPs ranged in scale, with some featuring the release of just one or two radiosondes from a single site and others featuring 10–20 radiosondes released over a night (at two sites) with the infrared camera and tethered balloon system also deployed. Some foggy periods occurred when no IOPs were conducted.

POWERING A FOG-MONITOR SITE

Fog-monitor sites are autonomous and powered by solar panels and batteries. The power management of these devices was a particular challenge during the rather dark and cloudy winters experienced. Stations were designed to shut down power-hungry applications, such as the fog spectrometer, when power was low. In addition, fog spectrometers were only switched on when RH was greater than 96% and switched off when it decreased below 94%, to save power. Using this technique, data availability was kept above 87%.

INITIAL ANALYSIS OF FOG OBSERVATIONS.

Initial statistics for the Shropshire sites have been gathered and can be compared with those from Cardington, for which fog morphology and statistics of fog are already well defined (Price 2011). During LANFEX, 27 cases of fog were identified over the Shropshire region; these cases were defined as those that formed overnight in initially clear-sky conditions and lasted at least 30 min at one or more sites. Table 3 shows some basic comparisons between the Shropshire sites, including which sites experienced the most fog and dense fog (visibility < 200 m). Jaybarns and Skyborry were the foggiest sites, but Jaybarns developed the densest fog by a significant margin. Recall that Jaybarns lies within the widest, most open valley. A preliminary analysis of temperature (not shown) indicates that, consistent with the findings of Vosper and Brown (2008), the temperature in the narrower valleys sometimes falls more quickly than the wider valleys during clear, calm evenings. One conclusion from our findings so far is that this more rapid cooling does not bias the narrower valley sites to being foggier; in fact, we have found the opposite. We also noted that when Jaybarns did cool more slowly during the evening SBL period, it was usually observed to continue cooling for longer into the night, and, in many cases, by 0200–0400 UTC, it was often then the coldest site.

Note that some occurrences of fog at the hilltop site Springhill could be due to the formation of low nocturnal stratus cloud, which then intersects the hill (this was observed on numerous occasions). However, note also that the occurrence of nocturnal stratus and stratocumulus clouds on clear nights was observed for some of the IOPs and was seen to inhibit the formation or persistence of fog in the valleys. This was observed for a brief period during IOP 12, presented below (Fig. 6).

Of the 27 fog cases identified, deep fog was observed to develop from shallow stable radiation fog in approximately 50% of cases. This figure is the same as that found for Cardington, Bedfordshire. It would seem, therefore, that the larger scale of the orography in Shropshire, compared to Cardington, had little influence on whether fog develops into a deep fog. However, since there are also small hills (approximately a few tens of meters in height) surrounding the Cardington site, we cannot conclude that orography is not important to the development of deep radiation fog. The time at which deep fog developed varied among sites, occurring first at Skyborry (average time ∼0400 UTC) and last at Jaybarns (∼0520 UTC). Other factors, such as the local vertical humidity gradient, may be important in controlling the vertical growth of fog, but further study is required to verify this.

If, at the first occurrence of fog, the temperature profile within the fog is saturated adiabatic (and thus is deep fog), this may be an indication that it has not developed in situ but was of nonlocal origin and advected over the site. This is because we expect in situ formation of radiation fog to occur in an SBL and therefore be thermally stable. Some fog observed during LANFEX fell into this category and was either deep from onset or very quickly became deep (within 1 h). The rapid transition from a stable to a saturated adiabatic temperature profile can be expected when a deep fog layer propagates into a clear stable region and mixes with it. Since the fog layer will cool radiatively more quickly than clear adjacent air, it may form a gravity current and flow into the clear region (Price et al. 2015). The analysis indicates that these advecting deep fog events were relatively rare in the Shropshire region but were more common at Pentre than elsewhere. Of the 10 fog events at Pentre, half were advecting deep fog. The only other site to experience advecting deep fog was Springhill, with one incidence. Studying the movement of fog throughout the Shropshire valleys will form an important element of future work.

