Filling the gaps

Posted on April 18, 2024 by lembrechtsjonas

Anyone working with microclimate data is familiar with time series data – repeated measurements over time at the same location.

And anyone working with time series has bumped into an important potential issue with them: gaps. More often than not, time series are incomplete. There could be erroneous measurements, sensor malfunctioning, sensor replacement, data transfer issues, memory issues and so on.

Over time, a whole toolbox of techniques has emerged to fill those gaps and make those time-series whole again. In a recent paper, we tested a series of these gap-filling methodologies for their accuracy. That question is important especially for microclimate networks, as here not only the temporal but also the spatial relationship between time series is playing a role, and filling gaps is thus not a trivial exercise.

In this paper, we applied and evaluated 12 such gap-filling methods to complete the missing values in a dataset originating from large-scale environmental monitoring. For this, we used the unique dataset of 4400 IoT-connected microclimate sensors that were deployed across Flanders as part of ‘CurieuzeNeuzen in de Tuin’, our large-scale citizen science project on heat and drought.

(a) The TMS-NB microclimate sensor was used in a large-scale citizen science project on microclimate monitoring. The sensor measures temperature at three heights, as well as soil moisture. Data transmission occurred via NB-IoT. (b) The WSN covered 4400 gardens across Flanders. Sensor locations are colored based on whether time series were complete (green) or had missing records (red).

Methods evaluated included Spline Interpolation, MissForest, MICE, MCMC, M-RNN, BRITS, and others, and the performance of these imputation methods was evaluated for different proportions of missing data (ranging from 10% to 50%), as well as a realistic missing value scenario.

Accuraccy estimates (Root Mean Square Error and Mean Absolute Error) for the twelve tested imputation methods

Interestingly, techniques leveraging the spatial features of the data (such as MC, MCMC and MissForest in the graph above) tended to outperform the time-based methods. Importantly, as well, real scenarios of missing values – with gaps often occurring in larger blocks – often resulted in a lower performance of the models than artificial scenarios with randomly missing points, especially for more traditional techniques such as MICE.

Of course, this result is not the final conclusion on the debate which gap-filling technique to use. The outcome strongly depends on the specifics of the datasets at hand, in our case a dataset of microclimate data of fairly short duration (only little seasonality), with relatively sparse temporal resolution (every 15 minutes) and unusually high spatial density (4400 sensors across Flanders). These features work in favour of techniques that take the spatial features of the data into account, and reduce the applicability of e.g., deep-learning techniques that might prove more robust for more complex time series with longer temporal window and higher temporal resolution.

So, let’s hope this exercise in gap-filling can help other microclimate enthusiasts in their search for good solutions!

 Source: Decorte et al. (2024) Missing value imputation of wireless sensor data for environmental monitoring. Sensors.

Posted in Belgium | Tagged , , , , , , | Leave a comment

A wealth of wildflowers

Posted on April 13, 2024 by lembrechtsjonas

One could wonder if the microclimate-based models of Haesen et al. would have predicted such a wealth of wildflowers in the garden of our new home!

Wild garlic

This is our first spring in our new garden, and we could thus have hardly foreseen that so many beautiful decorations would spring up in our little patch of forest!

Lily of the valley

Of course, that is if our distribution models would not have said so. In this recent paper in Ecology Letters, we used our high resolution (5 x 5 m) maps of forest microclimate to improve distribution models of forest understory plants.

5 x5 m might just be enough to give our garden its own little pixel on the maps, so it could be worthy to investigate…

Of course, I should not tell you that all that greenery was one of the reasons why we bought the place, so I am beyond excited that the garden is rewarding our decision so handsomely!

Bluebell
Wild garlic
Posted in Belgium, General | Tagged , , , , , | Leave a comment

A drastic shift in mycorrhizal communities

Posted on April 10, 2024 by lembrechtsjonas

One of the key findings of the Mountain Invasion Research Network (MIREN) is that mountain roads have a MASSIVE impact on their surrounding vegetation. Whole communities are changing, non-native species are moving in, and all kinds of species are moving hundreds of meters up and down along them.

Now (in this paper), we add a new and intriguing component to that story: belowground communities are changing fundamentally as well, and these changes are surprisingly consistent across the globe. In a first attempt to include the dynamics of belowground communities into our global MIREN storyline, we looked at the dominant mycorrhizal associations of plant communities along mountain roads.

We estimated the percentage of plant species in each of our MIREN plots from mountain regions across the globe that was associated with a certain mycorrhizal type, and analyzed trends in the dominance of each of these types along climatic gradients, contrasting disturbed roadsides with (semi-)natural interior vegetation.

Studied mountain regions and their mycorrhizal associations

Conclusions were surprisingly clear: while changes in vegetation are often relatively murky and contradictory between regions, trends in the dominant mycorrhizal type were relatively straightforward: roadsides overwhelmingly facilitate the establishment of communities associated with arbuscular mycorrhizal fungi (AMF), to the detriment of ericoid and ectomycorrhizal species. Even more, that beneficial effect on arbuscular mycorrhizae was especially strong in regions traditionally dominated by ecto- and ericoid mycorrhizal communities – often those with colder climates, such as our own northern Scandinavian research site.

A complex figure, perhaps, but the key story of the paper: the lower the proportion of AM associated plants in the adjacent vegetation (x-axis), the stronger the increase in AM-associated plants in the roadside (blue line above the 1/1-line). This pattern was especially strong in Norway, a region traditionally relatively poor in AM-associated species.

This ‘power to the AMF’ we see along mountain roads is very much in line with the known traits and preferences of these fungi and their associated plants. Indeed, roads tend to favour ruderal species, commonly AM or non-mycorrhizal (NM) plants, to the detriment of more perennial and woody plants, more likely to be associated with EcM and ErM fungi. We thus expected road disturbance to correlate with increased AM and NM representation and decreased EcM and ErM, yet were pleasantly surprised with the generality of that pattern.

MIREN surveys cover an impressively wide variety of mountain roads, yet observed patterns were surprisingly consistent among them. a) High elevation gravel roads in the Argentinian Andes b) Lowland gravel service road in the Norwegian Scandes c) High elevation asphalt road in Yellowstone, USA d) Lowland asphalt road in Chile with strongly invaded roadside vegetation e) High elevation gravel road in Tenerife f) Lowland snowed-in gravel road in the Norwegian Scandes.

A possible hypothesis behind our – for now unfortunately only observational – correlation would be that roadside disturbance leads to increased soil nutrient mineralization and/or reduced soil acidity, which AM-fungi are more apt at taking advantage of. Indeed, it is no coincidence that most ruderal plant species are associated with AM-fungi, with their relatively rapid colonization strategies, generalist nature, lower level of specialization for specific soil conditions, and preference for dynamic, disturbed habitats. The fact that these ruderal AM plant species are generally faster growing and more disturbance tolerant due to their ability to take advantage of vegetation gaps, thus could begin to explain the patterns we observed: increased nutrient availability, less extreme pH, more dynamic conditions, and reduced competition in roadsides all act in favor of AM plant species.

These intriguing patterns raised an obvious next question to us at the Mountain Invasion Research Network: what would the impact be on non-native species? Surprisingly perhaps, virtually all non-native species in all our studied mountain regions are AM- (or some NM-)associated. This does make sense, however, if one thinks about how strongly they are associated with roadsides; which we now showed to be the ultimate heaven for AM-associated plant species. Yet we went one step further: we also showed that these AM-associated non-native species can now also much more easily escape from the roadside into the adjacent vegetation if the latter is more AM-dominated. Another reason why invasion levels in our ericoid and ectomycorrhizal vegetation in northern Scandinavia are virtually zero?

Misty mountain road in the Argentinian Andes

Our results represent an important first global study of the role of anthropogenic disturbances in shaping mountain plant communities through the mycorrhizal fungi they associate with. While these findings might sound theoretical, they have important implications for vegetation restoration worldwide, as they suggest that roadside disturbance can change the fundamental make-up of EcM- and ErM-dominated plant communities, potentially shifting communities between alternative stable states of mycorrhizal dominance that could be very difficult to reverse.

The paper can be found here!

Posted in Science | Tagged , , , , , | Leave a comment

Microclimate, an important part of ecology and biogeography

Posted on April 9, 2024 by lembrechtsjonas

From the first days of SoilTemp, we had the ambition to turn this project into more than ‘just a database’. SoilTemp could become the much-needed glue to stick together microclimate ecologists and biogeographers from across the globe, and give them an opportunity to learn from each other and jointly move the field forward.

It was time for such a community, we believed, as microclimate research was rapidly evolving into a booming scientific discipline and, after several decades of crucial fundamental work, the field was finally truly turning global.

This vision for a collaborative future saw its first, and perhaps most important, pinnacle in 2022, when we managed to bring together more than a hundred microclimate enthusiasts at the first Microclimate Ecology & Biogeography (ME&B) conference in Antwerp. For a week, the spacious rooms of our beautifully old conference facility was buzzing with microclimate research from all across the globe.

The conference facility at the University of Antwerp in Belgium

So inspiring was the atmosphere, that we decided to write down what we learned at that conference and inform the world about the major strides forward that our field has taken and, importantly, where we believe the field is and should be heading next.

Thanks to a gigantic effort by 97 conference participants, the output of that story now got published: a perspective piece called ‘Microclimate, an important part of ecology and biogeography’. The main message of that paper got nicely summarized by lead author Julia Kemppinen on X.

In the article, we summarize the current status of microclimate ecology and biogeography as its own scientific discipline. First, we highlight the latest research on interactions between microclimates and organisms, including how microclimates influence individuals, and through them populations, communities, and entire ecosystems. We show how this is increasingly being studied from the tropics to the poles.

We discuss the importance of microclimates in ecosystem management under climate change. We showcase new research in microclimate management with examples from biodiversity conservation, forestry, and urban ecology.

Finally, we summarize the recent advances in data acquisition, such as novel field sensors and remote sensing methods. We discuss microclimate modelling, mapping and data processing, including the accessibility of modelling tools & advantages of different modelling approaches.

