Seems to work well, and pretty accurately, for me.

A brilliant tool, made better it seems.

For me, its value has been in quantifying the small but still significantly useable radiation during the deep-winter months. At those seasons, the sun is low in the sky and obstruction by surrounding buildings, trees and distant hilltops is critical.

PVGIS allows the observed local horizon (as viewed from any given point of interest) to be uploaded and incorporated in the results that come back as download. You upload a string of numbers as a txt file: these numbers represent measured angle of elevation of the horizon at given compass bearing. The system assumes they are equally spaced around the full 360o, so If you upload 36 numbers it decrees that they are spaced at 10o intervals. With a jagged horizon you need them closer than that, so upload a string of 120 numbers, giving 3o intervals. But you don't need to measure elevation at 120 points - just the SE to SW sector in deep winter, so 40 points measured, the rest uploaded as zeros or whatever you like - the sun doesn't 'see' the rest in deep winter so doesn't matter.

How to measure the horizon's angle of elevation at 3o intervals (the first number is taken as due East)? Obviously a theodolite will do it, and you have to relate it to true solar North, not map North or magnetic (compass) North.

I've experimented with taking photos with the camera exactly level. The print can be overlaid with a transparent sheet gridded in degrees upward and sideward from the print's exact centre point, which you've created e.g. by photographing a brick wall head-on, measuring the brick module and distance from wall, converting that by trigonometry into degrees. Then it's a problem of hoisting the camera aloft to perhaps a roof position (point of interest) that doesn't exist yet. Drones obviously, but haven't tried that yet.

Anyone got any bright ideas, remote hardware etc, to elegantly acheive this? My efforts so far v Heath Robinson!

Thanks ST. That one gives the longitudinal profile along a path - maybe there's a tool for the circular horizon profile from a fixed point.

Unfortunately map data, let along Google Earth, is no good for this purpose. It really has to be measured direct from the actual point of interest. It's completely sensitive to tiny local variations, like a building or a tree, tho I suppose map data/Google Earth could be reliable for a more distant land-formed horizon - but even then, whether it's bare land or tall-wooded makes all the difference.

Just moving the point of interest a couple of metres can make a massive distance to the patch of clear sky it 'sees', where that is (as usual) defined by nearby buildings/trees.

I bet I'm more Heath Robinson than you!
I really meant doing it photographically.

Posted By: SteamyTea

Posted By: skyewrightI prepared our custom horizon using a simple home made clinometer

Did it significantly improve PVGIS's estimate?

At a daily level it much better replicated when we 'loose' the sun in the early evening behind the ground that rises steeply to our NW, However, because of the time of day that happens it didn't make a big difference to the annual figures.
Since our horizon is less than 5 degrees from E to W we probably aren't a tough test case...

Posted By: SteamyTeaDid it significantly improve PVGIS's estimate?

It wouldn't 'improve' PVGIS's figures, only reduce them for when the sun is blocked. As skyewright says, would make v little difference to annual figures, which are totally dominated by when the sun is high in the sky.

But for deep winter, making use of what little there is (plenty, with a low southern horizon and suitable 10-day or so storage, for a not-quite-PH's space heating demand - or a 'done the best we can' refurb), all at low elevation, that shading is critical, maybe affecting the siting of collectors, if not ruling it out.

Even then any shading below 5-7o elevation doesn't matter because there's so very little power in the sun at such low elevation (in any season). Above that, well worth it.

Had a go with this. The result is way off in my case. My average over the last 6 years was approx 3100 kWh. PVGIS calculates 4030.

And an added wrinkle to the shading issue -
if there's partial shading to a PV array, like a small patch of shade from a chimney or tree that moves across the array over the hours, the whole array's output is degraded to match that worst-case patch. I don't think PVGIS attempts to account for that kind of extremely localised shading, whose effect is due to inverter technology. The effect can be minimised by a less-common inverter technology, where each individual panel has its own micro-inverter.