i don’t understand this. i’ll assume you know the difference between the data and the visual presentation of the data. so then why even make this point? is this whole thread just some sort of backdoor way of saying that you don’t want data to be obscured at all?
he just means it used to be a handful of circles in the old days. now the maps contain a lot of annoying false points that hide the more useful info
Yes, exactly what cassi said. When iNat was small there were very few obscured points so they were more just an oddity. I am not, in any way, saying that data shouldn’t be obscured.
It just seems that you think we’re against obscuration (though translating obscured dots on GBIF isn’t the best idea), we’re just trying to make maps more useful, @charlie just mentioned an example showing current system is not ideal, I didn’t read it as a mutiny against obscuration.
Here’s another example map of an obscured species. You can see how big the rectangles would be as one is shown in blue and that there would be a lot of rectangles with only a few points in them. Looking at this, I’m not sure if rectangles would be better or worse. It may depend on the case. Clearly the rectangles would greatly improve/simplify things where there are lots of observations. Where there would only be one rectangle for a single observation, I’m less convinced. That said, I would still rather see a rectangle that indicates there is a point somewhere within that rectangle than a random point. I especially prefer a rectangle over lots of random points. It seems a toggle between rectangles and points would maybe be most useful as well as an option to remove obscured observations from the map.
hmmm… i don’t think that’s all he’s saying.
how do you explain these other statements from that same post?
the data aren’t “wrong”. people who aren’t looking carefully may misunderstand certain presentations of the data, but there’s a difference between misunderstanding of data and bad data.
What’s your goal in focusing on that and not the topic?
the data are wrong because an intentionally wrong location is added to the map. I understand that the data is more imprecise than inaccurate, except that precision data does not always transfer, so it ends up just being wrong. The round circles are intentionally bad data, and a rectangle without the circles would not be.
i think a more appropriate interpretation is just that it’s a simplification of the data, taking just the obscured coordinates without the obscuration box – just like all the non-obscured points represent just the center coordinates without drawing big circles around them to depict positional accuracy (where that is recorded).
there are only so many variables you can depict in a given visualization without making it overly complex.
the goal is just to understand the perceived problem. if the problem isn’t clearly defined, it’s not possible to reach a reasonable solution.
i think you have a different interpretation. I don’t see any point in further discussion about this though. You clearly like the intentionally mismapped points, that’s fine, but it isn’t really the point of this feature request anyway. I think everyone here is clear that you don’t support it, and chances are it won’t happen anyway given how overloaded the iNat staff are.
the noisiness is one thing he notes as a problem, but this also comes from the “description of need” section of the original post:
so that’s the part i was trying to clarify.
I think it was excessive to this particular topic, I guess we need to stick to original point of having massive rectangles filled with dots that obscure other observations, as you pointed out maybe rectangle idea isn’t the easiest one to acheive, so just having two different layers sounds preferable, both being active on default but with ability to turn off.
If someone has time, it would be nice to see how it would look to replace obscured observations with rectangles in a variety of situations, especially those where there are lots of obscured observations making things messy. I assume you can download obscured observations with their obscured coordinates. Someone could download observations for an all observations in an area for a many taxon map and also an obscured species for a species map and then create some comparison maps to better assess how it would look and if it achieves the goal.
Maybe this was mentioned (I didn’t read through all the comments already posted), but my one concern with the rectangles is that they might eventually give away a location (only when there are many obscured observations from the same location). An example might be observations from one’s own yard - I personally obscure all observations from my yard.
Currently the system displays a random circle within a defined rectangle, and that rectangle always falls on the real location. Create two obscured observations from the same location and you get two rectangles and two circles. The rectangles will partially overlap, as they must fall on the actual location. Repeat the process many many times and eventually you will have a point where all the rectangles overlap - the true location. With the current system if would be quite difficult to determine that true location (I suspect it can be done), but with the rectangles I imaging it would be quite obvious - look for that dark spot and you know the true location (or at least close)
I do like where this is heading though. I’ve often thought it would be nice to have a smaller obscured zone, maybe 5km by 5km instead of 20. Or the ability to select between “obscure - temporary” and “obscure - permanent”. Sometimes I obscure an observation, but for what ever reason I only care for it to be obscured for a short time (example, I spotted a bear during hunting season and didn’t want to share that location during the hunt, or I observed an uncommon bird on a nest and didn’t want it to be disturbed) - in either case I would be fine if those obs automatically changed to “Open” after say 6 months. Even today I was looking though my obscured obervations and changed a handful of Obscured obs to Open, exactly because I thought they just mess with the maps and because I no longer think they need to be obscured.