Although a full discussion of the effects of turbulence falls outside the scope of this article, we briefly present some data to illustrate our initial findings. As discussed, most authorities agree that turbulence levels must be sufficiently low for the initial formation of radiation fog. The LANFEX database has the capability to offer guidance on the magnitude of turbulence under which fog may form. As an example, Fig. 6 shows vertical velocity variance (ww) measured at 2 m above ground at some of the main sites during IOP 12 (1–2 October 2015). We have chosen ww since it relates to the vertical transport of heat and humidity that will directly affect fog formation. IOP 12 was conducted in Shropshire during a persistent surface anticyclonic period that lasted several days. The boundary layer was relatively moist with high RH. Winds were light all night, with clear skies, except for a cloudy interlude of low stratocumulus for approximately 2 h part way through the night. Fog formed before and after this interlude at some of the sites. We can see the expected decrease in turbulence during the evening and overnight and also that Springhill, the hilltop site, remained more turbulent than the valley sites (though turbulence there did drop to almost the same levels as in the valleys between 0300 and 0615 UTC). Note that in each case of fog formation, ww at 2 m was at or below the value of 0.005 m2 s–2. Future presentation of further analysis will confirm whether this result is typical.

In contrast to measurements at Jaybarns and Skyborry, note in Fig. 6 that when fog appeared at Pentre, it was accompanied by a very rapid increase in turbulence (at the same time the temperature profile changed from stable to approximately saturated adiabatic). This is an example of an established deep and more turbulent fog advecting over the site. This fog probably started out life as a thermally stable, shallow radiation fog, deepened and developed a saturated adiabatic temperature profile, and then propagated into adjacent clear regions as a gravity current (Price et al. 2015). IR video clips (e.g., see the supplementary video taken at Skyborry) would appear to support the theory that this can occur in the Shropshire valleys. However, further analysis is required to confirm this. Note that during the cloudy interlude, turbulence remained weak at all valley sites (the boundary layer remained stable), but fog dissipated because of warming.

MODELING.

A key aim of LANFEX is to improve the skill of NWP fog forecasts, and therefore the modeling activities are integral to achieving this. The way the observations discussed above will be utilized to do this can be broadly classified into two themes:

  1. Improve the physical parameterizations used in NWP models—microphysics, turbulence, cloud macrophysics, land surface, radiation, and aerosol representation are all key to producing correct fog forecasts, and improvements to any of these will be beneficial

  2. Understand and evaluate the sub-kilometer-scale models that are starting to be used for NWP forecasts and will form the next generation of NWP models as computational resources grow—for example, the London Model (Boutle et al. 2016) or Delhi Model (Jayakumar et al. 2018)

The first of these themes is a traditional aim of any field campaign, but some of the new instrumentation deployed during LANFEX, such as the dewmeters discussed above, should help to provide additional constraints on model parameterizations not previously available. The second theme is a distinct feature of the LANFEX design, instrumenting a range of nearby sites within the domain of a ∼100-m-grid-length NWP model to evaluate the local-scale variability of fog and its representation in the model. This is something that has not been looked at by previous fog campaigns (e.g., Gultepe et al. 2009, 2016), which focused primarily on kilometer-scale NWP models.

