MONIM: the new Met Office Night Illumination Model

Transcription

1 Meteorol. Appl., 9 () DOI:./S MONIM: the new Met Office Night Illumination Model S. J. Revell & P. Hignett Met Office, Met Research Unit, Field Site, Cardington Airfield, Shortstown, Bedfordshire MK SY, UK This paper describes a new model developed by the Met Office to predict night-time light levels. The Met Office Night Illumination Model (MONIM) predicts light levels both in the visible (photopic) range and in the waveband to which night vision goggles (NVGs) are sensitive. The model will be used operationally for support of night-time flying operations. The model is described in detail and its light-level forecasts are compared with observations.. Background Within military aviation there is an increasing emphasis on night operations. In line with this there is heavy reliance on electro-optical devices, including night vision goggles (NVGs). NVGs are image intensifiers which amplify ambient light in the near infra-red and project the image onto a phosphor screen. Although the sensitivity of the NVGs is being continually improved, under any given set of conditions there is still a lower limit to the ambient light level below which they are ineffective. Therefore, there is a requirement to forecast night-time light levels, both for operational sorties and for longer-term planning. In the United Kingdom, and throughout NATO, this forecasting is routinely provided by the national, or relevant military, meteorological service. There are several Tactical Decision Aids (TDAs) in current operational use, in various countries, which have been developed for NVG support. Mostly these predict light levels in the visible, or photopic, region of the spectrum at wavelengths to which the human eye responds. Modern NVGs are most sensitive at near infrared wavelengths, so more accurate predictions of their performance could be achieved if forecasts were given for this spectral region. To meet this requirement the Met Office Night Illumination Model, MONIM, has been developed as a TDA for use by forecasters and flight planners to give aircrew guidance on the expected light levels in both the visible and near infrared. This helps ensure continuity with previous TDAs and practice in other countries. For operational sorties it is important that the TDA is simple to use and runs quickly, so that forecasts can be produced in a timely manner. The TDA is available to a wide range of users, both internal and external to the Met Office, so the TDA must be portable to other PC platforms and able to run on stand-alone systems. This means that it is not possible to have an automated data feed, so data entry must be kept to a minimum. Verification against observations is vitally important as it tests the model output and gives forecasters and endusers confidence in the TDA output.. Night vision goggles The NVGs used most commonly in the UK at the present time incorporate Generation III (Gen III) intensifier tubes. Typically the NVGs consist of two such tubes, to give binocular vision, although some systems work with only one. The Gen III image intensifier tubes use a gallium arsenide photocathode, which responds to light in the approximate range to 9 nm, with a maximum response in the near infrared region between and nm. This is significantly different from those wavelengths to which the human eye responds, which lie in a relatively narrow band between about and nm. This is known as the photopic region and, for scientific use, is defined by the CIE (Commission International pour l Eclairage) standard human eye response. This is the basis of photometry and all photopic measurements and units (e.g. lux and lumens), by definition, lie within the CIE response. There is only a small overlap between the photopic spectral response and the typical Gen III response, as shown in Figure. Light received by the image intensifier tubes is focused by an objective lens onto the photocathode, where electrons are released, so beginning a cascade effect. These electrons are accelerated and multiplied by a microchannel plate, which guides them to a phosphor screen, producing an intensified light image. This screen provides a virtually monochrome green image, so no

