a. Data sources

Ten Integrated Surface Flux Facility (ISFF) surface meteorological stations were positioned in the foothills around the Santa Catalina Mountains (Fig. 2). These stations measured temperature, humidity, and pressure at 2 m AGL and wind at 10 m AGL, between 22 June and 31 August 2006. Four of the stations were on sufficiently level and uniformly vegetated terrain to estimate surface sensible and latent heat fluxes, so they were equipped with high-frequency temperature and humidity sensors and 3D sonic anemometers (Table 1). The vegetation around the ISFF stations was typical for the upper Sonoran Desert: broadly separated paloverde and ocotillo trees dominated, interspersed with saguaro cacti. Thus, the land surface was largely bare soil, for which significant rainfall can dramatically alter the surface energy balance. The ISFF pressure sensor was a Vaisala PTB220B, a highly stable instrument with an accuracy of 0.25 hPa at the temperature range encountered during CuPIDO. (More information about the sensors and the siting of the ISFF stations can be online found at http://www.eol.ucar.edu/rtf/projects/CuPIDO/isff/.)

We also use meteorological data from two mountain stations: one is a continuously operating 30-m flux tower located on Mt. Bigelow and the other is an astronomical observatory on the Santa Catalina Mountains’ highest point, Mt. Lemmon. [The tower at Mt. Bigelow is operated by the Sustainability of semi-Arid Hydrology and Riparian Areas (SAHRA) at the University of Arizona; its instruments are described online at http://www.sahra.arizona.edu/research/TA1/towers/.] The meteorological data acquisition system at Mt. Lemmon was struck by lightning on 19 August 2006, resulting in a lack of Mt. Lemmon data between 19 and 31 August. All data were collected at 5-min intervals or better, except for the Mt. Lemmon data, which were hourly. Therefore, in any comparison that includes all 12 stations, all data are reduced to a 60-min resolution by centered averaging.

The altitude of the 12 stations ranges from 840 m on the southwestern side to 2779 m (the Mt. Lemmon astronomical observatory), a difference of 1939 m. We also use data from the National Lightning Detection Network (NLDN) and radiosonde data, both from the National Weather Service (NWS) Tucson office (TUS) located on the University of Arizona campus, and from the Mobile Global Positioning System (GPS) Advanced Upper-Air Sounding System (MGAUS), deployed as part of CuPIDO.

Surface measurements at the 10 ISFF stations, at the Mt. Bigelow flux tower, and at the Mt. Lemmon observatory were available for an overlapping period of 58 days in the summer of 2006 (22 June–18 August). Thus, the present analysis is focused on that period. This period can be subdivided in two parts: the first period (22 June–25 July) witnessed several orographic afternoon thunderstorms but the soil remained rather dry and the surface latent heat fluxes for the four ISFF flux stations remained small. We refer to it as the premonsoon dry period, even though the daily mean dewpoint exceeded the NWS’s threshold for monsoon conditions in Tucson (12.2°C) on several days (Fig. 3b). The average rainfall at the 10 ISFF stations between 26 July and 5 August 2006 amounted to 274 mm (Fig. 3b), much of it from nocturnal organized convection (Damiani et al. 2008). This exceptionally heavy rainfall resulted in near-saturated soils, local flooding, and a latent heat flux that far exceeded the sensible heat flux at any time of the day. We refer to the period of 26 July–18 August as the monsoon wet period.

High temperatures and a deep CBL prevailed during the premonsoon dry period (Fig. 3a). To establish the CBL depth, we use the operational 0000 UTC soundings, released at TUS about 3.8 h after local solar noon (LSN). The CBL depth is defined as the midpoint between two consecutive levels in a sounding where the potential temperature increases by at least 1 K, and that increase is sustained. The midpoint (rather than the base) is chosen because of the coarse vertical resolution of the soundings. Sometimes this method yields an unrealistically shallow CBL, especially on some days with thunderstorms, but Fig. 3a demonstrates that generally the CBL exceeded the height of the mountain during the premonsoon dry period, and that the CBL top above TUS was often near the mountaintop level during the wet period.