Here, we present a comparison of IOP 12 as simulated by two models—the operational Met Office Unified Model (UM) and research model Meso-NH, which contains identical physics to the French operational Application of Research to Operations at Mesoscale (AROME) model. Both models use a grid length of 100 m over domains similar to those presented in Fig. 1. The UM is initialized from its own analysis (at 1.5-km grid length) at 1200 UTC and uses a similar scientific configuration to that used in previous studies of fog and complex terrain (Vosper et al. 2013; Boutle et al. 2017). Meso-NH (Lac et al. 2018) is initialized from the 1.3-km AROME analysis (Seity et al. 2011), with an intermediate nest at 500-m grid length also being used and the 100-m domain not being initialized until 1800 UTC. Both models contain similar physical parameterizations, including a 3D turbulence representation, interactive land surface parameterizations, a subgrid cloud condensation scheme to allow partial cloudiness of a grid box, and a single-moment microphysics scheme including droplet sedimentation. The key differences come from the vertical resolution, which is much higher in Meso-NH (148 levels below 1,400 m, compared with 43 in the UM); the inclusion of droplet deposition processes in Meso-NH (von Glasow and Bott 1999; Mazoyer et al. 2017) but not the UM; and the inclusion of a single-species prognostic aerosol in the UM (Clark et al. 2008), used to represent aerosol–cloud interactions, compared to a fixed cloud droplet number concentration in Meso-NH.

Figure 7 shows the predicted visibility fields at 0600 UTC from each model. The fog is well constrained to the valley systems in both models, with hilltops (such as Springhill) remaining correctly fog free. However, there are interesting differences between the models. At Jaybarns, where both models are predicting low visibility, the UM values (<100 m) appear in better agreement with the observations than the Meso-NH values (∼200 m). The method of predicting the visibility differs between the models, being a simple function of water content in Meso-NH (Kunkel 1984) compared with a more complicated treatment based on aerosol and water content in the UM (Clark et al. 2008). Since many weather service customers, particularly in the aviation sector, require increasingly accurate forecasts of visual range, rather than just “fog” or “no fog,” it is intended that the LANFEX dataset can help improve model parameterizations of visibility, answering questions such as what level of complexity is required in the aerosol model for accurate visibility prediction.

Figure 7 also shows differences in the location of the fog, with the narrow valleys near Pentre and Skyborry containing lower visibility in the UM than Meso-NH, seemingly in better agreement with the observations. We therefore compare and contrast the model behavior in the wide valley of Jaybarns with the narrow valley of Pentre (see Fig. 2 for the cross sections). Figure 8 presents time–height plots of the cloud/fog water content observed by a ceilometer and simulated by the models. At Jaybarns, Meso-NH is clearly superior with the timing of fog onset (around 2300 UTC); however, the fog that develops is quickly too deep. A notable early result of LANFEX, presented in Boutle et al. (2017), is that the overly quick development of fog like this, which is common in numerical models (e.g., also see Maronga and Bosveld 2017), can be the result of poor representation of aerosol–fog interactions. Figure 9e demonstrates that the downwelling longwave radiation (proportional to the fog optical thickness) is considerably too high in Meso-NH during this early fog period, and the increase seen is instant after fog onset (rather than gradual, as expected when a fog layer gradually deepens), which is consistent with the conclusions of Boutle et al. (2017). In this case, the overdevelopment could also be linked to the overestimation of near-surface humidity prior to the onset of fog (Fig. 9c), motivating further study of the dewmeter observations or the overlying stratus cloud (at 600-m altitude; Fig. 8a), which arrives 2 h late in the Meso-NH model. This cloud caused a temporary dissipation of fog (Fig. 6), and it is possible that by arriving late in the model, it allowed the fog too much time to develop and also meant that the fog did not dissipate completely in the model when the cloud was present. The simulation of this overlying stratus is also clearly superior in Meso-NH compared to the UM. This illustrates a point noted previously (Hughes et al. 2015; Boutle et al. 2016) that model simulations are very sensitive to the macroscale parameterization of cloud cover, even at the 100-m scale where most models assume that cloud cover is all or nothing. As fog forecasts are so sensitive to this behavior, future LANFEX work will aim to develop cloud parameterizations for the 100 m–1 km grid scale.