I could be wrong, but I think there is zero risk of finding a location based on the rectangle it is in even if there are a thousand observations from the same spot. I’ve always assumed the obscured location would be a random point within a fixed rectangle and that these rectangles do not and can never overlap. That’s the logical way to do it. So, every obscured point from the same location would always be in the exact same rectangle. Thus, using the rectangle would in no way give away a location. Here’s an example of observations I’ve made of cultivated plants at California Botanic Garden that I was keeping track of for my research, most of which are obscured as they are rare plants. The obscured points form a single rectangle.
So, if this were shown as a rectangle rather than a bunch of points, all you would see for all of the obscured observations would be a single rectangle and you would be able to see what is underneath, which you can’t at all in the zoomed out image. The only difference between using points vs. rectangles in this case is that, with rectangles, the map is much more legible and no one can mistake the huge number of obscured points as being in the correct location. Clicking on the obscured locations as you see them on this map to get more information would just be silly as there are so many and because their location within the rectangle is meaningless. If you could just click on the rectangle and have it pull up a query of all those obscured observations, you would be much better served.
There is a predetermined grid on the Earth that defined these rectangles. Any obscured observation within that square would get that exact same rectangle. It used to be a circle anywhere within 1 KM of the location and in that case it was possible to infer a location from a large circle of numerous obscured observations in the same place. But it doesn’t work that way any more for that reason
as kiermorse noted, the boxes in the obscuration grid are fixed. so there’s not a risk of overlapping boxes revealing any obscured locations. however, visualizing obscured locations in the context of their 0.2 deg long x 0.2 deg lat obscuration boxes can in various ways reveal clues to the true locations of the observations, as noted in https://forum.inaturalist.org/t/option-to-display-obscuration-rectangles-grid-on-explore-map/7646.
i think “silly” is debatable. there are pros and cons with each approach, and they vary depending on the situation.
also, although the obscured location within the obscuration box is somewhat meaningless, the obscured locations themselves are not meaningless when viewed without the context of the boxes, as you can think of those locations as being accurate within roughly 20km in most cases.
Here’s a quick example of what a mix of obscured and accurate observations of plants at California Botanic Garden would look like using rectangles instead of obscured points. On left is what you currently see. On the right is a possibility of what you would see if changed. The rectangle represents all obscured observations from the garden and you can see a few points at the garden where some observations are not obscured.
since the devil is in the details, i’m happy to see that you’re actually doing the hard work of trying to visualize / prototype. it looks like in your example, you’re displaying a ~50% opaque green box. so i’m assuming you’re trying to match the color of the box with the color of the pin/circle markers, and you’re not going to scale the opacity of the box based on the number of obscured observations. is that how you’re thinking about it?
how do you propose displaying the boxes in each of these cases?
- you have both plants (green markers) and birds (blue markers)
- you have 500 obscured observations in one box and 1 obscured observation in the neighboring box
- instead of the pins, the density grid style markers (the red squares that appear at lower zoom levels) are shown on the map
- you’re viewing the map at a really low zoom level (ex. you’re viewing things at a world level rather than a city level), where the density grid style red squares are actually larger than the obscuration boxes
also, right now, the maps will automatically zoom in to allow all the points on the map to be displayed. if you’re visualizing obscured observations as boxes, does that automatic zooming now have to take into account the entire area of the boxes, too?
and…since there are currently 2 types of obscured observations – obscured by the observer and obscured at a taxon (/palce) level – would you lump those all together, or would you visualize the two types independently?