Sunshine Analysis

Current Sunshine Analysis Images

[See below for Background and Comments.]



Measurement of Bright Sunshine - Background

Many modern automatic weather stations can be fitted with a sensor to measure sunshine intensity (also known as solar irradiance), which gives a continuous electronic measure of the sunshine's brightness.

Solar sensors come in different types. The more sophisticated ones either point directly at the sun and are motorised to track its daily path across the sky or, alternatively, are used in fixed pairs with one sensor shaded from direct sunlight. There are further types that can detect the contrast level in shadows. All of these more advanced sensors can usually separate out total or global sunshine into its constituent direct and indirect components, but cost inevitably limits their use to high-specification stations. More affordable stations such as the Davis VP2 range can also be fitted with a solar sensor, but for cost reasons this is a single fixed sensor, which measures the brightness of the whole sky and therefore provides readings of overall (ie total or global) sunshine intensity only. (NB Solar sensors typically measure the visible wavelengths of sunlight. UV sensors measure light intensity in much the same way as solar sensors but are obviously only responsive to the UV wavelengths. In a sense, both 'solar' and UV types are sunshine sensors but measurements of Bright Sunshine usually relate only to visible light data.)

Traditionally, weather observers have logged daily totals of what is termed 'bright sunshine hours' (BS hours). This tem reflects the qualitative difference in brightness (and wellbeing) that we experience when the sun is shining brightly from a blue sky as contrasted with a dull, overcast day. Overall, a month with a  higher-than-average recorded total of BS Hours will obviously indicate a period with more frequent clear skies and less cloud cover than usual. Hence, the BS total is a useful extra parameter against which to interpret other weather statistics over e.g. a monthly period.

Measuring BS hours with any precision is not easy, by comparison with most other weather parameters. For air temperature or rainfall for example, provided you have a measuring instrument of accepted design and accuracy, there is little scope for uncertainty about the values it measures. But BS is different in kind from such parameters and its measurement is more tricky. While the contrast between blue sky conditions and heavy overcast is obvious enough, conditions that are neither of these two extremes are commonplace. Consider, for instance, the presence a layer of high cloud through which the outline of the sun can be more or less clearly seen, giving brightish conditions, and with shadows visible albeit maybe not sharply defined. Should this count towards the bright sunshine totals or not? It's easy to see why the measurement of BS totals is a relatively imprecise process.

In the past and before the widespread introduction of inexpensive electronic sensors, BS was measured by a fairly primitive instrument called a Campbell-Stokes recorder. This consists of a small glass sphere that focuses the sun's image on to a rectangle of sensitive paper. As the sun moves across the sky it marks a trace on the paper, but only when the sunshine is strong enough to create a sufficiently hot focus. This instrument actually provides a simple operational way to distinguish BS and non-BS conditions - if a trace is marked then that period classifies as BS and conversely no visible trace implicitly means that no BS was recorded for that period. This isn't of course an exact, scientifically well-defined distinction and it still requires an experienced user to set the paper correctly and, more importantly, to read the trace in a reliable manner. It's not easy, for example, on days with frequent short sunny intervals to locate the start and end of each individual trace consistently. And there are other imperfections with CS recorders such as the fact that sunshine is often too weak to burn a trace within a hour of sunrise or sunset even under conditions that would otherwise qualify as bright sunshine; traces also burn more readily in very dry compared to humid air.

Despite these limitations, for many years BS totals were almost by definition what was measured by a Campbell-Stokes recorder in the hands of an experienced user. But with the advent of affordable electronic solar sensors, life became more complicated. These sensors were a notable advance in allowing sunshine recording to move on from a simple yes/no of shining brightly or not (according to the definition of the CS recorder at least) to recording automatically and on a continuous scale, reading from zero up to a maximum (depending on latitude) of over 1kW/sqm. In an important sense, the detailed record of the electronic sensor provides all the information about the sunshine pattern throughout a day that anyone could reasonably want and the whole concept of bright sunshine hours becomes less relevant.