Winds were generally weak (<5 ms⁻¹) below the elevation of Mt. Lemmon, especially during the wet period (Fig. 3c). On some days stronger winds (>10 ms⁻¹) penetrated down to 1–2 km above Mt. Lemmon. On such days the interaction between large-scale mean flow and the mountain may have impacted the pressure, temperature, and flow distribution around the mountain. Yet on those days (except on 27–28 July) the CBL depth exceeded the mountain top height, and the observed circular asymmetry of surface flow at the ISFF stations was weak, so the interaction with the large-scale flow is ignored.

b. Pressure perturbations and the horizontal pressure gradient force

A pressure perturbation is a departure from a “mean” or “basic-state” value. The mean pressure is usually defined as the value that is hydrostatically consistent with the mean density profile. Pressure perturbations near mountains can be both hydrostatic and nonhydrostatic.

According to hydrostatic balance, a higher temperature over some depth in the atmosphere implies a lower pressure below this layer. During the summer the mature CBL is usually deeper than the altitude of Mt. Lemmon (Fig. 3a), thus, the CBL top is sketched above the mountain top in Fig. 1. To estimate the impact of a temperature anomaly within the CBL over the mountain (as shown in Fig. 1) on the air pressure below, we assume a constant virtual potential temperature (θ_υ) profile. Then the anomalous pressure (p′_ref) at the reference level (z_ref) due to a θ_υ anomaly (θ′_υ) is linearly proportional to the θ_υ anomaly and (to a good approximation) linearly proportional to its depth δ:

View Expanded

This relationship is derived from hydrostatic balance and the ideal gas law, assuming that θ_υ′ is height independent over the layer depth δ, that the perturbations are small compared to the mean values, and that the pressure on top of the CBL is uniform (p = p_top at z_top = z_ref + δ). In (1), R_d is the ideal gas constant for dry air, c_p is the specific heat under constant pressure, g is the gravitational acceleration, K is a constant (K = (R_d/c_p) = 0.286), and p_o ≡ 1000 mb.

The average height difference between the two mountain stations and the 10 ISFF stations is 1.6 km. Thus, setting the reference level at the average altitude of the ISFF stations in the foothills and δ = 1.6 km, then a θ_υ excess of 2 K over the mountain yields a pressure deficit of 1.0 mb, according to (1). Clearly this lower pressure would occur under the mountain bedrock (Fig. 1), but the horizontal pressure gradient force is nevertheless real.

Station pressure values are normally reduced to a common height above mean sea level (MSL) in order to evaluate the horizontal pressure gradient forcing (e.g., Reiter and Tang 1984; Tucker 1999). This is not feasible in the present study because of the large variations in the altitude of the stations on and around the Santa Catalina Mountains. The hydrostatic reduction to a common height would be too sensitive to the assumed temperature profile. Rather, we define a pressure perturbation from a temporal and spatial mean, and relate the horizontal gradient of this perturbation to that of the actual pressure. This only yields a pressure gradient (not a pressure), but that is sufficient for our purpose (i.e., to infer the forcing of anabatic wind). The only assumptions are hydrostatic balance, and a relatively small horizontal scale, specified below.

First, station pressure perturbations p′ are defined as departures from their 24-h station mean value, p_t. We assume that over the scale of the mountain, at any level, the horizontal variation of p_t and thus the mean geostrophic wind are insignificant; that is, ∂p_t/∂r ≈ 0, where r is the radial direction from the mountain, using cylindrical coordinates. (This applies to the azimuthal direction as well, but we assume circular symmetry, for simplicity.) But clearly p_t varies significantly with height. Next we remove p_s, the mean of p_t for all stations at any given time. This removes (semi) diurnal pressure variations, at least if the spatial structure of these variations is large compared to the network of stations used. Note that p_s is a function of time only, not any spatial dimension. Thus,

View Expanded

Wind is forced by a horizontal pressure gradient, thus the question is whether we can treat the observed difference in p″ between stations (Δp″) as a measure of horizontal pressure gradient ∂p/∂r. Treating the finite difference as a differential, we obtain the following from (2):

View Expanded

Assuming that both p and p_t are in hydrostatic balance, (3) can be expressed as

View Expanded

Here ρ is the air density, and ρ_t is its 24-h mean value. The term Δz/Δr is the average slope S of the terrain. Using the ideal gas law, and assuming that the perturbation ρ − ρ_t is small enough that differential calculus applies, (4) becomes

