Uncertainty in your GIS work

In response to this ’ we may assume Painted dogs (Lycaon Pictus) utilize certain vegetation type which may not be the case’ I would say that utilisation comes in different ways - do they use that vegetation type to hunt prey in? Do they use it as cover when sleeping out the hottest part of the day? To den in? To migrate throught (A safe corridor of sorts?)
Yes their spatial distribution may well be larger than shown in the data you collect but superimposing a vegetation or habitats map may show you overall patterns that will give clues as to what movement you may be missing out on tracking.
To try to cut down on spatial uncertianty without a GPS you can do spoor transects, set track traps.
You know generally in your area when denning time is, so if you can account for all packs in that time and maybe catch a glimpse of pups you can hedge a number on population growth.
In terms of population loss (snares, distemper, lion kills of dogs) they yes you will need more intensive monitoring.
We have monitored African Painted Dog for many years now. Please do get hold of megan@wildlifeact.com to ask more direct questions

As I’m new to this I don’t have any personal experience with data but I think I can understand the principles of these uncertainties, especially reading through some of the replies here.
I think there is always going to be some elements of uncertainty as we’re dealing with nature and there will almost always be variables beyond our control such as the weather, flooding, having to take readings at different times etc.
To reduce the sources of uncertainty we can try and take readings and samples on a more regular basis, increase the resolution of satellite images etc.
But external factors will always play some role and we just need to be aware of them and account for them when using the data.

The major uncertainty in my study is the temporal uncertainty. This is because the local population of my study subject, Great Egret are often supplemented by migratory population during the migratory season and yet, different population in tropic region tend to have seasonal movement in response to dry/wet season. I try to reduce the uncertainty by conducting my survey during non-migratory season to reduce uncertainty of counts cause by the migratory population but, the seasonal movement of some population might still generate some uncertainty in my study.

The coming source of uncertainty in my data collection could be temporal uncertainty. This may be due to the sometimes conduct of their thermoregulation at the same time of day (seasonal) for each lizard individual. As the weather can be unpredictable in the survey in different localities too. Furthermore, the weather pattern could affect the thermoregulation and behavior of the lizards. Therefore, as refer to @foggygoggle we may also be consider adding notes to indicate if it was hot or cold, sunny or cloudy, morning or evening, etc. As well as to minimize the source of uncertainty by trying to collect data at the same time of day, at the same time of the year, and with the same transect lines.

1 Like

From my own field data I can think of a few uncertainties. Inevitable ones include the fact that our GPS only has 3m accuracy, which is pretty good but of course, not perfect. There is also the fact that some of the data we record is the location of sighting an animal that is capable of locomotion — the location data connected to these animals is bound to be made inaccurate as the animal moves. Our team reduces uncertainty in the latter by increasing temporal resolution, taking measurements regularly over a span of time to get a more general idea of where these animals live. Of course, the spatial uncertainty remains, meaning that management procedures (etc) would likely have to cover a broader area than perhaps is necessary as a precaution.
If we keep taking measurements regularly for a longer span of time, we should slowly reduce uncertainty as to where these animals live, especially as seasonal changes occur.

I have seen many uncertainties in my field data in the form of Spatio - Temporal uncertainties due to insufficient spatial data covering the information about just the few location about the species distribution and also the data for a limited period of time causing the temporal uncertainties causing the biases in the final outcome. another uncertainties is due to the conversion of 3D actual bio-geographic data into 2D GIS mapping framework which we use to reduce through georefrencing,

When I think of my collected field data, the types of uncertainties that mat exist in them, if I get it right as I’m a new beginner to GIS, are : Location/Elevation uncertainty and Attribute uncertainty.
The Location/Elevation( of households & beehives) uncertainties are inevitable/always present because our field measurement instruments are imprecise and we are not careful when taking the measurements.
the Attribute ( the values observed at those Households’/beehives’ locations/elevations) uncertainties are deliberate simplifications(??).
The sources of uncertainties for Locations/Elevations may be reduced by more careful measurement and refined techniques employed from one field study to another.

My data are proxies for an undefined variable, soil ecosystem health, thus the uncertainty is great. However high uncertainty data is still more useful than no data. There is uncertainty around how closely the proxies are coupled to the true value spatially and temporally. They are deliberate simplifications, as they are far more economical than gathering more comprehensive data.
The uncertainties limit the generalizability of my results across space and time.
These uncertainties could be reduced by gathering more data, and gathering more different kinds of data, to explore consensus between multiple forms of measurement of soil ecosystem health.

I personally am new to the software and have no current project i am running that needs GIS, but i am learning a lot. Thank you all for sharing about your experiences, they are eye opening.