Figure 8 shows that the UM by contrast does not develop fog until 0500 UTC, which is considerably late at Jaybarns. Figure 9a shows that this is likely to be because the near-surface temperature is approximately 6 K too warm by 0000 UTC, in contrast to Meso-NH, which has a very good representation of the temperature evolution prior to the stratus cloud arriving. The dynamics of the valley flows (Fig. 10) appear well represented in both models, although the sudden drop in temperature coincident with the drop in wind speed to <1 m s–1 in the observations and the Meso-NH model is not apparent in the UM. This drop in temperature is likely to be due to the cessation of turbulence, suggesting that either the turbulence parameterization is too active in the UM [see Holtslag et al. (2013) for a recent review of problems with stable boundary layer turbulence in NWP models] or the vertical resolution cannot accurately represent the enhanced near-surface cooling [e.g., similar to the discussion in Vosper et al. (2013)]. The excess temperature in the UM is certainly a key reason for the delayed fog onset, but improvements to this must be coupled with improvements to the fog microphysics to prevent the overdevelopment of fog such as that shown in Meso-NH.

Finally, we consider the narrow valley of Pentre. Here, there is a very clear shift in wind direction, from easterly to westerly at 1800 UTC (Fig. 10d), as the onset of a downslope drainage flow commences. The UM is able to clearly reproduce this shift in wind direction, despite the simulated upslope flow during the daytime being too strong. Meso-NH also reproduces the direction change well, although ∼2 h too late in conjunction with the near-surface temperature drop. A similar model run where the 100-m model was initialized 5 h earlier (at 1300 UTC) gave a similar result, indicating that the initialization time was not responsible for the late forecast of wind change. Since the Meso-NH forecast the onset of drainage flows at the other three valley sites well, the result at Pentre is likely to be a local effect. The result highlights the challenges of forecasting boundary layer flow in regions of heterogeneous terrain.

SUMMARY AND OUTLOOK.

We have described the LANFEX field and modeling campaign to study radiation fog formation and evolution. A principal goal was to use a region of small hills as a natural laboratory, comparing and contrasting subtly different meteorological conditions in different valleys and deducing any effects these differences have on fog formation.

We have found that when fog transitions from shallow stable layers to deep fog, the magnitude of the surrounding orography does not appear to play a role, since we observed the same proportion of transitions to adiabatic fog in two regions of significantly differing orography: the Cardington site is flatter and more homogeneous than Shropshire. However, since shallow orography is present at Cardington, we cannot conclude that orography does not play a role in the vertical growth of fog. Further work is required to examine this, and data from very flat, open sites may prove useful in this respect.

There is also some evidence to suggest that where higher cooling rates were present during the evening (generally seen in the steeper narrower valleys), fog was no more likely to form as a result. In fact, in Shropshire the location with the greatest occurrence of fog (Jaybarns) was the widest and shallowest valley and sometimes experienced a slower cooling rate. Despite this, temperatures there would “catch up” with those in the narrower valleys, usually sometime after midnight, when it often became the coolest site, and persistent fog would often form. In this respect, it may be significant that Jaybarns was the lowest-lying site (Fig. 2) so that on a clear night it can be expected to be where the coldest air eventually accumulates and therefore experiences a higher occurrence of fog.

Observations of turbulence [vertical velocity variance (ww)] have been conducted for LANFEX data. Forming a better understanding of how turbulence affects the fluxes of heat and moisture, as well as the relationship between dew deposition and fog formation, is a central aim of LANFEX. A question arising from the analysis so far is whether a general turbulence threshold exists above which radiation fog will not form. Future presentation of results will address these issues and provide a fuller analysis of LANFEX turbulence observations [a report detailing the relationship between turbulence, humidity, dew deposition, and fog formation is given in Price (2018), which supports the initial result presented in Fig. 6].