But what is missing and what is understandably precious to many dedicated weather observers (and indeed to the increasing number of scientists looking for evidence of changes in weather patterns over the years) is any comparability to existing historical sunshine records. And many of us privately also cling on to the idea of bright sunshine as a worthwhile parameter because it has an obvious and intuitive connection with how sunny the weather felt over a period, even given that BS is difficult to define scientifically with precision.

For professional weather observers, there was a fairly straightforward answer to the comparability problem. It turns out that measurements of the direct component of sunshine (ie as opposed to indirect or total) can be correlated quite well with traditional BS records and in fact the WMO have set a value of 120W/sqm as the direct intensity that qualifies as bright sunshine. So counting the number of hours in a day that the direct irradiance exceeds this threshold should give a decent estimate of the BS total. (NB Many amateurs make the mistake of assuming that this same threshold can be applied to measurements of total/global irradiance, e.g. as logged by Davis VP sensors. Doing this will lead to serious errors in any resulting BS totals, sufficient to render the estimates largely meaningless.)

But less well-endowed observers with no access to an instrument able to measure direct irradiance still have a problem. What's described here is a possible solution that can generate useful estimates of BS hours from measurements of global solar irradiance as would be logged by eg a Davis VP/VP2 system.

Rationale for estimating BS Hours from global intensity data

The basic premise behind the method described here is really very simple: If we assume that the direct component of total sunshine intensity is generally the dominant component - which is broadly true - then there should still be a marked drop in total intensity whenever the sun is obscured by cloud cover. Any fall in intensity clearly won't be as dramatic as that observed by a direct intensity sensor (where a drop of >90% may be seen) but should nonetheless provide a useable indicator of whether the sun is obscured or not. So by a careful analysis of actual light levels compared to the calculated maximum levels under different sky conditions we should be able to define an algorithm that can detect BS conditions with reasonable accuracy.

To understand the algorithm, it's important to remember that total intensity sensors such as the Davis VP/VP2 type effectively take readings in a horizontal plane on the ground, whereas instruments measuring direct irradiance do so normal to the sun's image. As a result, readings of direct irradiance with a clear sky don't vary hugely as the the sun moves across the sky from lower to high angles of elevation and down again. (In practice, there are some variations due to the different depths and clarity of atmosphere that the light is passing through at different sun angles, refraction effects etc, but these are secondary issues.) It's because direct irradiance changes to only a limited extent once the sun is above the horizon that a simple numerical threshold of 120W/sqm can be used to define BS conditions from direct irradiance data.

In contrast, because total intensity is measured in a horizontal plane there is a considerable change in intensity through the course of each day as a result of simple geometry. In fact, global sunshine intensity varies - to a first approximation - as the cosine of the sun's elevation angle if the sky is clear. This is a well-established principle and mathematically quite straightforward even if care is needed with some of the calculations. So if you can compute the sun's angle for a given date and time (the angle obviously varies with the season and time of day) you can then estimate the expected intensity if the sun is not obscured. The next step is to calculate the relative intensity, ie actual measured intensity as a percentage of the computed clear-sky maximum value at any measured time-point and, according to our proposition, submitting this relative intensity value to our algorithm should indicate whether or not we have BS conditions..

Of course, we can't just assume that this approach will work, but need to validate it. So we've analysed a set of data collected at a UK site over several months for which simultaneous sunshine measurements by both VP solar sensor and CS recorder were made. From this data we were able to (i) confirm the correlation between relative intensity level and BS hours; and (ii) calculate a percentage threshold value that maximises the correlation and generates the best estimates of BS hours from the VP data. The method is not perfect. Under some sky conditions it overestimates BS periods and undershoots under others. Therefore it is not especially accurate at categorising any individual intensity value as BS or non-BS unless the distinction is reasonably clear-cut, ie a relative intensity significantly above or below the threshold value. But averaged over several hours and days, its performance is good and can generally match the CS data for monthly data to within ±10%. A promising parallel is also being found with estimates of BS hours from the InstroMet sunshine logger. A more detailed analysis of performance will be included in the technical notes on another page.

Further information

There will in due course be a couple more pages on this topic: