I love maps! It’s so easy to get lost in a good map! Maps play a central role in answering those fundamental questions – the ‘what’, ‘where’, ‘how’ and ‘when’ of ecosystem restoration. (I think we all know the ‘why’ for seagrass restoration or you wouldn’t be reading this blog…).
We all rely heavily on accurate and up to date maps to better understand the coastal seascapes where we work, and to plan specific restoration actions e.g. ‘what’ species of seagrass to plant and ‘where’ to plant it. For example, at the scale of one of our typical restoration plots (1000 m2 – 10,000m2) ‘remotely sensed’ maps are created by combining numerous drone photos into an orthomosaic. This helps us to measure the impact of management measures (i.e. the restoration successes and the failures). We can analyse these drone images and orthomosaics to monitor changes in seagrass cover showing the impact of our efforts.
But how do we make strategic decisions on where to prioritise seagrass restoration at the European scale? Well until now we haven’t really had a good map of European seagrasses and so that process has been quite a challenge! Enter the incredible team at Nantes Université who have spearheaded the work with a timely new publication – “An Initial Map of European Intertidal Seagrass”.However, I want to stress a key point – this initial map is called initial for a reason. Whilst it is the culmination of a LOT of hard work over many years (a truly phenomenal effort), it is also just the beginning of a great deal more work – the journey of no doubt numerous iterative improvements towards Version 2.0! You will have all heard the aphorism “Don’t let perfect be the enemy of good” or perhaps “All models are wrong, but some are useful.” Well, that’s certainly true for this map! At some point you just must draw the line and publish, and with the Regulation on Nature Restoration (Nature Restoration Law) coming into effect on Sunday 18th August last year it has meant that there really has been no time to waste in getting this map into the public domain.
It’s a tough call to make to decide when is ‘good enough’. Particularly because academics we often chase perfection! But the pursuit of perfection can sometimes prevent (or at the very least delay) the implementation of improvements, and we feel that is particularly true for this map because we believe we can expedite significant improvements to the final product by crowdsourcing knowledge from the wider seagrass community and the public at large!
To dive a little deeper, the challenge of creating maps at the ‘European scale’ is precisely one of scale. Whilst this Initial Map of European Intertidal Seagrass might be over 90% accurate (great!), the 10% that isn’t accurate still represents almost 1,400km2 of the European intertidal zone!
And what’s more, when it comes to using the map, the first thing you are likely to do when you open the map is to look at your local meadow (like I did), and it might be that it isn’t quite right (like mine wasn’t). I can therefore really appreciate that it can be quite jarring (When you know an area intimately, and when there are clear issues with that area, your impression will immediataly be a negative one!)
But this is exactly the point of publishing this map now. This is how we move forward together, and improve the final product from 91% accuracy to 92% accuracy, to 95% accuracy and beyond… Now that its published, the next step is crowdsourcing input into the map (You can open it here). By reaching out to everyones local knowledge we can go on to improve the map, and refine it so it better reflects seagrass distribution at the local scale, as well as the European.
In truth this interative improvement process already began during the process of creating this ‘initial’ map, the ‘final map’ has already been through a process of numerous interative improvements. Each time the final map was produced discussions with seagrass colleagues (myself included) identified seagrass meadows on the map where the final map which we knew weren’t quite right. This interative process was essential in improving the map up to this point, because to improve the end product we needed to know what was, and what was not seagrass! The take home message here is that the accuracy of remotely sensed outputs is improved enormously when it’s based on data that has been actually measured in the field, in fact field data is a fundamental part of the process. The more field data we get, the more we can improve the final product.
What’s more, the benefit of going through all this extra ground-truthing has shown us that the vast majority of issues with the creation of this map have come from three clear ‘problem’ areas:
The first problem is with what is known as ‘incorrect intertidal masking’ (in GIS ‘masking’ refers to the means of identifying areas of a satellite image to be included in an analysis). In this case the map was both missing seagrass areas, and including wrong areas (such as saltmarsh).
The second problem was with ‘salt and pepper pixels’ (in GIS ‘salt and pepper pixels’ are pixels whose intensity values are abnormally higher (salt) or lower (pepper) than those immediate surrounding them. For our seagrass map these were often isolated pixels in non-seagrass areas, but when salt and pepper pixels’ occur its often the case that the few pixels that are affected are very intensely corrupted (i.e. to white or black)And finally, the third problem was the misidentification of green algae or microphytobenthos as seagrass. This problem essentially comes down to not being able to distinguish ‘seagrass green’ from ‘algae green’ on sateliite images, or in some cases it occurs where the algae is sitting physically on top of the seagrass when the satellite image was taken (at low tide 3D habitats essentially become 2D for remote sensing purposes).
So what can we do about it?
Well the good news is that the first problem can be easily remedied both by local users, and by outreach to local knowledge. By co-creating a more accurate intertidal mask (i.e. to remove those saltmarshes – this is recommended in the paper) we can further improve the accuracy of this the map. But we need your local knowledge!
Additionally, those salt and pepper effects (except in those cases with colossal sensor errors) are usually pretty easily removed by applying a ‘nearest neighbour filter’ whuch works to remote those isolated corupting pixels from the map.
However, the misidentification(i.e. the what is seagrass green and what is alage green problem) isa mapping issue, but the rate of misidentification that we provided in the paper is deliberately a little conservative to try to ensure we don’t overestimate total seagrass cover.
So whats next?
Well looking ahead to Version 2.0 of ICE CREAMS it will likely include more and more small tweaks to the model, but we belive the largest improvements will come from image selection (i.e. intertidal masking properly, the use of cloud free images, selecting images for times of low tide etc) and also by combining ICE CREAMS with the outputs with other models and that better describe the wider ecological elements (i.e. using another model to distinguish species within the pixels we predict as seagrass).
So here’s our CALL TO ACTION, please let us know how the model works for your area. We want to hear from you! Together we can improve this this Initial Map of European Intertidal Seagrass but only if we combine the technlogy with local seagrass knowledge of from across our European community, and indeed the wider public!
RJ