A key aim of LANFEX is to improve numerical weather prediction for fog events so that their often negative impact on human activities can, at least, be partly mitigated. We have presented some early results that have identified areas of weather prediction that need to be improved. The modeling work so far has shown that NWP models in the region of 100-m-grid-length range can reasonably reproduce the local-scale variability that leads to the onset and development of fog in a network of small valleys. It has also shown that adequate representation of the vertical resolution is equally important to the horizontal resolution. Clear deficiencies in model parameterizations—aerosol activation, turbulence representation, cloud micro- and macrophysics—are in need of further development, and LANFEX provides a complete set of observations from which to improve them. Boutle et al. (2017) provide an example, using LANFEX data, of how changes to microphysics can improve the simulation of fog evolution from shallow to deep. Future work will more thoroughly investigate the differences between models for IOP 12, using some of our more novel observations, such as dew deposition, to further constrain the fog microphysics. Intercomparisons of other LANFEX IOPs (such as IOP 1) will determine the generality of these results and help to evaluate the ability of high-resolution models to simulate both the local and nonlocal mechanisms of fog development.

To date the LANFEX dataset has proven most useful in studying the initial formation and midlife development of radiation fog, which was the main aim of the project. However, the study of fog dissipation should also be given high priority, and while we note that the LANFEX dataset has potential to study this, fewer data were collected for the fog dissipation phase such that it is anticipated further observation campaigns are required to study this.

ACKNOWLEDGMENTS

Our thanks go to all of the people involved in the LANFEX campaign, including the landowners on whose land we placed equipment, many of whom asked for nothing in return. Dave Bamber, Bernard Claxton, Tony Jones, Nicky Osborn, and Jeff Norwood-Brown are acknowledged for their help with technical and administrative matters and during setup, takedown, and intensive observation periods. D. K. E. Smith was supported by NERC Grant NE/M010325/1. We thank the editor and anonymous reviewers for their detailed comments, which helped to improve this paper.