2 S. J. Revell & P. Hignett Relative Response Wavelength (microns) (ii) (iii) (iv) moonlight has proportionally more red and near infrared light. Twilight. This is light scattered from the sun below the horizon. Starlight. Taken over the whole sky hemisphere, the average starlight at the earth s surface is relatively redder than sunlight. The light level is not constant but varies to some extent according to location and time of night. The brighter planets, particularly Venus, will also have an influence. Airglow. This is a faint luminescence of the upper atmosphere and often produces more radiation at the surface than starlight. It is very uniform across the sky and should not be confused with the much brighter aurorae. Figure. Relative responses for CIE Human Eye (solid) and Generation III NVG (dotted). colour discrimination is possible. Gen III intensifier tubes have a maximum overall gain, electrons produced to electrons generated at the photocathode, of. However, an automatic gain control is used to limit the maximum average brightness of the phosphor screen to a typical luminance of cd m. This lies in the mesopic region of the human eye response and is the transition zone between normal daylight vision (photopic response) and fully dark-adapted vision (scotopic response). Effectively, the overall gain of the NVGs responds inversely to the ambient light level, with the consequence that a steady reduction in ambient light appears as a perceived increase in the image noise, and hence a degradation in the image contrast, rather than a reduction in the image brightness. Also protective circuitry is incorporated to prevent excessive blooming from very bright sources such as flares, explosions, afterburners, etc.; however, it is still necessary to avoid bright sources, such as a full moon, as this can saturate the tubes causing them to cut out. At very low light levels a point will be reached when the image becomes too noisy for operational use. However, even before this point, visual acuity through NVGs declines rapidly at light levels below mlx (Pincus & Task 99). In flight with NVGs it is difficult to detect a gradual increase in cloud that will lead to a lowering of the light level, with a consequent reduction in NVG performance. A principal function of the TDA, therefore, is to warn NVG users of the potential occurrence of conditions leading to light levels in the range that will significantly compromise the goggles performance. Night-time radiation in the visible and near infrared comes from four main natural sources: (i) Moonlight. This is reflected sunlight but, because the reflectance of the moon s surface in the near infrared increases relative to that in the visible, In addition to these four natural sources of light, NVGs will also be sensitive to cultural lighting, which is scattered light from domestic, industrial, road and street lighting. Cultural light is reflected from the base of thick cloud so has most impact under cloudy conditions. The spectral output of common lighting types varies; the older low pressure sodium vapour lamps give fairly monochromatic orange-yellow light to which Gen III tubes are not very sensitive. However, the newer fluorescent, mercury and high pressure sodium lamps have a much broader spectral output and will therefore have a greater effect on NVGs. Overall, there is no sharp threshold to NVG performance at low light levels. Several factors become important, including the eyesight of the individual user, others are similar to those that affect visibility with the unaided eye. For example, the presence of dust or fog will reduce visibility and targets will be difficult to detect if the visual contrast between the target and its background is poor.. Calculating night-time light levels The techniques for calculating light levels are conceptually straightforward and involve two aspects: astronomical and atmospheric. The light incident at the top of the atmosphere is determined by the position of the sun and moon relative to the earth. The amount of light that reaches the surface is reduced by scattering and absorption due to gases, aerosol and cloud particles in the atmosphere. Previous TDAs used by the Met Office to predict night-time light levels for NVG support have predicted the photopic light levels at the earth s surface from empirical relationships based on the moon s phase and position (Turton & Stone 99). Similar empirical relationships were used to calculate twilight from the sun s position. Since the model relied on empirical relationships calculated for the photopic region only it was not possible to use these models to calculate light levels in other wavebands.

3 Met Office Night Illumination Model MONIM MONIM uses a different approach; having calculated the top of atmosphere (TOA) irradiance for the photopic and NVG wavebands (see section.), a simple spectral model, SPCTRAL (Bird & Riordan 9), is used to calculate the attenuation of the light through the atmosphere. SPCTRAL is a column transmittance model and, in the form used in MONIM, allows for extinction by Rayleigh scattering, ozone, water vapour and other mixed gases. Extinction by aerosol and cloud can be added in a straightforward manner. There is an implied assumption that the surface albedo is characteristic of a land surface and is approximately equal to.. For moonlight and twilight calculations the lunar and solar zenith angles are needed. They are generated by an almanac subroutine which takes latitude, longitude and time as input, and outputs the azimuth and zenith angles for the sun and moon, and the lunar phase; leap years and century years are accurately accounted for. Calculations are carried out only for the period when the sun is more than. o below the horizon... Moonlight In SPCTRAL the surface irradiance is calculated by multiplying the top-of-atmosphere irradiance by a series of transmittances to allow for extinction by Rayleigh scattering and gaseous absorption. The direct and diffuse components are calculated separately and then summed to give the total. The top-ofatmosphere lunar spectral irradiance is taken as the extra-terrestrial solar spectrum multiplied by the lunar spectral reflectance as used in LOWTRAN. The annual variation in the earth sun distance is allowed for, but no account is taken of the relative motion of the moon around the earth; the error incurred is less than.