View Expanded

or

View Expanded

Here T_υ − T_υ,t is the virtual temperature departure from the 24-h mean, for any station. Fujita et al. (1962) argued, without giving details, that observed pressure perturbation gradients between stations (Δp″/Δx) can be treated as horizontal pressure gradients [the term on the left in (5)], in other words, that the last two terms in (5) can be ignored. These two terms depend on the magnitude of the diurnal cycles of pressure and temperature, respectively. To address whether they can be ignored, we scale (5) for the differences between a mountain station and a foothill station in the Santa Catalina Mountains, and at times that the diurnal cycle terms in (5) reach their maximum value. It will be shown in section 3 that Δp″ peaks at 1.0 mb, over a distance Δr of 15 km on average, and that the maximum pressure departures from the 24 h mean [largely due to (semi) diurnal tides] at stations around the mountain, p − p_t, average at 1.4 mb. The average slope between the mountain top and the ISFF stations is S = 0.10. Thus, the first term on the right in (5) scales as 7 × 10⁻³ Pa m⁻¹, and the second term scales as 2 × 10⁻³ Pa m⁻¹. The maximum values for (T_υ − T_υ,t) average at 3.8 K, thus the last term on the right in (5) scales as 13 × 10⁻³ Pa m⁻¹. This scale analysis indicates that the observed spatial difference in p″ is not just due to a horizontal pressure gradient, but also due to the diurnal temperature cycle. Thus, we use (5) to estimate the diurnal variation of the (hydrostatic) horizontal pressure gradient ∂p/∂r between the mountain and the 10 foothill stations. Because the temperature sensor at the Mt. Lemmon astronomical observatory was biased by exposure to sunshine during part of the day (see section 3), we use the Mt. Bigelow tower as the reference mountain site. To evaluate the virtual temperature and pressure departures from their 24-h mean values in (5), we use average values of the two sources for which the pressure difference is computed (i.e., Mt. Bigelow and the select ISFF station).

c. Thermally forced orographic circulation

Mahrt (1982) developed the momentum equations for katabatic flow in terrain-following coordinates. The equations can be applied to upslope flow. Thus, anabatic wind can be driven by two terms: a buoyancy term and a hydrostatic thermal wind term. The buoyancy term is due to local excess in θ_υ and vanishes when the slope disappears. The thermal wind term drives a sea breeze over flat terrain, for instance. It yields upslope flow when the CBL contains a higher θ_υ over the mountain. Unfortunately, we cannot estimate the thermal wind term because we do not know the horizontal variations of the CBL depth. Because the CBL was typically weakly capped in CuPIDO, we cannot assume that its top is flat. There is some evidence from simultaneous MGAUS soundings that the CBL top domed over the mountain, as sketched in Fig. 1. This is attributed to local surface heating over the mountain.

A sufficient way to quantify the forcing of anabatic flow is to consider the horizontal vorticity equation (e.g., Miao and Geerts 2007):

View Expanded

Here D/Dt is the total derivative, η = (∂u/∂z) − (1/r)(∂rw/∂r) is the horizontal vorticity around the mountain, and (u, w) the velocity vector in a radial cross section. A toroidal circulation, or ring vortex around the mountain, is driven by a radial gradient in θ_υ. According to Eq. (1), this gradient is proportional to the hydrostatic horizontal pressure gradient force, discussed in section 2b.

The toroidal circulation includes low-level, convergent, anabatic flow, and divergent flow at some level aloft. We cannot estimate the divergent flow aloft, but the location of the ISFF stations allows an estimation of the low-level convergent component (called the mean anabatic wind υ_n) as follows:

View Expanded

where C = ∮ ds is the circumference of a polygon obtained by connecting the 10 midpoints between the 10 adjacent ISFF stations (Fig. 2), ds is the incremental distance along this irregular decagon, and υ_n is the horizontal wind component normal to vector ds. The reason for the use of midpoints is that it positions the stations centrally near the decagon sides, and thus the station winds yield the best available estimate for υ_n along corresponding decagon sides. By definition, υ_n > 0 for winds blowing into the decagon. The mountain-scale convergence −∇ · v is linearly proportional to υ_n, according to the divergence theorem (e.g., Holton 2004) and (7):

View Expanded

Here A is the area contained within the decagon, A = 576 km².

a. Surface pressure and temperature variations around the mountain

The diurnal pattern of the “single” pressure perturbation (p − p_t) is dominated by the diurnal and semidiurnal tides, each with an amplitude of about 1.0 mb (Fig. 4a). These oscillations appear very similar for each site, but both frequencies, especially the diurnal component, have a smaller amplitude at the two mountain stations. The differences in diurnal variation of p″ (p − p_t − p_s; Fig. 4b) between stations should almost entirely be due to boundary layer processes, since the upper-atmospheric (semi) diurnal tidal variations have a large-scale structure that is essentially uniform for our cluster of 12 stations (e.g., Hagan et al. 2002).

