Welcome to the 5th practical.

In this practical we will be using a combination of Google Earth Engine and R to understand biodiversity patterns. It is essential that you found your way trough the first practical as these practical sessions intentionally have an increasing complexity.

Step 1: Required softwares and datasets: overview

Softwares

As mentioned on the starting page, you’ll need Rstudio and a working account for code editor.

Problem setting

As introduced in class, we will use camera trap data: these are cameras where each time an animal passes, a picture is taken and AI is used to identify the animal. After some data cleaning and quality checks, the total amount of unique species (species richness) can be calculated for each camera (so each location). The question in this practical will be: for a dataset collected in Costa Rica: which gradients in species richness exist? Based on what we saw in class, we assume here that if there are gradients, they might depend on NDVI. Additionally, because the dataset is acquired on the flanks of a volcano, we will also check the effect of elevation.

Datasets

This is an overview of the problem identification:

building block	decision
Geographic unit	points
temporal scale	2019(!)
Assumption	Species richness depends on the amount of energy available. Here we will use NDVI as a proxy for net primary production. Additionally, we assume that richness decreases with temperature, and thus will decrease when elevations are higher
Dimensions	We’ll consider the elevation and NDVI
Dimension description	elevation: ? , temperature: NDVI: calculated from the LANDSAT 8 mission

(!)In reality the data was collected intermittently since 2007 but for the purposes of this practical, the data are aggregated and strongly simplified. Dealing with the temporal dimension of this dataset and with the particularities of data cleaning for camera trapping lies beyond the scope of this course. We assume, for the purposes of this exercise, the data is suitable for analysis against a landsat image of the year 2019 (see later).

And we’ll use following data structures and sources:

Dataset	Type	Source	Access point
TEAM-camera data Costa Rica	vector	downloadable from canvas, original source: https://www.wildlifeinsights.org/team-network
NDVI	?	LANDSAT 8	Google Earth Engine: see below
?	?	?	?

---

Step 2: accessing the data

** SPECIES RICHNESS**

We have processed the camera trap data, so that it is compiled into a shape-file that you can download from canvas: download the folder and unzip in your working directory.

Then you can load the shapefile into google earth engine so that you can do analysis with/on it:

Now it needs a few minutes to load (you can check the progress in the right panel ‘tasks’).

Once it is finished, we can load it and visualize it on the map. We will also draw a square around it which will be useful to calculate the NDVI. Loading the camera points and drawing the region of interest around them is shown here:

NDVI MAP

The points are relatively close together, the NDVI datasets that are readily available in the Google Earth Catalogue have a rough spatial resolution (e.g. the based on MODIS satellites has a 500m resolution.

We will thus have to make it ourselves based on satellites with a better spatial resolution. One of the options is the LANDSAT 8 mission which collects spectral information in the red and near-infrared spectrum (the bands we need to calculate NDVI) at 30m resolution. All the information on this product is given here.

/// You should have the script ready where you have imported the table (the camera points) and drawn a rectangle around them 
// let's also visualize the points; 
Map.addLayer(table, {color: 'FF0000'},'qgis shapefile');
//and center the map around the rectangle we made: 
Map.centerObject(clip);

//Now we can start calculating NDVI based on the landsat 8 reflectances in the red and near infrared bands (use the link above to find out which bands these are) 
var landsat = ee.ImageCollection("LANDSAT/LC08/C02/T1_L2");
print(landsat); // if you want to print this to the console, you'll get an error: indeed the imageCollection is way too big
//so, let's first select on the region and the dates of interest (only within the clip, and the whole year of 2019)
var landsat = ee.ImageCollection("LANDSAT/LC08/C02/T1_TOA").filterBounds(clip).filterDate('2019-01-01', '2019-12-31');
print(landsat); //great this works
// in the tropics, and in mountaineous areas, clouds can be an issue when calculating NDVI. So let's sort the image collection according to cloud cover, and then take the image with the least clouds
var bestlandsat = ee.Image(landsat.sort('CLOUD_COVER').first());
print(bestlandsat);

// now let's visualize this image: 
// Define the visualization parameters.
var vizParams = {
  bands: ['B4', 'B3', 'B2'],
  min: 0,
  max: 0.3
};

// Center the map and display the image.
Map.addLayer(bestlandsat, vizParams, 'true color composite');

//now let's calucalte ndvi of this image
// Compute the Normalized Difference Vegetation Index (NDVI): we need to define the nir and red bands: find out which ones these are and replace the X and Y below with the correct band numbers: 
var nir = bestlandsat.select('BX');
var red = bestlandsat.select('BY');
var ndvi = nir.subtract(red).divide(nir.add(red)).rename('NDVI'); // this is (NIR-RED)/(NIR+RED)


// Now let's visualize this ndvi map we made:                   
var ndvizoom = ndvi.clip(clip); //only cut out the region that what we need
var ndviParams = {min: -1, max: 1, palette: ['blue', 'white', 'green']}; // set visualization parameters
Map.addLayer(ndvizoom, ndviParams, 'NDVI image'); //plot the NDVI


// GREAT! now we can simply extract the values of NDVI for the different points and export it to a csv: 
var camera_ndvi = ndvizoom.reduceRegions(table, ee.Reducer.mean( ),30); // the '30' here indicates the resolution at which you want to collect the data, i.e. in this case the resolution of the dem , in this case 90m
Export.table.toDrive({
  collection: camera_ndvi,
  description:'camera_ndvi',
  fileFormat: 'CSV'
}); //we write the elevation values per camera location to a csv
// after running this it should appear in your 'tasks' here to the right: click on 'run': after finalization it will appear in your google drive

Now we are ready for the next step: extracting the elevation data for each point