REFERENCES

REFERENCES
Agarwal
,
M.
,
T.
Maze
, and
R.
Souleyrette
,
2005
:
Impact of weather on urban freeway traffic flow characteristics and facility capacity
.
Centre for Transportation Research and Education Rep.
,
22
pp., www.ctre.iastate.edu/reports/weather_impacts.pdf.
Bartoková
,
I.
,
A.
Bott
,
J.
Bartok
, and
M.
Gera
,
2015
:
Fog prediction for road traffic safety in a coastal desert region: Improvement of nowcasting skills by the machine-learning approach
.
Bound.-Layer Meteor.
,
157
,
501
516
, https://doi.org/10.1007/s10546-015-0069-x.
Bergot
,
T.
,
2013
:
Small-scale structure of radiation fog: A large-eddy simulation study
.
Quart. J. Roy. Meteor. Soc.
,
139
,
1099
1112
, https://doi.org/10.1002/qj.2051.
Boutle
,
I. A.
,
A.
Finnenkoetter
,
A. P.
Lock
, and
H.
Wells
,
2016
:
The London Model: Forecasting fog at 333 m resolution
.
Quart. J. Roy. Meteor. Soc.
,
142
,
360
371
, https://doi.org/10.1002/qj.2656.
Boutle
,
I. A.
,
J.
Price
,
I.
Kudzotsa
,
H.
Kokkola
, and
S.
Romakkaniemi
,
2017
:
Aerosol–fog interaction and the transition to well-mixed radiation fog
.
Atmos. Chem. Phys.
,
18
,
7827
7840
, https://doi.org/10.5194/acp-18-7827-2018.
Clark
,
P. A.
,
S. A.
Harcourt
,
B.
Macpherson
,
C. T.
Mathison
,
S.
Cusack
, and
M.
Naylor
,
2008
:
Prediction of visibility and aerosol within the operational Met Office Unified Model. I: Model formulation and variational assimilation
.
Quart. J. Roy. Meteor. Soc.
,
134
,
1801
1816
, https://doi.org/10.1002/qj.318.
Cuxart
,
J.
, and
M. A.
Jiménez
,
2012
:
Deep radiation fog in a wide closed valley: Study by numerical modeling and remote sensing
.
Pure Appl. Geophys.
,
169
,
911
926
, https://doi.org/10.1007/s00024-011-0365-4.
Duynkerke
,
P. G.
,
1999
:
Turbulence, radiation, and fog in Dutch stable boundary layers
.
Bound.-Layer Meteor
.,
90
,
447
477
, https://doi.org/10.1023/A:1026441904734.
Fitzjarrald
,
D. R.
, and
G. G.
Lala
,
1989
:
Hudson valley fog environments
.
J. Appl. Meteor.
,
28
,
1303
1328
, https://doi.org/10.1175/1520-0450(1989)028<1303:HVFE>2.0.CO;2.
Fuzzi
,
S.
, and Coauthors
,
1992
:
The Po valley fog experiment 1989: An overview
.
Tellus
,
44B
,
448
468
, https://doi.org/10.3402/tellusb.v44i5.15561.
Guedalia
,
D.
, and
T.
Bergot
,
1994
:
Numerical forecasting of radiation fog. Part II: A comparison of model simulation with several observed fog events
.
Mon. Wea. Rev.
,
122
,
1231
1246
, https://doi.org/10.1175/1520-0493(1994)122<1231:NFORFP>2.0.CO;2.
Gultepe
,
I.
, and Coauthors
,
2007
:
Fog research: A review of past achievements and future perspectives
.
Pure Appl. Geophys.
,
164
,
1121
1159
, https://doi.org/10.1007/s00024-007-0211-x.
Gultepe
,
I.
, and Coauthors
,
2009
:
The Fog Remote Sensing and Modeling (FRAM) field project
.
Bull. Amer. Meteor. Soc.
,
90
,
341
359
, https://doi.org/10.1175/2008BAMS2354.1.
Gultepe
,
I.
, and Coauthors
,
2016
:
An overview of the MATERHORN fog project: Observations and predictability
.
Pure Appl. Geophys.
,
173
,
2983
3010
, https://doi.org/10.1007/s00024-016-1374-0.
Haeffelin
,
M.
,
Q.
Laffineur
,
J.-A.
Bravo-Aranda
,
M.-A.
Drouin
,
J.-A.
Casquero-Vera
,
J.-C.
Dupont
, and
H.
De Backer
,
2016
:
Radiation fog formation alerts using attenuated backscatter power from automatic lidars and ceilometers
.
Atmos. Meas. Tech.
,
9
,
5347
5365
, https://doi.org/10.5194/amt-9-5347-2016.
Herman
,
