As a beginner, Cloud cover, which affects the image’s visibility, is a source of uncertainty in satellite imaging, such as landsat images.
The estate i work on was given a habitat cover based on a person and AI interpretation of the satellite imagery. The person sitting at the computer in a different country had enough understanding our climate to know AI was wrong to classify an area as crop, as we can’t grow alot of crops in the north but they did not know enough local ecology to understand the cover was from Bracken. This plant grows rapidly in the summer to look like a green forest and then dies off in august to cover the area in a brown matt. The area that pant covers can change rapidly in a mast seed year.
If this information wasn’t check with locals on the ground and ground truthed, it would lead to mis information and unreliable analysis. Spatial data coming from satellite needs to be taken in the context of the time of year and with local understand of what is happening on the ground.
I think uncertainty (temporal, spatial and thematic) has a significant hand in many environmental related studies. I’m making a very generalized statement here, but especially considering the rate of climate change.
Also I remember coming across a reply about an AI error, given the whole 4IR, maybe how can problems like misidentifying categories be fixed
In satellite imagery, there could be great uncertainty because many ecosystems are changing rapidly particularly at the border with human interactions. So for example a satellite image or GIS map might show a mixed forest/grassland area, but even 3 or 6 months later these images might be wholly inaccurate because now that area has been converted for agricultural use. This would be an example of temporal uncertainty, I believe.
1 Like
Uncertainty must always be considered in our analysis because when we are collecting data, sometimes we are selective and we lose sight of important data, that can represent some discovered. It is important we keep in mind all situations that may present.
In my study, i used satellite imagery for to do comparisions from the antropogenic activities and their enviromental impact over time, in this way is possible distinguish all the changes. if you compare an image from 10 years ago with a current image, the impact is evident. Due to the above, i don´t think that satellite imagery may have uncertainty because uncertainty rest with the researcher.
I’m not sure there is anything folks can do to reduce sources of uncertainty entirely. When I collect annual presence/absence data for a particular species, I am collecting data points in time and space that are subject to temporal uncertainties. (Probably more than I could even begin to point out…)
How do we negate or minimize uncertainties to guide management decisions? Do we do the best we can with what we have??
The bird survey I will be carrying out will have spatial uncertainty due to the nature of the environment, security challenges and habitat loss we now face. It will be a seasonal survey (Dry Season and Rainy Season) so temporal uncertainty will be taken to account.
To reduce these uncertainties, the research work will be carried out in areas that are safe and thoroughly accessible. The region is pretty much the same (Savanna) so thematic won’t account for much.
In my work there is thematic uncertainty since at times is difficult to clearly define landscape categories, and also temporal uncertainty because surveys are conducted only once. These generalisations are mostly due to time and resources constraints, and might affect the accuracy of the areas defined for conservation and other uses. We could reduce uncertainty by making use of current satelite data and also by monitoring closely any changes in the landscape.
In collecting data on vegetation cover, we could experience temporal uncertainty due variation in land cover as seasons change. To take an accurate management decision, images can be taken at various seasons.
When we want to determine vegetative units from satellite imagery, there are inevitable uncertainty caused by the inability to accurately measure the boundaries. In this case the boundaries on the map produced do not correspond exactly with those on the field. To reduce these uncertainty, you have to go on the field several times to correct.
I have experienced uncertainty to define border to specific features on map. like forest cover. and it can get subjective to accurately demark it. this will effect the precision of the map I am trying to make. I cant think of any solution for it.
Uncertainty in one aspect of my work is around the structure of the forest where the Critically Endangered mangrove finch is found. Due to the now very restricted area where this species is now found, totally just 30ha, very fine scale differences in the forest structure are useful to see in comparison to the radio tracking data we have collected, especially since in recent years we have detected dieback in the mangroves at this site. It is hard to get this fine scale information and compare to the past (looking for changes in the forest over time) from existing aerial image files due to the lower resolution files of earlier photos and at times cloud cover restricting the area we are interested in. This limits our ability to review accurate change over time in the ramaining habitat of this Critically Endangered species and therefore impacts on our decision making for conservation management.
And combining different sources! Nowadays, there are multiple satellites where you can access the data (some are free, but others are for sale) and then mix the information to get a better spatial and temporal precision 
1 Like
Do you think it is relevant to the amount of uncertainty given by the lack of night surveys? I mean, if your species or study object it’s mostly diurnal (as I think dolphins are) I don’t think the time between day and night could cause a lot of problems in your analysis, right?
For the record I’m not a dolphin expert 
so maybe I’m wrong about their behavior, it’s just an idea
And I have a question, why do you think adding acoustic records it’s going to cause spatial uncertainty?
There is lots of uncertainty associated with my data - species distribution data and climate data. The species data is taken as a large-scale national survey, and are not associate with fine-scale location data. The country is divided into many grids and all observations taken within that grid are normalised to its’ centre- meaning two observations in totally different habitats is centralised to the same place, and therefore are not always reflective of key information like habitat preference, number of individuals, time of day, etc. Climate data comes with its own, numerous uncertainties - such data is the result of modelling at different scales, and often extrapolated away from weather stations. It is also often averaged over months, seasons, or years, and so often has temporal uncertainty. When making predictive models, it is important for me not to lose sight of these many uncertainties inherent to the data.
Imagine there are often not very clear answers to such a key question, haha. We have to maintain always in our minds the uncertainty in the data, especially when making interpretations and inferences from our models. Knowing clearly what you do and don’t know, and how confident you can be about your data both pre- and post-analysis helps to draw the boundaries of your knowledge (i.e., what conclusions can we draw, and which can we not?). Much easier said than done, of course!
As I work from South America, we have all types of uncertainties, in some places more than others. But in general, it’s a common situation. For example, because of the weather, some places are too cloudy, so no matter when the satellite takes images is not going to catch good resolution. So in some cases, this type of uncertainty is inevitable. In this case, if possible, it’s necessary to take additional data from exploratory trips to the specific area, using GPS, and as well when there are available drones.
Thank you for your advised. I will try with my best. Thank you.
Earlier, I had no clue about uncertainties. This is for the first time I have come across this term. After reading this I feel uncertainties are very common when dealing with geospatial data. As natural parameters are not in our control. In our study, I think we have temporal as well as spatial uncertainty. It becomes difficult to mark a shapefile for a given protected area because the vegetation changes over time.
I will require more help on how to figure out uncertainties and how to deal with them.
Thanks & Regards,
Rutuja
In my field data exist uncertain because of the lack of precision of the devices that are used for the geolocalization of the point of the quadrants. This kind of uncertainty is inevitable because depends on the signal of the satellites and the accuracy of the GPSs and it is a generalizing of the decimals numbers of the coordinates. That could affect the representation of the data if two or more quadrants are very close to each other and can be reduced using more accurate devices.