The difference in p″ between the mountain and the foothill stations is roughly periodic, with an amplitude of nearly 1.0 mb: the pressure is anomalously low over the mountain near midnight and peaks ∼2 h after sunrise. It is anomalously high at the mountain sites throughout the afternoon, peaking 4–5 h after LSN. A closer examination of Fig. 4b shows that pressure perturbations are largely controlled by station altitude (color coded and shown on the right side in Fig. 2). The highest of all ISFF stations (i.e., station 10), is closest to the Mt. Lemmon curve, while the perturbation pressure trend of the low-elevation west-side stations (i.e., stations 1, 2, and 3) is most dissimilar to Mt. Lemmon’s.

The dependence of Δp″ on station elevation differences appears to be the result of differences in the diurnal virtual temperature variation integrated over the depth of the local boundary layer. We do not have temperature profile data, but within the well-mixed CBL, surface θ_υ is a sufficient surrogate. Mt. Bigelow has the smallest diurnal temperature variation (Fig. 4c), mainly because the nighttime cooling is less, thus, it is anomalously warm in the second half of the night and anomalously cool in the afternoon. The temperature variation at the Mt. Lemmon astronomical observatory is not as reliable, the probe apparently was exposed to the sun, thus the daytime heating is exaggerated (Fig. 4c), and therefore also the negative T ′ at night. Station 2 has the largest diurnal temperature range, because of its low elevation, and because it is located in a local valley (Fig. 2): cold air drains, thus this station records the strongest cool anomaly near sunrise. Local terrain concavity matters (Geerts 2003): a station such as station 9 on a slight local ridge on the piedmont east of the Santa Catalina Mountains is relatively warm near sunrise (Fig. 4d).

As discussed in section 2b, the observed differences in p″ (Fig. 4b) cannot be interpreted as horizontal pressure differences. The hydrostatic horizontal pressure difference (∂p/∂r)Δr between Mt. Bigelow and the surrounding stations is estimated using (5), and its diurnal variation is shown in Fig. 5a. A clear diurnal cycle is present, with about the same magnitude as that for differences in p″, but the phase has shifted: the lowest pressure occurs over the mountain (relative to the surrounding stations) close to LSN, about 5 h later than the maximum p″ deficit over the mountain (Fig. 4b). The period when the horizontal pressure difference is negative (lower pressure at Mt. Bigelow) is referred to as the anabatic forcing period. This period starts 2 h after sunrise, peaks close to LSN, and ends shortly before sunset (Fig. 5a). Its counterpart, the “katabatic forcing period,” peaks between midnight and sunrise; the katabatic forcing period has a flatter peak than its anabatic counterpart and lasts some 3 h longer than the nighttime.

The surface heat fluxes in Fig. 5c are an average for the four ISFF stations with flux capability, and for the 58-day period. None of these four stations is located on the mountain. Time series of fluxes derived from the Mt. Bigelow flux tower contained some unrealistically high values even for 30-min averages and thus were not included. It is assumed that the average surface flux from the four foothill stations is representative of the overall mountain area encompassed by the decagon in Fig. 2.

We now examine three details of the diurnal variation of horizontal pressure differences (Fig. 5a). First, the anabatic forcing period seems nearly coincident with the positive surface sensible heat (SH) flux cycle (Fig. 5c), and thus the cycle of net incoming solar radiation at the surface. This seems rather early; in fact its peak occurs a few hours before LSN for stations on the east side of the mountain. Thus, the anabatic forcing period for a mountain the width of the Santa Catalina Mountains appears to peak earlier than the sea-breeze forcing, which is proportional to the temperature difference between the marine BL and the CBL over the adjacent landmass, and that difference peaks a few hours after LSN (Abbs and Physick 1992). This observation will be revisited.