%. The lunar surface is a non-lambertian reflector so the reflectance is weighted by a function dependent on the lunar phase. This weighting function, from the USAF Geophysics Laboratory Handbook, is equal to one for a full moon (zero phase angle) but is not symmetrical, being slightly larger for a waxing moon (negative phase angle) than the corresponding values for a waning moon (positive phase angle). This asymmetry accounts for differences in the reflectance across the moon s surface. By integrating the horizontal lunar spectral irradiance at the earth s surface over the CIE and NVG responses illustrated in Figure, the band irradiances are calculated. For the photopic region the illuminance is calculated by multiplying by the luminous efficacy factor, defined such that at nm, watt of radiant power is the equivalent of lumens of luminous power. SPCTRAL requires specification of the integrated column amounts of water vapour and ozone. These were set at. cm and. cm respectively. At the wavelength range used here gaseous absorption is comparatively weak. Varying the water vapour column from a minimum of to a typical global maximum of cm changes the illuminance by typically.% and the NVG irradiance by.%. Similarly, varying the ozone amount from to a typical maximum of. cm changes the illuminance by % and the NVG irradiance by.%... Twilight As twilight is scattered solar radiation received when the sun is below the horizon, an exact treatment would be complex. At mid-latitudes typical of the UK, the period over which twilight is the main source is relatively short and the irradiance changes rapidly with time (approximately an order of magnitude every minutes). At higher latitudes, where there is persistent twilight for extended parts of the year, the light levels tend to be well above the threshold of NVG use. In both cases the absolute accuracy requirement is not high since the light levels at twilight are well above the operational minimum for NVG use (typically around to mlx) and an empirical approach is sufficient. Using observational data on twilight photopic light levels originally obtained by Brown (9), equation () expresses the horizontal illuminance, I, at the Earth s surface, in millilux (mlx), as a function of the solar zenith angle, z. log I =. z.z +.99z. () At the time of producing the TDA there were insufficient observational data to generate an analogous expression for the NVG band. On the assumption that the spectral distribution of twilight is similar to that of the diffuse sky radiation when the sun is just above the horizon, which can be calculated, a scaling factor was derived to generate the NVG band irradiance from the illuminance given by equation (). This is evaluated in section, and could be modified in the light of further observational evidence... Starlight and airglow Under a moonless night sky, when the sun is well below the horizon, starlight and airglow are assumed to account for all available natural light. The variability in light levels from starlight and airglow is not well characterised. Predicting exact values of starlight would require extensive knowledge of the stars which are above the horizon at any given time. Contributions from lower magnitude stars would almost certainly have to be parametrised since the numbers involved would be too high. A recent study carried out in the Met Office concluded that variations in starlight owing to changes in stellar configuration are not significant compared with variations in airglow ( J. Swift, personal communication). Taken over the whole sky hemisphere,

