class: top, left, inverse, title-slide # Mapping biodiversity changes across
spatio-temporal scales ##
PhD Presentation ###
Student: François Leroy
Supervisor: Petr Keil ### Czech University of Life Sciences
Prague --- class: inverse, center, middle Focus on ## Species Richness --- class: inverse, center, middle ## Species Richness is an extensive variable = proportional to spatial and temporal scales --- # Species richness and spatial scale : the SAR .pull-left[ <br><br> - **SAR** = Species Area Relationship (Arrhenius 1921) <br><br><br> - Linear on a log-log scale <!-- <br> --> <!-- - The slope indicates the spatial `\(\beta-diversity\)` --> <!-- <br> --> <!-- - Can be used to extrapolate species richness (*e.g.* Kunin *et al.* 2018) --> ] .pull-right[ <img src="data:image/png;base64,#images/sar.png" width="800" height="430px" /> .credit[Brown *et al.* 2007] ] --- # Species richness and temporal scale: the STR .pull-left[ <br><br> - **STR** = Species-Time Relationship <br><br><br> - Can also be well described by a power law <!-- <br> --> <!-- - Emerges from both sampling and ecological processes --> <!-- <br> --> <!-- .center[**And... we don't know much more about it**] --> ] .pull-right[ <img src="data:image/png;base64,#images/str.png" width="552" height="430px" /> .credit[STR of the Breeding Bird Survey for different North American states, White 2004] ] --- # Species richness, space and time: the STAR .pull-left[ <br> - **STAR** = Species-Time-Area Relationship <!-- <br> --> <!-- - Results from both environmental variability and biological processes --> <br><br><br> <!-- The STAR raises the question: --> <!-- <br> --> .center[**Temporal and spatial scales highly drive species richness**] ] .pull-right[ <img src="data:image/png;base64,#images/star.png" width="352" /> .credit[Plant species richness as a function of interacting space and time, Adler & Lauenroth 2003] ] --- class: inverse, center, middle What about the relationship between space, time and... ## Species richness <span style="color: red;">trends</span> ? --- # Biodiversity changes and spatial scales <br><br> .pull-left[ * According to the taxa: .center[*"Species richness change can increase, decrease, reverse or be unimodal accross spatial scales."*] * For North American birds, larger spatial scales seem to increase positively the species richness (Oikos, Chase *et al* 2019) <br> ] .pull-right[ .center[ <img src="data:image/png;base64,#images/chase_2019.PNG" width="100%" /> ] ] --- # Biodiversity changes and temporal scales... <br><br><br> .center[...nothing in sight in the scientific literature] --- class: inverse, center, middle # Problematic: ### What are the links between <span style="color: red;">species-richness trends</span> and <span style="color: red;">spatio-temporal scales</span>? -- <br><br><br> <b>Taxon:</b> Birds <b>Location:</b> So far Czech Republic <b>Methods:</b> Machine learning (CART, GLM...) --- class: inverse, center, middle # Goal Use heterogeneous avian biodiversity datasets to model species richness trends across continuums of space and time scales .center[ <img src="data:image/png;base64,#images/star.png" width="50%" /> ] --- # Datasets <br> We need heterogeneous datasets with: * Time series (for the trend) * Several spatial scales * Several temporal scales <br><br> 2 datasets: 1. The **atlas** data from the Czech Society for Ornithology (Česká společnost ornitologická). Courtesy of Vladimír Bejček, Karel Šťastný and Ivan Mikuláš. 2. Local time series from the **Breeding Bird Survey** (Jednotný program sčítání ptáků). Courtesy of Jiří Reif. --- # The Atlas dataset .pull-left[ .center[**Temporal scales**] 4 time periods, 3 different time spans: * M1 = 1973-1977 (**5 years**) * M2 = 1985-1989 (**5 years**) * M3 = 2001-2003 (**3 years**) * M4 = 2014-2017 (**4 years**) ] .pull-right[ .center[**Spatial scales**] Ranging from less than **100 Km** `\(^2\)` to **80 000 Km** `\(^2\)` (the entire Czech Republic) ] --- # The Atlas dataset ![](data:image/png;base64,#index_files/figure-html/unnamed-chunk-10-1.png)<!-- --> --- # The BBS dataset * Local time series * 350 transects * 20 census points per transect * From 1982 to 2020 * At least 2 censuses per transect and per year * 4 different **spatial scales** * **Temporal scales** ranging from 0.5 year to decades --- # The BBS dataset
--- class: inverse, center, middle # Data analysis --- # Species richness trend across spatial scales * For each spatial scale, we computed the **species richness trend** (slope of linear regression) * 3 points in time (1987, 2002, 2015) * Harmonization of the two datasets --- # Species richness trend across spatial scales <img src="data:image/png;base64,#images/atlas_maps_boxplot.png" width="95%" /> --- # Species richness trend across spatial scales <img src="data:image/png;base64,#images/boxplot.png" width="95%" /> --- # Species richness trend across spatial scales .pull-left[ * Increasing species richness trend with spatial scale * Coherent with Chase *et. al* 2019 * Easier extirpation at smaller scales * Lower proportion of extinction with increasing scales (Keil *et. al* 2018) * More habitat heterogeneity at larger scales ] .pull-right[ <br> <img src="data:image/png;base64,#images/boxplot.png" width="95%" /> ] .center[ <br> **Question:** is this result observed for a continuum of spatial scales? ] --- class: inverse, center, middle # Modeling --- # Random Forest 1. Understand the drivers of species richness changes 2. Predict the species richness Predictions will allow me to fill the boxplot gaps, spatial and temporal gaps of the datasets <br> **Formula:** ```r randomForest(sr ~ Lat + Long + AREA + Date + polypoint_ratio + time_span) ``` --- # Partial plots .center[ *Species-time relationship (STR) through time* <img src="data:image/png;base64,#images/str_overtime.png" width="50%" /> ] --- # Predicitons * For <span style="color: blue;">50 `\(Km²\)`</span> and <span style="color: blue;">10 000 `\(Km²\)`</span> .pull-left[ <br> <img src="data:image/png;base64,#images/50sqkm.png" width="100%" /> ] .pull-right[ <br> <img src="data:image/png;base64,#images/10000sqkm.png" width="100%" /> ] --- # Predicitons * For <span style="color: blue;">50 `\(Km²\)`</span> and <span style="color: blue;">10 000 `\(Km²\)`</span> .center[ <img src="data:image/png;base64,#images/boxplot_with_predictions.png" width="87%" /> ] --- # Next steps <br> * Take into account the sampling effort <br> * Take into account environmental covariates to understand at which scales they interact with biodiversity dynamic <br> * Explore the link with temporal scales <br> * Enlarge the model to Europe <br> * Study other biodiversity metrics and other taxa (lepidopterans, amphibians...) --- class: inverse, center, middle # Thank you for your attention .footnote[ Email: leroy@fzp.czu.cz Twitter: @FrsLry ] --- class: inverse, center, middle ## Supplementary slides --- # Observed vs. predicted .center[ <img src="data:image/png;base64,#images/obs_vs_pred.png" width="80%" /> ] --- # SAR and STR .pull-left[ .center[**The SAR**] <img src="data:image/png;base64,#images/sar_rf.jpg" width="100%" /> ] .pull-right[ .center[**The STR**] <img src="data:image/png;base64,#images/str_rf.jpg" width="100%" /> ]