Deep Species Distribution Modeling From Sentinel-2 Image Time-Series: A Global Scale Analysis on the Orchid Family

Abstract

Species distribution models (SDMs) are widely used numerical tools that rely on correlations between geolocated presences (and possibly absences) and environmental predictors to model the ecological preferences of species. Recently, SDMs exploiting deep learning and remote sensing images have emerged and have demonstrated high predictive performance. In particular, it has been shown that one of the key advantages of these models (called deep-SDMs) is their ability to capture the spatial structure of the landscape, unlike prior models. In this paper, we examine whether the temporal dimension of remote sensing images can also be exploited by deep-SDMs. Indeed, satellites such as Sentinel-2 are now providing data with a high temporal revisit, and it is likely that the resulting time-series of images contain relevant information about the seasonal variations of the environment and vegetation. To confirm this hypothesis, we built a substantial and original dataset (called DeepOrchidSeries) aimed at modeling the distribution of orchids on a global scale based on Sentinel-2 image time series. It includes around 1 million occurrences of orchids worldwide, each being paired with a 12-month-long time series of high-resolution images (640 x 640 m RGB+IR patches centered on the geolocated observations). This ambitious dataset enabled us to train several deep-SDMs based on convolutional neural networks (CNNs) whose input was extended to include the temporal dimension. To quantify the contribution of the temporal dimension, we designed a novel interpretability methodology based on temporal permutation tests, temporal sampling, and temporal averaging. We show that the predictive performance of the model is greatly increased by the seasonality information contained in the temporal series. In particular, occurrence-poor species and diversity-rich regions are the ones that benefit the most from this improvement, revealing the importance of habitat's temporal dynamics to characterize species distribution.

Rank-based deep learning from citizen-science data to model plant communities

“…4), and it would likely be less detailed without ubiquitous citizen-science observations. Predictive performance may improve if rank-based DNNs were con ned to the relevant species and trained with detailed remote-sensing predictors 20,33 . Maps of potentially dominant species may be useful for forest rangers to identify locally competitive species 47 , but comparable observation probabilities are also advantageous in other contexts.…”

“…Compared to SDMs, DNN-based modelling frameworks offer interesting new perspectives. They allow considering the spatial con guration of a landscape as predictor 21 , and they can cope with huge data sets 20,21 . Additionally, DNN-based approaches typically model the distributions of many species jointly, which can be an e cient way of handling spatial biases.…”

“…Driven by their success in data science 16 , deep neural networks (DNNs) have become an increasingly popular alternative to model biodiversity from observational data [17][18][19][20][21][22][23] . Compared to SDMs, DNN-based modelling frameworks offer interesting new perspectives.…”

Predicting species distributions in the open ocean with convolutional neural networks

“…Treating the problem as a classification task allows estimating the conditional probability of y (the observed species) given that an observation has been made in the environment x. It has the advantage to be (asymptotically) invariant to the spatial sampling effort but it is sensitive to the taxonomic reporting bias (the fact that some species are more observed than others) [45]. In the absence of taxonomic reporting bias, the estimated probabilities would converge to the relative probability of each species given the environment.…”

Use of remote sensing and image processing for identification of wild orchids

“…The developed approach was successful in assessing 4 of the preferred orchid species. An ambitious project to globally map orchids was reported in 2022 (Estopinan et al, 2022). Again, using satellite imagery data, they built a dataset for training and validation of a species distribution model (SDM).…”