4 S. J. Revell & P. Hignett the average starlight at the earth s surface is relatively redder than sunlight. Airglow is a faint luminescence that occurs as a result of the interaction between certain constituents of the upper atmosphere and external factors such as sunlight or charged particles. The interaction results in a physical process in which photons are emitted; spectrally, airglow has emissions at green ( nm) and yellow (9 nm) wavelengths, with a broad red and infrared emission from about to nm. The intensity of airglow is comparable to that of starlight but is affected by solar variations and consequently there will be a long-term variation in line with the solar cycle. Significant spatial and short-term temporal variations have also been reported (Roach & Gordon 9). Owing to the difficulties in accurately forecasting starlight and airglow, MONIM, in common with other TDAs, uses a single constant value of millilux to represent the combined contribution. The equivalent irradiance for the NVG band has been generated by applying a scaling factor calculated from a model average starlight spectrum... Effects of clouds The SPCTRAL model calculates the irradiance at the Earth s surface for clear sky conditions. Cloud cover can have a significant effect and needs to be taken into account; however, the limitations imposed by a stand-alone application mean only a very simple approach can be adopted. MONIM currently uses the same parametrisation for the effect of cloud on the surface illumination as its predecessor ILLUMW. An attenuation factor for various cloud types was calculated by parametrising the calculations made in the US Atmospheric Sciences Laboratory ILLUMA model (Duncan et al. 9). By running the ILLUMA model for a limited range of cloud types at set heights, attenuation factors were derived as a function of lunar elevation. These calculations were made for fully overcast conditions. The four different cloud types were: cirrostratus at 9 ft, altostratus at 9 ft, stratus at 9 ft and nimbostratus at ft and these are labelled High Cloud, Medium-Level Cloud, Low Cloud and Solid Cloud respectively. The cloud types were limited to this set of four to minimise the amount of detailed information that has to be entered to run the model. Therefore only broad guidance can be given on the likely impact of cloud... Lunar eclipse During a lunar eclipse the natural light levels will be reduced. MONIM predicts the occurrence of an umbral eclipse but does not predict the light levels during the eclipse.. The operational TDA MONIM has been developed as a Windows-based application to run under Microsoft Windows NT. It does not rely on any automatic data-feed and can, therefore, run on stand-alone PCs... Model inputs For ease of operation the inputs required for the model have been kept to a minimum. First, the start and end dates for the required forecast are entered. This is done by clicking on the appropriate dates from a calendarbased display. Next the location for the forecast is entered. The model is initialised with a comprehensive set of UK-based locations but other locations can easily be added. For each new location added the latitude and longitude are required; these can be specified either in degrees and decimals or degrees, minutes and seconds. If the output is required in local time rather than UTC then a UTC Offset can be specified for each location. The offset is the time difference, in hours, between the local time and UTC. After running the model the forecast cloud cover can be entered manually. A cloud type (high, medium, low or solid) can be entered at hourly intervals or the default clear sky value can be kept... Model outputs The output is presented in both tabular and graphical formats. The light levels and the astronomical data (sun and moon elevation and bearing and lunar phase) can all be displayed as a time-series with data points plotted every minutes. One -hour period from midday to midday is displayed at any one time. If cloud forecasts have been entered, the light levels for both clear sky and the forecast cloud conditions are shown. Hourly values of the illumination levels and the astronomical data can also be displayed in tabular format. The sunrise, sunset, moonrise, moonset and twilight times are also presented in tables. The sunrise, sunset and twilight times are determined by the elevation of the sun. Sunrise and sunset occur when the zenith angle of the centre of the sun is 9, i.e. it is (the sun s angular semi-diameter) below the horizon. Civil twilight occurs at a solar zenith angle of 9, nautical twilight at and astronomical twilight at. Moonrise and moonset occur when the lunar zenith angle is 9 + s p, where s and p are the moon s angular semi-diameter and parallax respectively. The parallax is calculated by the model but for simplicity the moon s semi-diameter is set at. The semi-diameter can vary from. to. depending on the distance from the earth to the moon; however, setting it to a

