Post provided by Apolline Louvet

Tree base in an urban area, partly occupied by spontaneous vegetation. Picture credit: Apolline Louvet.

Seed banks play a key role in plant metapopulations, however, detecting seed banks remains challenging. The current method of assessing the effect of seed banks on plant metapopulation dynamics focuses on the presence of standing vegetation. This has direct implications for plant metapopulation models, which require data on the absence of a seed bank to understand plant dynamics.

Our paper, ‘Detecting seed bank influence on plant metapopulation dynamics’ in Methods in Ecology and Evolution, introduces a new metric on plant metapopulations, which assesses the seed bank contribution to the global observed dynamics. In this post, we recall what led us to develop this metric.

Plants, seed banks and tree bases

It all started with a monitoring program of tree bases along streets in Paris. As tree bases are urban green spaces, they exhibit special features, such as high levels of fragmentation, and frequent extinction events caused by city gardeners weeding them. Moreover, they can potentially act as corridors between green spaces. Therefore, in order to broaden the understanding of plant distribution and plant dynamics in tree bases, floristic inventories were performed in more than 1000 tree bases in Paris, over a duration of 10 years.

The results of this monitoring program were already used to uncover the drivers of plant distribution and plant dynamics in tree bases. Since seeds banks can play an important part in disturbed environments, we wanted to investigate the contribution of the seed bank (if present) to plant dynamics, using the data from the monitoring program.

Plants growing in urban tree bases are often considered as metapopulations. The plants can generally not grow on the road or on the sidewalk, and the action of city gardeners result in high extinction probabilities. And here arose our first issue: we only had presence/absence data for standing vegetation. However, most existing estimators for metapopulation models which use only presence/absence data for plants also require the absence of a seed bank. In particular, we cannot use them to infer whether a seed bank was present.

Parameter estimation in presence of a seed bank

Luckily, a new estimation method has been introduced in a recent article, which allowed us to estimate metapopulation parameters when a seed bank is present. This estimator only uses presence/absence data for plants. We wanted to use this estimator to assess, for each plant species, whether a seed bank was present or not. Since many tree bases were monitored, and this during a long observation duration, it seemed at first that we would be able to directly apply the estimation method to our dataset. However, there was no reason to assume that metapopulation parameters were identical from one street to another. Therefore, even though more than 1000 tree bases were inventoried, we could only apply the estimator to smaller subsets of 200 to as low as 31 tree bases. And that is how our second issue arose: we did not have enough data for the estimator to be accurate enough for our purpose. We could not draw any conclusion on the presence of a seed bank, nor estimate how the seed bank was contributing to plant dynamics.

Illustration of the intra-patch metapopulation dynamics in the model used. Full arrows indicate transitions that always occur, while dashed arrows indicate events that only occur with a certain probability. Picture credit: Apolline Louvet.

Detection issues arise

Instead of waiting for 20 more years in order to have enough data, we investigated whether there were certain parameter sets for which the accuracy issue was particularly important. It turned out metapopulation parameters related to seed bank presence were mostly ill-estimated when one of these two situations occurred:

  1. There were a lot of colonization events, that is, new seeds entering the tree bases from an external source
  2. At each generation, a fraction of the seeds just produced almost always germinated

Under these two situations, the presence of a seed bank is erased by colonization or germination events, which bring new seeds. The seed bank died? This is not a big deal, as new seeds will come from outside the metapopulation, or as there is standing vegetation which will produce new seeds. In terms of observed dynamics, metapopulations with these parameters are not really that different from metapopulations without seed banks.

The SBCE probability

Therefore, we focused on the type of event which was typical of the presence of a seed bank: a delayed germination, without any colonization or seed bank death event happening in between. The metric we introduced measures the frequency of this kind of event. We called it the Seed Bank Characteristic Event probability, or SBCE probability in short.

With this new metric, we could not completely answer whether a seed bank was actually present or not. But this was not a problem, because we were initially interested on assessing the impact of the presence of a seed bank on the observed plant metapopulation dynamics. Therefore, we used this metric to investigate whether a seed bank was contributing significantly to the observed dynamics.

Application to real datasets

A typical tree base along a street in an urban area. Picture credit: Apolline Louvet.

In our article, we explain how the metric can be applied to real datasets. Among other things, we detailed how to interpret the values it takes, and we indicate some precautions that have to be taken while designing the data collection step. One key result of our analyses is the fact that our approach is suited to datasets coming from citizen science, even though these datasets are expected to contain more observation errors than others. We also applied our method to the dataset of plants in urban tree bases.


In order to assess the impact of the presence of a seed bank on the observed plant dynamics in urban tree bases, we introduced a new metric, the SBCE probability. Our approach can be applied to other plant metapopulations, and highlights that in some situations, detecting the effects of a phenomenon can be easier than detecting the phenomenon itself.

To read the full study, visit the Methods in Ecology and Evolution article: “Detecting seed bank influence on plant metapopulation dynamics”.