Second, the slight depression in the composite anabatic forcing period at ∼1500 local time (LT;2200–2300 UTC) may be related to thunderstorm activity. Lightning data from the NLDN suggest a strong diurnal modulation of monsoonal thunderstorms in the Tucson, Arizona, area: they almost all occur between 2200 and 0300 UTC (Watson et al. 1994). Orographic thunderstorms may generate a cold pool and pressure rise over the mountain. We revisit this later as well.

And third, we question the validity of one of the findings of Fujita et al. (1962). They studied summertime pressure variations around the San Francisco Peaks, a mountain of similar size near Flagstaff, Arizona. The abstract of Fujita et al. (1962) states that “. . . a very small low-pressure area formed over the heated side of the mountain slope . . .” They attribute the lower pressure on the southeastern slopes in the morning to the higher net radiation and thus a higher surface heat flux at that time. Our data do not confirm this pressure response. The stations on the east side of the Santa Catalina Mountains tend to have the highest pressure relative to Mt. Bigelow 2–3 h before LSN (mostly blue stations in Fig. 5a), while those on the western flanks of the mountain (mostly colored red in Fig. 5a) experience the strongest anabatic wind forcing just after LSN. In other words, the six eastside stations have a rather high pressure in the morning [−0.2 mb compared to the four westside stations, at 1000 LT, and a lower pressure just after LSN (−0.4 mb at 1300 LT)]. Further evidence comes from the contrast between two stations: 5 (east) and 4 (northwest). They are located at roughly the same elevation on opposite foothills of the Santa Catalina Mountains. In the morning the horizontal pressure difference is negligible (Fig. 6a), while in the evening the eastern station (i.e., 5) actually has a slightly lower pressure, even though, as expected, its temperature perturbation is lower than at the NW station (i.e., 4) at that time (Fig. 6b). The key factor appears to be the small altitude difference between stations 4 and 5 (Fig. 2). Since the prevailing winds were southeasterly during the 58-day period, and the stratification is strongest around dawn (Fig. 7), it is possible that the anomalously high pressure at station 5 (and the other eastside stations) around dawn, compared to that at station 4, is due to some upstream blocking and flow splitting. In short, the horizontal pressure variations are controlled not by slope orientation but mainly by station altitude.

b. Solenoidal forcing

Three different data sources are used to examine the solenoidal forcing [the rhs (6)] over the Santa Catalina Mountains. First, we explore the diurnal variation of θ_υ based on station data (Fig. 7). Only 11 stations are shown in Fig. 7; Mt. Lemmon was excluded since its record lacked humidity measurements and its temperature data are of poor quality (section 3a). The diurnal θ_υ variation is plotted against altitude; when the CBL is well developed, the x axis in Fig. 7 can be considered to be distance from a (bell shaped) mountain; the lower stations are farther to the right. The nocturnal development of a low-level cold pool is obvious. Little nocturnal cooling occurs at Mt. Bigelow, which was in the residual mixed layer on most days. Of particular interest is the increase of θ_υ toward the mountain (toward the left in Fig. 7a) between about 1200–1700 LT, when the CBL is well developed. This provides clear evidence of solenoidal forcing. The mean θ_υ difference between Bigelow and the 10 foothill stations during the afternoon (1200–1700 LT) is 1.5 K. Clearly the daytime CBL did not reach above the elevation of Mt. Bigelow on some days in the 58-day period examined here, according to the 0000 UTC TUS soundings (Fig. 3a), thus the excess θ_υ at Bigelow could be attributable to vertical stratification. The θ_υ anomaly (i.e., the departure from the 11-station mean) for just those days with a deep CBL is shown in Fig. 7b. Mt. Bigelow still is warmer in the afternoon: the mean θ_υ difference between Bigelow and the 10 foothill stations between 1200 and 1700 LT is 1.2 K for the deep CBL days.

