I’d to apologise for the long wait. Dirt roads (usually with shades of ground, be it brownish, yellowish) are sometimes very obvious from imagery, than, let’s say, an asphalted highway - usually grey in colour. Side note: rains will wash away any dust accumulation on asphalted surfaces.
Today’s post will be focussed more on highways criss-crossing the unending rice fields specifically. Since I’m from the state of Kedah, my statements would reflect a mapper’s bias from the socio-geographical area of where I live. It may not reflect or translate well to be modelled in other states of Malaysia.
First of all, previous remote mapping team (GlobalLogic for Grab) wanted to map highways which not yet mapped, with deadlines being set. This is unfortunate: some highways were mapped higher in hierarchy. On top of that, since the number of local mappers are really small, these incorrectly mapped highways cannot be amended as early as possible. It’s been a year, plus; and still I managed to discover incorrectly mapped highways. Leaving changeset discussions may not bear fruit - the related mappers have been inactive and already stopped mapping altogether.
The domino effect is obvious: the AI import team on behalf of Facebook and others, tend to follow these incorrect mapping practice.
It’s not always the case where highway=residential would follow a highway=unclassified in parallel, from end-to-end. Fair, sometimes dirt roads do lead to residentials. But do we really have to map a residential highway from one extreme end to another extreme end?
Not necessarily. See above example. Sometimes houses can be found in “island”-like plots. The highway on the left of the canal should be split to indicate limits accordingly. There are even access roads direct to some houses from an asphalted highway=unclassified; which I would map as highway=residential ways instead. Only machineries (think of harvesters, tractors and the likes) would rather make use the remaining stretch of these dirt roads, than, say, even underbone motorcyclists. And, why not, include the humble hobby cyclists as well.
And of course, there will be real-world cases that might not fit these textbook cases. Feel free to bring them up here, and we’ll discuss.
I would like to take this opportunity to beg, hopefully, that the team would carefully check edit histories, before making edits solely based on imagery, without any additional cues. I don’t fancy seeing my ground truth efforts went direct into the bin unless there are strong proof showing otherwise (I’m happy to be proved incorrect, with proper evidences if available). I reckon other active, local mappers would share the same sentiment too.