G. R.
, and
R. S.
Schumacher
,
2016
:
Using reforecasts to improve forecasting of fog and visibility for aviation
.
Wea. Forecasting
,
31
,
467
482
, https://doi.org/10.1175/WAF-D-15-0108.1.
Holtslag
,
A. A. M.
, and Coauthors
,
2013
:
Stable atmospheric boundary layers and diurnal cycles: Challenges for weather and climate models
.
Bull. Amer. Meteor. Soc.
,
94
,
1691
1706
, https://doi.org/10.1175/BAMS-D-11-00187.1.
Hughes
,
J. K.
,
A. N.
Ross
,
S. B.
Vosper
,
A. P.
Lock
, and
B. C.
Jemmet-Smith
,
2015
:
Assessment of valley cold pools and clouds in a very high-resolution NWP model
.
Geosci. Model Dev.
,
8
,
3105
3117
, https://doi.org/10.5194/gmd-8-3105-2015.
Jayakumar
,
A.
,
E. N.
Rajagopal
,
I. A.
Boutle
,
J. P.
George
,
S.
Mohandas
,
S.
Webster
, and
S.
Aditi
,
2018
:
An operational fog prediction system for Delhi using the 330 m Unified Model
.
Atmos. Sci. Lett.
,
19
,
e796
, https://doi.org/10.1002/asl.796.
Kunkel
,
B. A.
,
1984
:
Parameterization of droplet terminal velocity and extinction coefficient in fog models
.
J. Climate Appl. Meteor.
,
23
,
34
41
, https://doi.org/10.1175/1520-0450(1984)023<0034:PODTVA>2.0.CO;2.
Lac
,
C.
, and Coauthors
,
2018
:
Overview of the Meso-NH model version 5.4 and its applications. Geosci
.
Model Dev
.,
11
,
1929
1969
https://doi.org/10.5194/gmd-2017-297.
Maronga
,
B.
, and
F.
Bosveld
,
2017
:
Key parameters for the life cycle of nocturnal radiation fog: A comprehensive large-eddy simulation study
.
Quart. J. Roy. Meteor. Soc.
,
143
,
2463
2480
, https://doi.org/10.1002/qj.3100.
Mazoyer
,
M.
,
C.
Lac
,
O.
Thouron
,
T.
Bergot
,
V.
Masson
, and
L.
Musson-Genon
,
2017
:
Large-eddy simulation of radiation fog: Impact of dynamics on microphysics
.
Atmos. Chem. Phys.
,
17
,
13 017
13 035
, https://doi.org/10.5194/acp-17-13017-2017.
McCabe
,
A.
,
R.
Swinbank
,
W.
Tennant
, and
A.
Lock
,
2016
:
Representing model uncertainty in the Met Office convection-permitting ensemble prediction system and its impact on fog forecasting
.
Quart. J. Roy. Meteor. Soc.
,
142
,
2897
2910
, https://doi.org/10.1002/qj.2876.
Menut
,
L.
,
S.
Mailler
,
J. C.
Dupont
,
M.
Haeffelin
, and
T.
Elias
,
2014
:
Predictability of the meteorological conditions favourable to radiative fog formation during the 2011 ParisFog campaign
.
Bound.-Layer Meteor
.,
150
,
277
297
, https://doi.org/10.1007/s10546-013-9875-1.
Nakanishi
,
M.
,
2000
:
Large-eddy simulation of radiation fog
.
Bound.-Layer Meteor.
,
94
,
461
493
, https://doi.org/10.1023/A:1002490423389.
Price
,
J. D.
,
2011
:
Radiation fog. Part I: Observations of stability and drop size distributions
.
Bound.-Layer Meteor.
,
139
,
167
191
, https://doi.org/10.1007/s10546-010-9580-2.
Price
,
J. D.
,
2018
: On the formation and development of radiation fog: An observational study. Cardington Tech. Note 93, 23 pp.
Price
,
J. D.
, and
R.
Clark
,
2014
:
On the measurement of dewfall and fog-droplet deposition
.
Bound.-Layer Meteor.
,
152
,
367
393
, https://doi.org/10.1007/s10546-014-9930-6.
Price
,
J. D.
, and Coauthors
,
2011
:
COLPEX: Field and numerical studies over a region of small hills
.
Bull. Amer. Meteor. Soc.
,
92
,
1636
1650
, https://doi.org/10.1175/2011BAMS3032.1.
Price
,
J. D.
,
A.
Porson
, and
A.
Lock
,
2015
:
An observational case study of persistent fog and comparison with an ensemble forecast model
.
Bound.-Layer Meteor
.,
155