A second piece of evidence for the solenoidal forcing comes from an analysis of the T_υ difference between the Mt. Bigelow tower and the corresponding (linearly interpolated) pressure level in the 0000 UTC TUS sounding. The TUS radiosondes are launched from a site located 27 km to the SSE of Mt. Bigelow. The National Weather Service soundings typically are released 55 min before the nominal time, and the ascent rate is 3–4 m s⁻¹, thus we used the 2315 UTC data from Mt. Bigelow in this comparison. At 2315 UTC (3 h and 45 min after LSN) the CBL normally is well developed, and orographic cumulus convection is likely. A total of 55 days of the 58-day period had sounding data at the level of Mt. Bigelow, and for those days T_υ was on average 1.40 ± 1.53 K warmer at Mt. Bigelow than at the radiosondes. In comparison, the air was 0.68 ± 1.03 K cooler over Mt. Bigelow than aboard the radiosonde at 1115 UTC (∼1 h before sunrise) for 56 of the 58 days with 1200 UTC sounding data. The afternoon sounding–Mt. Bigelow T_υ difference corresponds well with the station data θ_υ difference mentioned above.

A third piece of evidence for the solenoidal forcing comes from the one day in the CuPIDO campaign that a series of MGAUS radiosonde pairs was released simultaneously from an upwind corner of the Santa Catalina Mountains and from its peak. On this day, 17 August 2006, the mean wind below 5 km MSL was just 1.9 m s⁻¹ from 168°, thus the Windy Point release site was generally upwind of Mt Lemmon (Fig. 2). The soundings were released hourly from 4 h before LSN until LSN (Fig. 8). The average CBL top was below the elevation of Mt. Lemmon for the Windy Point soundings, but a shallow CBL did develop above Mt. Lemmon, suggesting a doming CBL top as sketched in Fig. 1. The air was unusually moist for Tucson on this day, with a cumulus cloud base just ∼250 m above Mt. Lemmon. All Mt. Lemmon soundings ascended through cumuli (explaining the large θ_υ variability between soundings), while none of the Windy point soundings penetrated cumuli. Below the cloud base, the Windy Point θ_υ was lower than that at Mt. Lemmon in some sounding pairs, and higher in other sounding pairs (Fig. 8a). On average, the low-level θ_υ difference was of the expected sign, but insubstantial (<0.4 K, Fig. 8b). Above the Cu cloud base, larger θ_υ excesses over Mt. Lemmon were present at some levels in the composite soundings, reflecting net cumulus buoyancy. The lack of low-level θ_υ difference even in the last sounding pair (when the CBL top reaches above Mt. Lemmon) suggest that Windy Point is essentially within the mountain warm core.

These three pieces of evidence confirm the presence of solenoidal forcing with a magnitude of 1–2 K over the width of the mountain. This corresponds with a pressure deficit of 0.5–1.0 mb over the mountain, according to (1). This magnitude chimes with the observed peak horizontal pressure deficit over Mt. Bigelow (Fig. 5a). The pressure difference of 0.5– 1.0 mb over a distance of ∼15 km (the characteristic width of the Santa Catalina Mountains) is comparable to that of solenoidal circulations of matching scales over flat terrain. Florida sea breezes, for instance, are marked by a surface pressure gradient of 0.2–0.5 mb (10 km)⁻¹ (Atkins and Wakimoto 1997; Kingsmill and Crook 2003). For gust fronts this figure ranges between 1–2 mb (10 km)⁻¹ (Mueller and Carbone 1987; Atkins and Wakimoto 1997; Kingsmill and Crook 2003).

c. Anabatic wind

Daytime upslope winds over a heated mountain are highly variable, due to turbulence driven by thermals in the CBL. Also, even the slightest mean advective wind in the mountain environment will yield wind toward the mountain on the upwind side of the mountain and vice versa on the opposite side. Thus upslope flow estimation requires data from around the mountain and some temporal averaging. The 10 ISFF stations were generally well positioned around the mountain (Fig. 2) to compute the mean anabatic wind υ_n and mountain-scale convergence, as defined in Eqs. (7) and (8).

The diurnal variation of the mean anabatic wind in the foothills of the Santa Catalina Mountains is shown in Fig. 5b. The average surface (10 m AGL) flow is katabatic (negative) all night long, starting 1.5 h before sunset and ending 2 h after sunrise, averaging 0.3 m s⁻¹ during this period. Anabatic flow picks up swiftly during the morning, peaks 2 h before LSN, and remains rather steady until ∼1300 LT. At its peak the mean anabatic wind is 0.5 m s⁻¹, which corresponds with a mountain-scale convergence of 0.9 10⁻⁴ s⁻¹.

