Observed atmospheric electricity effect on clouds

The atmosphere’s fair weather electric field is a permanent feature, arising from the combination of distant thunderstorms, Earth’s conducting surface, a charged ionosphere and cosmic ray ionization. Despite its ubiquity, no fair weather electricity effect on clouds has been hitherto demonstrated. Here we report surface measurements of radiation emitted and scattered by extensive thin continental cloud, which, after ∼2 min delay, shows changes closely following the fair weather electric field. For typical fluctuations in the fair weather electric field, changes of about 10% are subsequently induced in the diffuse short-wave radiation. These observations are consistent with enhanced production of large cloud droplets from charging at layer cloud edges.


Introduction
The existence of an atmospheric electric field in fair weather 1 has been known for over two centuries [1][2][3][4] with its common universal time diurnal variation in clean air established globally by the Carnegie survey ship in the early 1900s [5,6]. Any role for fair weather electricity in atmospheric processes, and clouds in particular, has, however, never been demonstrated. Classical meteorological investigations speculated on electrostatics as a factor in cloud development [7,8] but, more recently, fair weather atmospheric electricity has been suggested as a physical 'missing link' coupling solar activity with climate [9][10][11]. Physical mechanisms linking fair weather fields with clouds therefore deserve further investigation.
Upper and lower edges of layer clouds are sensitive to fair weather electrification [12], as this is where the air conductivity drops substantially between the clear and cloudy air due to ion removal [13], and droplet charging results [14,15]. The vertical fair weather current, from which the fair weather field originates, passes through the cloud edge conductivity boundary. At a sharp cloud boundary, the fair weather 1 'Fair weather' is used in atmospheric electricity to describe situations in which there is no strong local charge generation. It extends from serene sunny conditions with, formally, no weather, to overcast but non-convective nonraining or non-foggy conditions. current can cause appreciable charging [16]. Using radiation measurements made beneath an extensive layer cloud, changes in cloud properties have been observed to occur shortly after fair weather electric field fluctuations, consistent with chargeinduced changes in droplet properties.

Cloud droplet formation
Formation of cloud droplets requires a water vapour supersaturated environment and particles able to act as cloud condensation nuclei (CCN). In regions with supersaturation greater than the critical supersaturation (typically 1-2% above 100% relative humidity), a condensing droplet will grow, but in regions having insufficient supersaturation, the droplet will evaporate [17]. After condensation, droplets grow by vapour diffusion and droplet-droplet collision (coalescence), the latter providing more rapid growth as droplet size increases.
Both condensation and coalescence can be influenced by charge [18]. Condensation can be facilitated at lower supersaturations if the condensation nucleus concerned is sufficiently highly charged to reduce evaporation [16]; coalescence to large droplets is enhanced by electric forces between charged droplets [19], which, because of electrostatic image forces, are always attractive at small separations whatever the relative polarities of the colliding particles [20]. Fair weather droplet charging should occur at the edges of all liquid layer clouds, but, as small droplets generally co-exist with drizzle and rain, electrically-induced changes at the smallest droplet sizes will usually be obscured. To observe electrically-enhanced growth effects, the cloud droplet distribution needs to be dominated by small droplets. Thin, extensive layer clouds can provide a suitably narrow droplet distribution, biased towards small droplet sizes [21], with fair weather droplet charging occurring at the upper and lower cloud boundaries.
Two responses will occur from droplet number and droplet size changes. Firstly, in response to increasing the droplet number concentration, more scattering of incident solar radiation will occur, which will increase the diffuse shortwave radiation (S d ) at the surface. Secondly, for a thin cloud, the long-wave emissivity decreases slightly with droplet size (at ∼−0.1% per 0.05 μm increase in effective radius) [21], hence a small charge-enhanced increase in the droplet size distribution will slightly reduce the emitted downward longwave radiation (LW d ). (Droplet growth may also increase the short-wave scattering.) Increased droplet charging at the lower edge of a thin cloud can therefore lead rapidly to both an appreciable S d increase and a small LW d decrease.
As well as the short-wave and long-wave surface radiation cloud responses, the initiating atmospheric electricity changes at the cloud can be detected at the surface. In fair weather conditions, the surface electric field, conventionally measured as the potential gradient (PG), is related to the global circuit vertical current density J c by where F is the PG and σ the air conductivity [14]. At the upper and lower cloud boundaries, the space charge density ρ is where ε is the permittivity and σ varies with height z across the cloud-air boundary [12,16]. For constant surface air conductivity, PG variations are therefore directly linked to variations in ρ at the cloud through equations (1) and (2). Conductivity changes across the cloud boundary are typically a factor of ten. In addition, the weak radar returns showed that large drops were absent 3 , suggesting circumstances in which coalescence or condensation processes might be directly observed. Figure 1(a) shows the cloud over the southern UK on 14th February, observed by the MODIS satellite. Figures 1(b) and (c) show time series of surface radiation, wind and PG measurements, beneath the cloud at Reading (51.442 • N, 0.938 • W) for the 13th, 14th and 15th February. Figure 1(b) shows the surface broadband short-wave and long-wave radiative fluxes. The clear conditions on 13th February are apparent from the diurnal cycle evident in the global solar irradiance (S g ), and the small diffuse solar irradiance (S d ); on the 14th February, however, there was a much reduced diurnal cycle in S g , with fully overcast conditions causing the shortwave radiation to be entirely diffuse (S g = S d ). There was also negligible diurnal variation in the downward long-wave radiation (LW d ) on the 14th February. Figure 1(c) shows the wind speed at 10 m (u 10 ), typical of steady local conditions. Appreciable, but not exceptional, high frequency variability was apparent in the PG, unlikely to be of local origin because of the quiescent conditions.  Figures 2(c) and (d) present the same data portions after high pass filtering, retaining variations with timescales of less than 1 h. A correlation is clearly evident between the filtered PG and S d or LW d . The possibility of an instrumentation cause can be rejected as additional sensors (for S g , LW u , net radiation R n , and soil heat flux G) operating at the site used identical electronics [22] and the same logging system as for S d and LW d . In these additional measurements (not shown), both the S g and R n data were dominated by the diffuse short-wave radiation and were also correlated with the PG, but G and LW u , which vary slowly because of damping effects of the soil, were not. The effect was therefore radiative, originating in the lower atmosphere.