,
301
327
, https://doi.org/10.1007/s10546-014-9995-2.
Roach
,
W. T.
,
R.
Brown
,
S. J.
Caughey
,
J. A.
Garland
, and
C. J.
Readings
,
1976
:
The physics of radiation fog: I—A field study
.
Quart. J. Roy. Meteor. Soc.
,
102
,
313
333
, https://doi.org/10.1002/qj.49710243204.
Robinson
,
P. J.
,
1989
:
The influence of weather on flight operations at the Atlanta Hartsfield International Airport
.
Wea. Forecasting
,
4
,
461
468
, https://doi.org/10.1175/1520-0434(1989)004<0461:TIOWOF>2.0.CO;2.
Rodhe
,
B.
,
1962
:
The effect of turbulence on fog formation
.
Tellus
,
14
,
49
86
, https://doi.org/10.3402/tellusa.v14i1.9528.
Román-Cascón
,
C.
,
G. J.
Steeneveld
,
C.
Yagüe
,
M.
Sastre
,
J. A.
Arrillaga
, and
G.
Maqueda
,
2016
:
Forecasting radiation fog at climatologically contrasting sites: Evaluation of statistical methods and WRF
.
Quart. J. Roy. Meteor. Soc.
,
142
,
1048
1063
, https://doi.org/10.1002/qj.2708.
Seity
,
Y.
,
P.
Brousseau
,
S.
Malardel
,
G.
Hello
,
P.
Bénard
,
F.
Bouttier
,
C.
Lac
, and
V.
Masson
,
2011
:
The AROME-France convective-scale operational model
.
Mon. Wea. Rev.
,
139
,
976
991
, https://doi.org/10.1175/2010MWR3425.1.
Steeneveld
,
G. J.
,
R. J.
Ronda
, and
A. A. M.
Holtslag
,
2015
:
The challenge of forecasting the onset and development of radiation fog using mesoscale atmospheric models
.
Bound.-Layer Meteor.
,
154
,
265
289
, https://doi.org/10.1007/s10546-014-9973-8.
Taylor
,
G. I.
,
1917
:
The formation of mist and fog
.
Quart. J. Roy. Meteor. Soc.
,
43
,
241
268
, https://doi.org/10.1002/qj.49704318302.
Tudor
,
M.
,
2010
:
Impact of horizontal diffusion, radiation and cloudiness parameterization schemes on fog forecasting in valleys
.
Meteor. Atmos. Phys.
,
108
,
57
70
, https://doi.org/10.1007/s00703-010-0084-x.
van der Velde
,
I. R.
,
G. J.
Steeneveld
,
B. G. J. W.
Schreur
, and
A. A. M.
Holtslag
,
2010
:
Modeling and forecasting the onset and duration of severe radiation fog under frost conditions
.
Mon. Wea. Rev.
,
138
,
4237
4253
, https://doi.org/10.1175/2010MWR3427.1.
Vautard
,
R.
,
P.
Yiou
, and
G. J.
van Oldenborgh
,
2009
:
Decline of fog, mist and haze in Europe over the past 30 years
.
Nat. Geosci.
,
2
,
115
119
, https://doi.org/10.1038/ngeo414.
von Glasow
,
R.
, and
A.
Bott
,
1999
:
Interaction of radiation fog with tall vegetation
.
Atmos. Environ.
,
33
,
1333
1346
, https://doi.org/10.1016/S1352-2310(98)00372-0.
Vosper
,
S. B.
, and
A. R.
Brown
,
2008
:
Numerical simulation of sheltering in valleys: The formation of nighttime cold-air pools
.
Bound.-Layer Meteor.
,
127
,
429
448
, https://doi.org/10.1007/s10546-008-9272-3.
Vosper
,
S. B.
,
E.
Carter
,
H.
Lean
,
A.
Lock
,
P.
Clark
, and
S.
Webster
,
2013
:
High resolution modelling of valley cold-pools
.
Atmos. Sci. Lett.
,
14
,
193
199
, https://doi.org/10.1002/asl2.439.
Ye
,
X.
,
B.
Wu
, and
H.
Zhang
,
2015
:
The turbulent structure and transport in fog layers observed over the Tianjin area
.
Atmos. Res.
,
153
,
217
234
, https://doi.org/10.1016/j.atmosres.2014.08.003.
Zhou
,
B.
, and
B. S.
Ferrier
,
2008
:
Asymptotic analysis of equilibrium in radiation fog
.
J. Appl. Meteor. Climatol.
,
47
,
1704
1722
, https://doi.org/10.1175/2007JAMC1685.1.

Footnotes

A supplement to this article is available online (10.1175/BAMS-D-16-0299.2).

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Supplemental Material