It is interesting to note that most mass convergence occurs before LSN, while solar radiation and surface heat fluxes peak at LSN (Fig. 5c) and the surface and BL temperatures typically peak a few hours after LSN. In fact the maximum θ_υ difference between Mt. Bigelow and the foothills stations occurs at 1430 LT (Figs. 4c and 7). Thus, one would expect that the solenoidal forcing (essentially the horizontal temperature differences in the CBL) for a toroidal heat island circulation around the mountain and the resulting surface anabatic flow and mountain-scale convergence also peak a few hours after LSN. The early development of anabatic wind (Fig. 5b) is roughly consistent with its pressure forcing (Fig. 5a). Apparently mass convergence in the boundary layer over an isolated mountain is not entirely driven by local surface heating.

The anomalously early peaking of the anabatic wind (and its pressure forcing) may be related to cumulus development over the mountain. One can argue that orographic cumulus development causes net deep-column heating and thus low-level hydrostatic pressure decrease and enhanced convergence. Yet three CuPIDO case studies presented in Demko et al. (2009) do not reveal this. By comparing orographic cumulus growth rate with mountain-scale convergence, they show that the only measurable impact of cumulus evolution on mountain-scale near-surface convergence is the divergence following the collapse of a cumulus tower. The 58-day composite seems to confirm this. Near 1500 LT the mean flow briefly becomes slightly katabatic (Fig. 5b). A separation of the 58-day period between days with/without afternoon thunderstorms proves that this divergent flow is a feature of thunderstorm days only (Fig. 9a); the divergence probably is due to the spreading of cold pools from thunderstorms, which usually form in the early afternoon (Fig. 10). It is consistent with the weakened anabatic forcing at this time (Fig. 5a), in particular on thunderstorm days (Fig. 9b), but the effect of thunderstorms is more apparent in the wind field than the pressure difference field, because the cold pool may spread over the foothill stations as well, thus removing the pressure difference effect.

d. Impact of the surface sensible heat flux on solenoidal forcing and anabatic wind

Surface heat fluxes were dramatically different between the 2006 premonsoon dry period and the monsoon wet period. This allows us to study their impact on the solenoidal forcing and the resulting circulation (Fig. 11). The peak daytime sensible heat flux halved from ∼200 to ∼100 W m⁻² following the heavy rains early in the wet period (Figs. 11e,j). The latent heat flux increased, but while this may affect moist convection, it does not directly affect the solenoidal forcing. The amplitude of the diurnal surface temperature cycle decreased by about 40% in the wet period (Figs. 11b,g), because of increased cloudiness and soil moisture. The amplitude of p″ differences between stations decreased accordingly, except for Mt. Bigelow (Figs. 11a,f). This comparison yields strong evidence that the observed diurnal pressure variations are driven by surface heating.

The amplitude of the diurnal cycle of the horizontal pressure difference between the foothill stations and Mt. Bigelow also decreased by about 40% from the dry period to the wet period (Figs. 11c,h), and its phase remained essentially unchanged. The nocturnal katabatic wind was substantially stronger during the dry period (Figs. 11d,i), which is consistent with the higher nocturnal cooling rate. Nocturnal cloudiness and rainfall were not uncommon during the wet period.

The daytime mean anabatic wind was not substantially stronger during the premonsoon dry period. The transition from katabatic to anabatic wind was steeper during the dry period, consistent with the rapid transition from katabatic to anabatic pressure forcing (Fig. 11c) and the rapid increase in surface sensible heat flux (Fig. 11e). Also, the anabatic wind was more likely to continue into the afternoon during the dry period, compared with the monsoon period (Figs. 11d,i), presumably because fewer thunderstorms erupted. But in both periods, the anabatic wind and its horizontal pressure forcing started about 2 h after sunrise and the anabatic wind peaked 1–2 h before LSN.

In short, this comparison indicates that daytime sensible heat flux strongly controls the amplitude of the solenoidal forcing (expressed in terms of a θ_υ difference or horizontal pressure difference), but the strength of the resulting mountain-scale convergence appears less sensitive to surface heating, at least for the Santa Catalina Mountains in summer. We hypothesize that this lack of sensitivity is due to the fact that excessive surface heating increases the chances of moist convection, which produces divergent flow around the mountain. We explore this hypothesis further in the next section.