Most importantly, we thus ask the question: what is next for microclimate ecology and biogeography? We identify major knowledge gaps that need to be filled for further advancing microclimate investigations, applications, and methods in the fields of ecology and biogeography. First of all, global microclimate research should be conscious of its biases. For instance, forest and tundra biomes are well represented in the microclimate literature, while microclimates matter to many terrestrial organisms across all terrestrial biomes. Second, it is also important to note that in the English-written scientific literature, microclimate ecology and biogeography are largely represented by studies, researchers and institutions of European, North American and Australian origin. We emphasize that these knowledge gaps and biases are important to consider in all future research that aims for a genuinely global coverage in microclimate investigations. This is key for making ecology and biogeography a more global endeavour.

That said, our most urgent research gaps include 1) spatiotemporal scaling of microclimate data, 2) quantifying and understanding the mismatches between macroclimate and microclimate in predicting responses of organisms to climate change, and 3) the need for more evidence on the outcomes of microclimate management.

The ME&B-excursion brought us to the gigantic Ecotron experimental facilities of the University of Hasselt. A more mechanistic understanding of microclimate and its role in ecosystem functioning is high on the list for ‘what’s next’ in microclimate science

With all that in mind, this is clearly an optimistic paper: we show how we can come together with our own expertise and background, from all over the world, and rapidly move a field of research forward. Who would have thought, ten years ago, that microclimate would have become such a crucial parameter in all of ecological research? Who would have thought that so many of us would now routinely incorporate a microclimate perspective in their research? We haven’t saved the world just yet, but at least we brought a global community together with that goal in mind.

As such, the ME&B-conference was indeed truly a pinnacle of our work at SoilTemp. However, mind you, this was just the beginning of microclimate science as a global discipline. Now, our ‘snowball’ is rolling downhill, and increasingly gaining momentum. Wanna jump on? A perfect opportunity would be the néxt ME&B-conference, happening this August in Helsinki.

You wouldn’t want to miss it, as there is still so much more to learn!

Posted in Science | Tagged , , , , , | Leave a comment

Scouting the cliffs of Freÿr

Posted on April 2, 2024 by lembrechtsjonas

On a sunny day in early spring, our team took the car to one of the most monumental locations in Belgium: the cliffs of Freÿr. These beautiful slabs of ‘naked’ limestone of up to 120 meter, overlooking the river Meuse, provide one of the most – if not the absolute most – favourite rock climbing locations in the whole of Belgium.

Overview of ‘les Rochers de Freÿr’ towering over the river Meuse.

With easily a 1000 climbing routes, ranging from the easiest to the hardest, the rock is legendary in Belgium rock climbing milieus. It is there that we headed to for the final scouting for a new – and extremely exciting – global survey: ‘MIREN Rocks’ (more on that – and how to join – here!).

Scouting the area for the perfect survey sites with local guide David

Indeed, as we speak, we finalized the protocol for MIRENs’ global rock survey, using our experience from the cliffs of Freÿr as an example. We checked for routes to sample, estimated variability in orientation, difficulty and botanical diversity and, most importantly, tested out ways to install microclimate sensors on a cliff face.

Microclimate monitoring on the cliffs with the TOMST thermologgers

It’s the latter that perhaps makes me most excited: we can now plug in one of our trusted TOMST loggers – the simpler thermologger one – in a gap in the cliff face, and as such start measuring the true temperatures our cliff plants are experiencing. The ultimate cross-over between my two favourite networks, MIREN and SoilTemp.

The three-layered shield helps reduce the error resulting from direct solar radiation. Nevertheless, it will be hard to get accurate reads of the local temperature due to the extreme radiation absorbed, reflected and emitted by sunbathing cliffs

Now, that protocol is ready for action, also thanks to the extensive contributions by experts from across the globe. As such, the protocol has grown into a true community effort, and we hope it will be a resource for many climbers and ecologists to help us track rock cliff vegetation and the implications of rock climbing on it.

Selecting the optimal climbing routes for our upcoming monitoring

So, take this as a wake-up call: are you an ecologists, botanist or the like, and do you like to climb rocks? Then please join MIREN Rocks! All information can be found here.

Cliff surveys would go fast in some places if you count the number of plants on them – where it not for the issues related with getting up there!
Asplenium sp. – lover of cliffs
In some spots, the cliffs drop directly into the Meuse
The surrounding Natura 2000-forests were starting to be covered in beautiful spring flowers – here Anemone nemorosa
Posted in Belgium | Tagged , , , , , , , , , | Leave a comment

A new beginning

Posted on March 26, 2024 by lembrechtsjonas
My new home in Utrecht

Last week, I spent a delightful day at my soon-to-be new home university in Utrecht. A day packed with discussions on scientific ambitions, teaching plans, and meeting new colleagues and ongoing research in the group. The first tulip of the year made it all feel like an official new beginning.

The real start will have to wait till May 1st, but you bet that I’m already making plenty of exciting new plans in advance!

Posted in Netherlands | Tagged , , , , , | Leave a comment