5 Met Office Night Illumination Model MONIM Model Predictions (mlx) Photometer Data (mlx) Figure. Scatter plot of observed and predicted mlx values. Dotted lines are ± %. constant value introduces errors of less than minutes in the moonrise and moonset times and does not affect the illumination calculations... Illumination levels For the photopic region the light level outputs are given as the illuminance on a horizontal surface in millilux (mlx). For the NVG band the photometer used for validation (see section. for more details of the instrument) defines an arbitrary unit, termed Nvis-millilux (Nvis-mlx), which provides the user with a measure of the light level available, whose behaviour is comparable to the photopic value expressed in millilux. The definition of Nvis-millilux derives from the calibration of the photometer, which is typically carried out against a lux K source. The output of the NVG band is set according to a weighting factor derived from an average starlight spectrum, such that the millilux and Nvis-millilux values will converge at light intensities and spectral distributions typical of starlight. Typical natural light levels vary from. mlx for an overcast moonless night to mlx in direct sunlight. Twilight covers the range from about mlx to mlx and NVGs will operate down to approximately mlx.. Model validation.. Photometer measurements In order to validate the results of the model, measurements of night-time light levels were made with a Hoffman ANV-A photometer. The ANV-A is an all-weather photometer designed specifically for measuring ambient night sky illumination. It has two photosensors, one which operates in the photopic range and one which mimics the sensitivity of NVGs. The photometer is connected to a laptop computer which logs the light levels measured by both photosensors at -minute intervals. Before the start of the trial the photometer was calibrated at the National Physical Laboratory against a K tungsten filament lamp over the range to mlx. The measurement uncertainty was found to be ± %. The standard error on the means of the measurements of the ratio between the photopic detector reading and the Nvis detector reading was ±.%. In order to minimise the effects of artificial, or cultural, lighting it is important to choose a site which is well away from large areas of habitation. A Met Office site in Dumfriesshire, Scotland was chosen. The site is at Eskdalemuir,. N,. W, where the only local source of lighting is from the near-by observatory, which is screened behind a stand of trees. With the exception of a -week period in August, the nighttime illumination data have been recorded continuously from January to April... Comparison with observations Figures and are scatter plots of cloud-free observations against the MONIM predictions for light levels in the photopic and NVG bands respectively over the full observation period from January to April. The observations, sampled at -minute intervals, have been matched to the nearest MONIM prediction, made every minutes. The dotted lines enclose the ± % region. The assumptions made in section. to generate twilight values for the NVG band have led to a systematic under-prediction of the corresponding Nvismlx values on Figure. From an operational perspective

6 S. J. Revell & P. Hignett Model Predictions (Nvis-mlx) Photometer Data (Nvis-mlx) Figure. Scatter plot of observed and predicted Nvis-mlx values. Dotted lines are ± %. Frequency (%) < >. Range Figure. Histogram of frequency distribution of starlight and airglow for the photopic (black) and Nvis (grey) ranges. this is less critical than over-prediction, and can be rectified in a subsequent version. There is no evidence of an analogous systematic bias in the photopic levels shown in Figure. In this dataset, twilight occupies a relatively short part of the night, which accounts for there being comparatively fewer points greater than about mlx. Below this level moonlight is the major light source for the bulk of the observations. On both Figures and there is reasonably good systematic agreement between observation and prediction in the to mlx range. The behaviour at the lowest values arises from the minimum prediction under cloud-free skies of mlx in the photopic region, corresponding to. mlx for the NVG band. The lowest levels are encountered when the moon is below the horizon and the only sources are starlight and airglow. Figure shows the frequency of occurrence of the observations made under these conditions. The range and shape of the distributions is similar, with the ratio of their means (photopic to NVG) equal to., compared to the modelled value of.. Meteorological factors can account for some of this observed variation. For example, a reduction in the horizontal visual range to km, if maintained over a depth of km, causes a reduction in the illuminance of %. The factor is somewhat less for the NVG band because of the lower aerosol scattering at these wavelengths. These figures were generated by running the SPCTRAL model with parameters appropriate to a pure scattering sulphate aerosol. Unobserved cloud is also a likely factor. A significant influence, although it cannot be verified directly here, is simply the natural variation arising from the motion of the planets and brighter stars, and changes in the airglow. The assumed fixed values of mlx, and the corresponding. Nvis-mlx, are a little higher than the means observed at this location, but further data from different sites would be needed before any change could be justified.