Analysis
The filtered data suggest a cloud response after the atmospheric electricity changes, as the PG changes generally occur before the radiative changes. Figure 3 demonstrates this lag effect at higher temporal resolution, using 20 s averages of the filtered data, through the use of compositing. 'Compositing' 4 averages together data obtained either side of a defined event, allowing the mean behaviour in response to many such events to emerge from the natural variability. (This variability usually obscures the effects around a single event.) Here, the LW d and S d data are composited around local PG extrema to investigate the radiative response to PG changes, using data during the steady daytime conditions from 10 UT to 1530 UT with timescales longer than 2 h removed.
The compositing reveals both a significant S d increase and LW d decrease 1-2 min after a PG increase. If a PG increase were followed by an increase in droplet size and/or number around the cloud base, such an increase in S d and decrease in LW d would be a consequence. Proportionality between PG change and the sub-cloud radiative changes, allowing for a lag of 100 s, is summarized in the scatter plots in the lower panels of figure 3. For the averaged PG increase of 40 V m −1 (∼25% of the mean), the subsequent S d change was 6 W m −2 (∼10%) and the LW d change was −0.3 W m −2 (∼0.1%).
The lag in the radiative response cannot arise from instrument response times as these are much shorter than the observed lag 5 . 4 'Compositing' is sometimes also known as a 'superposed epoch' analysis. 5 All the instruments were continuously sampled at 1 Hz and 12 bit resolution. Specifically, the short-wave and long-wave radiation measurements were made using Kipp & Zonen CMP11 and CNR1 radiometers respectively, with a solar tracker and shade ball used to occult the short-wave radiometer for the S d measurement. The PG was recorded with a Chubb Instruments JCI131 field mill. The radiometers' specified 95% time responses are 5 s (CMP11) and 18 s (CNR1), and the −3 dB cut-off frequency for the JCI131 is 7 Hz. Thus the PG response is effectively instantaneous with 1 Hz sampling, and, for averaging periods of 20 s or more, the radiometer response time effects will be small. In any case, both radiometers will respond several times more rapidly than the atmospheric lag time of ∼100 s observed. The observed lag is unlikely to be the result of a geometric effect in which the radiation instruments sample downstream clouds whilst the field mill samples directly overhead. A simultaneous measurement of the direct solar beam (S b ) was also made using a Kipp & Zonen CH1 pyrheliometer (5 • view angle, 7 s time response to 95%) mounted on the solar tracker. During overcast conditions this samples the cloud close to the sun's position, where there is spatial anisotropy in the diffuse radiation source. No evidence for a lag effect in this region of the sky was found in the S b data, and negligible variability was apparent in S b compared with that present in S d . Furthermore, for this alternative explanation that the PG and S d modulation were simply due to propagating cloud thickness changes, there would be an inverse relation between PG and S d , which is the opposite of that observed.

Conclusions
These observations provide evidence of a direct effect of global atmospheric electricity on clouds. Because of the global presence of the fair weather current, similar effects may occur more generally in layer clouds. For typical variations in global fair weather current, the resulting change in thin clouds would correspond to the 6 W m −2 found here, which would affect the atmosphere's local radiative balance. Our results also provide evidence for charge-induced cloud microphysics, for example aiding coalescence [19] to larger raindrops. The initiation of raindrop coalescence remains an unsolved problem in cloud physics [23].