Evolution of Stratocumulus Over Land: Comparison of Ground and Aircraft Observations with Numerical Weather Prediction Simulations

Abstract

Forecasting of low cloud continues to challenge numerical weather prediction. With this in mind, surface and airborne observations were made over East Anglia, UK, during March 2011 to investigate stratus and stratocumulus advecting from the sea over land. Four surface sites were deployed at various distances inland aligned approximately along the flow. In situ data include cloud-droplet measurements from an aircraft operating off the coast and a tethered balloon 100 km inland. Comparisons of thermodynamic and cloud properties are made with Met Office operational model simulations at horizontal resolutions of 4 and 1.5 km. The clouds contained droplet concentrations up to 600 cm$^{-3}$ within polluted outflow off Europe. These measurements were compared to three different model schemes for predicting droplet concentration: two of them perform well at low to moderate concentrations but asymptote to 375 cm$^{-3}$. Microwave radiometers at the ground sites retrieved liquid water paths that reduced with distance inland and were generally below 200 g m$^{-2}$. The modelled water path performs well upstream but more erratically far inland. Comparisons of thermodynamic profiles are made within both Lagrangian and Eulerian frameworks and show the model predicted changes in equivalent potential temperature generally within 1 K, with occasional errors of 2 K or more. The modelled cloud-top temperatures were in good agreement with the observations down to $-$7 $^{\circ }$C, but the magnitude of the temperature inversion, although good at times, was too small by on average 1.6 K. The different simulations produced different cloud-top water contents due to a combination of resolution and scientific upgrades to the model, but they generally underestimate the amount of cloud water. Major changes, such as the mesoscale temporary cloud breaks on 2 March 2011 and the complete clearance on 4 March, were seemingly predicted by the model for the correct reasons.

Appendices

Appendix 1: Ground Instrumentation

The $30\,\hbox {m}^{3}$ helium-filled tethered balloon payload used during COALESC comprised a light-weight turbulence probe and a DMT CDP. The CDP is essentially the same as the aircraft unit but modified to operate at lower air speeds (Price 2011). The turbulence probe is a miniaturized version of that described in Lapworth and Mason (1988). The turbulence probe can provide three-component vector winds up to 20 Hz and this can be combined with its fast-response platinum-resistance thermometer (PRT) to compute heat flux. We use the mean wind speed from the turbulence probe to calculate the CDP sample volume. In addition to the PRT, the turbulence probe also houses a slow-response thermistor and a Vaisala HMP44 humicap, both mounted in an aspirated housing. The CDP and turbulence probe were mounted 3 m apart on the vertical tether beneath the balloon (and 60 m beneath the balloon itself), such that small errors in flow velocity across the CDP sample area are possible. But as the CDP data are averaged over at least 60 s, then flow velocity and time synchronization ($\approx$1 s) errors are negligible.

Continuous column-integrated vapour and liquid water monitoring with passive microwave radiometers was established at Weybourne, Denver Sluice in addition to the permanent deployment at Chilbolton. The measured brightness temperatures (at approximately 23 and 32 GHz) are first converted into optical depths and then a regression method is used to convert the optical depths into LWP values. This regression, detailed in Price (2003), is based on LWP values from integrated radiosonde ascents that assume adiabatic ascent. The same retrieval algorithm has been applied to all three radiometer datasets. During the experiments Vaisala RS92 radiosondes were released from Weybourne, Denver and Cardington at 60–90 min intervals. Radiosondes transmit data on pressure, temperature, relative humidity and winds at 0.5 Hz with a 5–6 m height resolution.

There were cloud-base ceilometers installed at Cardington, Denver (both Vaisala LD40) and Chilbolton (Vaisala CT75K). Shortwave (Kipp & Zonen CM22 pyranometers, 0.2–2.8 $\upmu$m) and longwave (Kipp & Zonen CG4 pyrgeometer, 4.5–40 $\upmu$m) broad-band radiometers are part of the permanent instrumental fits at Cardington and Chilbolton. They measure hemispherical irradiance with a nominal 3 $\%$ accuracy. At Cardington both downwelling and upwelling irradiances are measured; at Chilbolton only shortwave downwelling and longwave upwelling irradiances are measured. The Chilbolton 35 GHz cloud radar, whereby typical liquid cloud droplets act as Rayleigh scatterers, provides a measure of cloud-top height.

For aerosol measurements on the ground at Cardington, a Grimm 1.109 aerosol monitor sampling at 3.5 m height was operated. This measures across the diameter range 0.25–30 $\upmu$m and so covers the accumulation and coarse mode. Its minimum size is similar to that of the aircraft PCASP and so the total concentrations from the two instruments can be compared.