In my limited experience of mapping, and once again thinking of the learning goal I have decided for this course (i. e. how to create a map of the distribution of the sponge species I studied during my PhD using data that are available online), I would say that temporal uncertainty could be the main type for my data. In fact, my target sponge species are very sensitive to environmental variations and entire popolutations have been seen to disappear after intense weather events that affected the study locations. Unfortunately, this is a cause of uncertainty that cannot be avoided, at least in case of analyses carried on over an extended time period, and it can affect negatively the study of these sponge populations, as it happened already in the past.
I would expect thematic and spatial uncertainties to be less relevant for my target sponge species because they tend to not form patches and their growth form is massive, therefore it is easier to determine their edges compared to encrusting species.

In my research uncertainty could come from a lack of precision from certain instruments- for example our Garmin GPS unit does not have a very fine spatial resolution. Uncertainty could also come from inaccuracy in defining the dominant species range- it’s hard to characterize the extent of different types of seagrass beds in more estuarine water using senses other than touch (texture/size of blade) and visual identification (which can be imprecise because you are snorkeling around underwater in murky water)

Uncertainty in my research could come from errors in the GPS system or temporal uncertainty (if the movement patterns fit the entire population for the entire year and not just the few individuals sampled during the sampling period). Additionally, if the resolution of one map is fine scale, and you are overlaying it onto a course scale map, there may be some uncertainty about the accuracy of the overlap.

Using satellite imagery when looking at sea-level rise can cause uncertainty because the images won’t all be taken at MHWM. This makes it difficult to map sea-level rise if data is insufficient.

The source of uncertainty in my data is temporal and spatial. Conducting algae surveys of reef flat is heavily dependent on weather, tides, and staff capacity. Even when you try to plan everything out ahead of time, there will always be changes to the data collection schedules and we always miss surveying a portion of our study site or we have to redo a survey at a later time or date. We try to reduce these uncertainty by planning our surveys around the time when the weather is the most favorable, we have alternative plans in case a certain area is too rough for a survey and we can focus our resources to a calmer site until the weather permits. We also try to replicate the same surveys about two times out of the year.

Part of the challenges when working in a floodplain area would be spatial and temporal uncertainty. Most areas are seasonally flooded and inundated making camera trapping effort difficult. GPS collars are all subjected to locational errors including hand-held GPS devices. Even with drone mapping, image resolution may depend on the weather condition. Satellite imagery that are produced during a certain period of time become inaccurate for example with seasonal flooding making surrounding area very variable. Hence, part of minimizing uncertainty introduced mainly by simplification due to expensive cost of mapping a big area to high resolution is by extrapolating the habitat data. Simplification and generalisation of map resolutions make habitat analyses inaccurate. However, shaping the objectives of the whole project and realigning data collection schedule to spread out across the timeline helps deal with the variability. Another step used to reduce source of uncertainty would be to use drone imagery in mapping small areas which can be done to high resolution (3D).

Uncertainty may be caused due to a lack of precision on a landscape scale when a boundary of a research area needs to be drawn. I try to minimize this by zooming in as much as possible to have most details available so I can draw the area as exactly as possible.

Identifiying the base level from which heights are taken (the “Datum”) is always a tricky one and varies around the world. Here in the UK, Topography (ie. land heights) are based on the Ordnance Survey MHWM (taken from a tidal gauging station in Newlyn, Cornwall) but bathymetric depths (used on Marine Charts) are taken from Lowest Level of Astronomical Tides). When working on information to do with Sea Levels, you need to work out which Datum you are going to use (ie. relate sea levels to the normal “land” heights, or to depth above the ocean floor and the normal level used to calculate this - ie Astronomical tide level).
The added complication around the UK coast is that the tidal level varies so there is no one single conversion factor you can use to convert marine chart bathymetric/tidal data to OS topgraphic data.
This is then compounded around the world which will all use different datums for their topgraphic heights. - all very complicated !/

Due to wetlands being more than just a point in space, different people may measure from different sides/ends and have the same wetland at slightly different places. This would be classified as spatial uncertainty. They are not inevitable as each researcher can decide which point or side they take the coordinates from. It just usually adds more work and nobody really looks into it.

Temporal and spatial uncertainty would be the most common in the research that I am currently involved in. This would be due to the changes in population size, vegetation cover, weather patterns, as well as differences in vegetation growth within the study area which are examples of temporal and spatial uncertainty. Some of the uncertainties are inevitable as one cannot control that changes in weather or the rate of vegetation growth within an extensive study area. Taking measurements during similar times of the day, year, or weather conditions can help reduce uncertainty.

1 Like

Since I work with isotope stable values of Penguin blood, that integrate about two weeks of dietary, there’s a basal time uncertainty I have to deal with; added to the inevitable time gap between sampling events.
The spatial uncertanty could be due to GPS misalignments while tracking the Peguins foraging trips.
I can’t see a way out of this problems, other than always consider methodic error in the interpretation of results and stick to sample habits as clean and methodic as posible.
Since I don’t pretend creating a high resolution isoscape, I don’t think this uncertanty could deteriorate the resulting map.