7 Light Level (mlx) Met Office Night Illumination Model MONIM : : : : : : : : : (a) Light Level (Nvis-mlx) : : : : : : : : : (b) Figure. Comparison of measurements (solid) and clear-sky predictions (dotted) for a full moon at Eskdalemuir on /February for (a) the photopic and (b) the NVG ranges. Cloud cover (crosses) is plotted on the secondary axis... Interpretation of specific cases On any given occasion some interpretation of the MONIM output is necessary to give guidance on the actual conditions that may be encountered. Assessment of cloud impact is difficult when the cloud cover is broken and when the moon is close to the horizon. Use of the cloud attenuation factors can provide an estimation of the worst conditions likely. To illustrate some of these points three cases have been chosen, as described below. The first case, shown in Figure, is for a full moon on / February. Sunset and sunrise were at : UTC and : UTC respectively and moonrise and moonset were at : UTC and : UTC respectively. The basic presentation in Figure (and subsequently in Figures and ) is the same as that provided to the user by MONIM. It is important that information can be assimilated quickly, and so a logarithmic scale is used as the ordinate, for clarity. This emphasises the critical region of low light levels, whilst still clearly indicating those periods when levels will change rapidly. Figure a is a time series showing the photopic photometer readings (solid line) and the MONIM photopic predictions (dotted line) for clear skies. The hourly observation of cloud cover is plotted in octas on the secondary y-axis. Figure b is as Figure a but the NVG band observations and clear sky predictions are plotted. Although the moon is close to the horizon at both dusk and dawn the cloud effect differs. At dusk the cloud cover of to octas seems to lead to a lower observed light level than predicted. However, at dawn there is an enhancement of the light level which may be due to reflection from the base of the relatively low amount of cloud present. The second case, shown in Figure, is from / December when moonrise was at :9 UTC and the moon phase was around %. Sunset and sunrise were at : UTC and : UTC respectively. Here there was substantial variation in the cloud cover throughout the night, the moon is not close to the horizon at either dawn or dusk and the general effect of the greater cloud cover is to cause a reduction of the light level relative to the clear sky prediction. The last case, shown in Figure, is from / February when there was a new moon. The sunset and sunrise times were : UTC and :9 UTC respectively. In this case there is natural light from starlight and airglow only. The fixed values used by

8 S. J. Revell & P. Hignett Light Level (mlx) : : 9: : : : : : : (a) Light Level (Nvis-mlx) : : 9: : : : : : : (b) Figure. Comparison of measurements (solid) and clear-sky predictions (dotted) for a % moon phase at Eskdalemuir on / December for (a) the photopic and (b) the NVG ranges. Cloud cover (crosses) is plotted on the secondary axis. Light Level (mlx) : : : : : : : : (a) Light Level (Nvis-mlx). : : : : : : : : (b) Figure. Comparison of measurements (solid) and clear-sky predictions (dotted) for a new moon at Eskdalemuir on / February for (a) the photopic and (b) the NVG ranges. Cloud cover (crosses) is plotted on the secondary axis.

9 Met Office Night Illumination Model MONIM MONIM for both the photopic and NVG bands are a little higher than the observations, which show a slow modulation through the night, consistent with the range shown in Figure.. Summary MONIM has been developed as a quick-response tool for forecasting night-time light levels in both the photopic and NVG wavebands. The main advantages of MONIM over existing models is that MONIM can be used to calculate light levels in the NVG band and the physically-based model can be more easily developed to meet changing requirements and to better represent the effects of cloud. The model has been verified against observations using a photometer located at Eskdalemuir. A quantitative comparison of light levels from the photometer and MONIM shows good agreement, especially at lower light levels where the predictions are more critical to end-users. Some disagreement occurs at higher light levels around dusk and dawn and these will be investigated further. Acknowledgements This work was funded by the Ministry of Defence Meteorological Support Group Research and Technology Programme. The authors would like to express their gratitude to the Met Office staff at the Meteorological Research Unit, Cardington and at the Eskdalemuir Observatory for their assistance in collecting the photometer data. References Bird, R. E. & Riordan, C. (9) Simple solar spectral model for direct and diffuse irradiance on horizontal and tilted planes at the Earth s surface for cloudless atmospheres. J. Climate Appl. Meteorol. : 9. Brown, D. R. E. (9) Natural illumination charts. Washington D. C., Department of the Navy, Bureau of Ships, Report No. -. Duncan, L. D., Sauter, D. P. & Miller, A. (9) Natural illumination under realistic weather conditions ILLUMA. EOSAEL, Vol, TR-. Atmospheric Sciences Laboratory, White Sands Missile Range, NM -. Pincus, A. R. & Task, H. L. (99) Measuring observed visual acuity through night vision goggles. SAFE Symposium Proceedings 99, th Annual Symposium, September pp.. Roach, F. E. & Gordon, J. L. (9) TheLightoftheNight Sky. D. Reidel Pub. Co. Turton, J. D. & Stone, G. D. (99) Forecasting night-time illumination. Meteorol. Mag. : 9. 9