[talk-au] Tracks flagged as missing from government data

[Andrew Harvey]

On Wed, 18 Aug 2021, at 4:40 PM, Little Maps wrote:
> Andrew, the 1:25,000 Vic gov topo mps show tracks/driveways on private 
> properties in a different colour to those on public land and the map 
> legend clearly distinguish the two. So hopefully there is a 
> public/private field in the dataset that can be used to distinguish the 
> two. Cheers Ian

There's nothing I can see in the Vicmap Transport dataset to distinguish this, if you know of another dataset that would help I'm keen to have a look. I tried looking at the crown land data from Vicmap, and while this would certainly help, it misses some areas where public forestry tracks occur.

Another option would be to use existing OSM mapped features for parks and reserves, which could be used to prioritise those on public land, I'll touch on the private land tracks below...

On Wed, 18 Aug 2021, at 6:11 PM, Little Maps wrote:
> Apologies for repeated posts on this issue,

Oh no need to apologise, the more eyes on this the better.

> but a data dump of 250,000 
> ways is worth some discussion I believe…

Agreed.

> Andrew, if it is not possible 
> to separate public and private roads using tags in the Vic gov 
> database, can the Vic Tracks MapRoulette Challenge please be pulled 
> down immediately? If you scan around the state, it’s obvious that the 
> great majority of the “unmapped tracks” are on private property and not 
> on public land. Even in the forested highlands of Gippsland, there are 
> far more “unmapped tracks” on private property around the margins of 
> the forests, and in the surrounding farmlands than in the forest itself.
> 
> With no better data available, it seems reasonable to suggest that this 
> MapRoulette Challenge includes 100,000-150,000 roads/tracks/driveways 
> on private land (maybe more), with no indication of that fact to inform 
> well intentioned mappers, and no suggested tagging to indicate 
> access=private. Many of the private roads are indeed short driveways 
> that have no through connectivity, but many are longer and create 
> through ways. In private forestry plantations in W Vic for example, all 
> of the private internal roads are included in the challenge, which 
> creates a wide grid of new “public” roads. I’ve only looked at the Vic 
> challenge so far and have no feedback yet on the challenges in other 
> states. 

highway=track are documented as forestry, agricultural or fire trails, so shouldn't be considered as public roads by data consumers.

Tracks on private land are useful too, for deliveries, fire management and emergency access.

Whenever I'm out surveying I always try to set foot, bicycle and motor_vehicle access tags on highway=track since I don't think you can safely assume any particular access "defaults" apply unlike other highway classifications where there are reasonable access defaults.

So in my view there is no harm in mapping tracks we know exist but don't yet know access restrictions, in fact I think there is a lot of benefit to getting them in and then later mappers can follow up with setting access tags once they are known.

> A couple of months ago, Microsoft’s mapping team was told to cease and 
> desist after they mapped a few 1000 private roads without indicating 
> private access (they responded to that request admirably). This 
> challenge dwarfs that issue 100-fold. I’m confident that the intentions 
> were good but this implementation is fundamentally flawed. The fact 
> that the data dump and challenge were sponsored and paid for by a 
> government department with no notification or discussion from the 
> Australian mapping community until after the fact makes the issue even 
> more problematic in my mind.

I saw the discussion in brief, but my recollection was that involved tagging some roads as unclassified or residential where data consumers do assume by default they are public access.

OSM has always been someone does a bit of mapping and others build upon it, rarely are things mapped perfectly from the first upload, it only becomes a problem when some tags data consumers have assumed defaults.

>From my side I do consider this a soft launch, by uploading and publishing the challenges on MapRoulette it provides a good medium to distribute the results to the mapping community to look over. So I am very open to further feedback here which I can use to rebuild the challenges.

So far there is no organised mapping happening to work through these challenges.

On Wed, 18 Aug 2021, at 6:25 PM, Warin wrote:
> For NSW the tracks can be compared with the DCS Base Map so as to confirm there existence and possible level of highway = track/unclassified etc.

I though it was the same data as on the Base Map, but I never checked that. Regardless I don't hold much trust in the DCS Base Map and only use it as an additional data point where all else fails, so still it's a good idea to check on imagery. Some tracks become overgrown to the point there are now highway=path and then sometimes disappear completely but still remain in these government datasets.

> For WA if you think the 'name' is descriptive then change 'name=*' to 'description=*' ... this way the data is not lost. 

Good idea. It takes a bit of comute to process these, but I'll make the change and reupload.

On Wed, 18 Aug 2021, at 6:26 PM, Little Maps wrote:
> And one final, short post… the 122,886 “unmapped tracks” in the SA 
> tracks map roulette challenge are also predominantly private roads on 
> private land, especially in farming areas. Again, the challenge wrongly 
> assumes that access is public not private. 

I had assumed they were on on public land, but regardless I don't think this changes anything, if they still are visible on imagery and confirmed through this data, then the tracks do likely exist. Do you suggest that maybe tracks on private land shouldn't be mapped at all or are you suggesting that they shouldn't be mapped without any access tags set? As I previously stated I think we are better off mapping tracks on private land without access tags set until confirmed, but keen to see this discussed.