Appendix 2: Aircraft Instrumentation

A wing-mounted CDP was used to count cloud droplets and determine their size for nominal diameters of 2–50 $\upmu$m over 30 size channels. The CDP measures the forward scatter (4–12 $^{\circ }$) from droplets interacting with 658 nm light from a diode laser. Channel 1 is ignored in the CDP data analysis, such that the minimum size detection was 3.5 $\upmu$m diameter. With a fixed laser cross-sectional area ($0.24\,\hbox {mm}^{2}$), knowledge of the sample volume is directly related to the air flow velocity.

Liquid water content can be measured by integrating the CDP size distribution but also by using the Nevzorov hot-wire devices. The efficiency of the Nevzorov is reduced in drizzling clouds where drop sizes are in excess of 50 $\upmu$m because the drops are too large to be collected and hence completely evaporated by the hot wire. A baseline drift of the instrument also needs to be removed during cloud-free periods due to the effects of pressure, temperature and true air speed. Water vapour mixing ratio on the aircraft on 1, 2 and 16 March was measured with a SpectraSensors Water Vapor Sensing System (WVSS-II) using a technique of absorption spectroscopy. The manufacturer quotes an accuracy of 5 $\%$. Rosemount/Goodrich 102 total air temperature sensors were used as a measure of air temperature and the true air speed and wind components were derived from a five-port turbulence probe mounted on the nose of the aircraft.

An elastic backscattering lidar with daytime capability, suitable for aerosol and cloud observations, was fitted to the aircraft. It operates at 355 nm, has depolarization, and is mounted for nadir view (Marenco et al. 2011). Clouds are detected using a system of thresholds for the uncalibrated range-corrected signal $P$, capable to highlight large spikes due to the beam hitting an optically dense target. A cloud top is detected if: $P(R)> 4000; P(R)> 3P (R\ - 200)$; and also if no other cloud top has been found between ($R\ - 500$) and $R$, where $R$ is the range from the lidar to the observed layer (in metres). Once a cloud is detected, its top range $R_{c}$ is set as the first point starting at $R$ and moving inward to ($R\ - 200$), where $P(R_{c}) < 1.5P (R\ - 200)$. Clouds identified within 500 m of the terrain are automatically discarded as surface return, although this limit was a little too high for some of the very low cloud in Fig. 3.

A DMT passive cavity aerosol spectrometer probe (PCASP) was used to count and size dehydrated aerosol particles out of cloud for diameters 0.2–3 $\upmu$m i.e. covers the accumulation mode. The total concentration of the PCASP can be used as a measure of particulate pollution. Because the PCASP covers the accumulation mode that contains the majority of cloud condensation nuclei at supersaturations typically found in stratocumulus, the PCASP concentration measured below cloud can be related to the cloud-droplet concentration (Jones et al. 1994).

Appendix 3: Droplet Generation Schemes

The MURK visibility scheme (Clark et al. 2008) is used as a basis of a formulation to diagnose $N_\mathrm{d}$; this was implemented in model version PS26, allowing better sensitivity to air mass. This converts aerosol mass ($m_\mathrm{aer}$), a forecast product in itself, into aerosol number ($N_\mathrm{aer}$) using the method shown in Haywood et al. (2008),

$$\begin{aligned} N_\mathrm{aer} = n_{0}\left( \frac{m_\mathrm{aer}}{m_{0}}\right) ^\rho \end{aligned}$$

(1)

where $n_{0}$, $m_{0}$ and $\rho$ are tuneable parameters: $\rho$ was fixed at 0.5; for PS26–30, $n_{0} = 2\times 10^{9}$ and $m_{0} = 1.8956\times 10^{-8}$; for PS31 and later versions $n_{0} = 5\times 10^{8}$ and $m_{0} = 1.458\times 10^{-8}$. $N_\mathrm{aer}$ is then converted into $N_\mathrm{d}$ using the Jones et al. (1994) curve (which asymptotes to $375\,\hbox {cm}^{-3}$) that has been used in the climate version of the MetUM,

$$\begin{aligned} N_\mathrm{d} = 3.75\times 10^{8}\left[ 1.0-\mathrm{exp}(-2.5\times 10^{-9}N_\mathrm{aer})\right] . \end{aligned}$$

(2)

In the analysis presented here we have used the PS26 and PS31 formulation of Eqs. 1 and 2 to predict $N_\mathrm{d}$. We also included a third test method that uses Eq. 2 but an alternative to Eq. 1 called the CLASSIC aerosol scheme used for climate modelling (Bellouin et al. 2011). The CLASSIC scheme assumes a log-normal aerosol size distribution with fixed mean radius (0.11 $\upmu$m) and standard deviation (1.35) for a particular aerosol species (in this case the Haywood et al. (2008) VISURB aerosol) so that only $N_\mathrm{aer}$ increases with increasing $m_\mathrm{aer}$. This is in contrast to Eq. 1 where the mean radius implicitly increases with $m_\mathrm{aer}$ (as well as number) because $\rho$ is fixed. The value of $\rho$ is that prescribed by Clark et al. (2008) on the basis of observations in Haywood et al. (2008).