[Talk-GB] Mapping of Kielder Forest(s)

[James Derrick]

On 29/08/2021 12:11, Russ Garrett via Talk-GB wrote:
> This large area of forest has been meticulously mapped as many
> hundreds of small polygons, mostly tagged landuse=forest. The polygons
> are split at narrow breaks in the trees as seen in either Bing imagery
> or OS StreetView.

As one of the mappers who have spent many hours on the Kielder area, I 
appreciate that you've taken the time to ask the list before embarking 
on wholesale change.

Originally many of the forest areas were imported from OS Opendata (I 
think), in rectangular blocks. This could be seen where a large 
continuous plantation was truncated on an arbitary line, likely an 
import selection area.

As part of a much wider personal project (14 years and counting...) to 
map Northumberland in field-by-field details, these areas were improved 
manually by chopping into individual blocks, and adding the fire breaks 
/ forest haul roads as gaps in the continuous areas, some as separate 
areas marked as meadow/ scrub/ heath/ etc.

This detail was added to give information to walkers and riders (horse/ 
pony/ fit mountain bikers) where access may be easier - walking along a 
forestry haul road is easy; across the ridges created by a commercial 
planting machine is very hard.

> ## The problems with this approach
> 1) The use of name tags here is not ideal - it violates the "one
> feature, one element" principle. From a rendering perspective this is
> also bad: at higher zooms, there is an unavoidable mess of identical
> name labels, and at low zooms a label is missing when it should
> clearly be shown - these are big forests and well deserving of a label
> at low zooms.

I can see that adding a master relation to bring individual stands of 
landuse=forest trees together could be useful, however where do you 
start and stop?

The Forestry Commission signage on the ground suggests they manage the 
wider landscape with areas / plantations / species collectively known by 
several names, not just Kielder.

Do you have access to this level of hierarchical data (I don't even with 
ground survey, beyond specific areas - e.g. Hawkhirst, Bakefin) or are 
you suggesting one level of "Kielder"?

> 3) Kielder is a working commercial forest, with a significant area of
> trees being felled and replanted on a yearly basis. There is no
> guarantee that the forest will keep the same gaps after replanting,
> and so constant tedious fiddling with the outline polygons will be
> required to keep the map accurate.

That is indeed an issue - stands are regularly clear felled, and as the 
dam was replanted many years ago, many areas are ready for cropping at 
similar times.

My general approach is to map from imagery, but do recognise from my own 
walks on the ground that areas may be clear felled and appear barren. 
These typically are replanted relatively quickly (in the 10-20 year 
lifetime of a softwood tree) and rarely change use (e.g. continued 
forest, not meadow) due to the physical geography (terrain, altitude, 
thin rocky soils).

By breaking the overall forest into individual stands of landuse=forest, 
it seems relatively simple to change the status of one area as it changes.

I'm not aware of a tag lifecycle of forestry, but have seen newly 
planted forest areas tagged as natural=scrub, before turning into 
landuse=forest as the saplings grow from the corduroy lines of a 
planting machine. Not sure what a clear-felled area could be modelled as.


> These landuse polygons should be merged into much larger polygons,
> which encompass smaller breaks in the trees. These larger polygons can
> then be combined into a multipolygon which represents the entire
> forest.

Are you really suggesting combining these areas into one large block, 
adding haul roads as highway=track and calling it Kielder?

If so, such an extraordinarily blunt approach would appear to throw away 
detail, and with it many hundreds of hours of effort. :-(

The stands of trees are not linear blocks with equidistant uniform haul 
roads - the physical geography is much more complex, hence the complex 
areas of landuse=forest. There are old farmsteads, bothys, stone walls, 
shedding rings, gullys, all which punctuate the land.

This is the rich detail I've tried to add - imperfectly, but still with 
care.


> I appreciate that a lot of work has gone into mapping this area and it
> is obviously a shame to have to revisit it all, but I think these
> changes are important to allow the map to be easily maintained in
> future, and to give these forests the rendering prominence they
> deserve.

I'd suggest respecting the hundreds of hours that have added detail 
useful for walkers and statistical analysis.


> Comments are welcome, and I won't do any further mapping in this area
> until people have had a week or so to comment.

Thanks - as your idea as presented seems a very bad one!


James
-- 
James Derrick
     lists at jamesderrick.org, Cramlington, England
     I wouldn't be a volunteer if you paid me...
     https://www.openstreetmap.org/user/James%20Derrick