Pasture biomass mapping
24 Nov 2020
The uplands areas of the Jura Mountains are characterised by grassy pastures (or grassland) along the series of valleys that make up the mountains. These pastures have isolated single trees (usually conifers) or groups of trees. The valley areas are usually divided into a series of farms each of which has a chalet d'alpage (alpine chalet) that provides accommodation in the summer during the grazing season and in some cases milking facilities for a farm's herd. Herds are traditionally milk herds but are increasingly becoming beef and/or breeding herds.
The pastures are surrounded by forests, both forest plantations and forests based on natural forests, many of which are managed in the sense that commercial logging is carried out.
The mountain pastures are an important defining aspect of the Jura Mountains (some 20% of the uplands are covered by pastures). It is therefore interesting and useful to understand how the pastures change from late-spring to early autumn while they are being grazed.
Ideally, it may even be possible to map and monitor the movement of the herds so that vistors can avoid encountering herds while visiting the mountains, much of which forms part of Jura Mountain national parks. We are mainly interested in the area (see main map) to the north-east of Saint-Cergue in the Swiss communes of Saint-Cergue, Arzier and Bassins that forms part of the Parc Jura Vaudois.
Using satellite imagery for vegetation mapping is well established and used widely, notably for understanding grazing dynamics. The predominantly smooth, grassy areas of pastures in the Jura Mountains can be mapped visually using Bing and Google Earth satellite imagery and relief mapping using a Digital Elevation Model (in our case the high-resolution Swisstopo swissAlti3D product) as shown on the main map. These grassy areas are drawn (see the editing version of the pasture map) using an approximately 10-meter wide buffer between a grassy area and the surrounding forest and then removing groups of trees and their 10-metre buffer. Areas covered by isolated single trees are not extracted.
Our pasture map shows the grassy pastures for several areas. In many cases, satellite images but more generally ground truth can be used to identify the temporary fences that sub-divide pastures. These fences are generally set up in early-spring and are often removed in early-autumn so that a herd can graze over a large area. The grassy areas are themselves often part of a larger area, called a farm, that has a boundary, usually in the form of dry stone walls or temporary fencing that replaces degraded walls.
Without going into too much detail, it is simply indicated that the Jura Mountains pasture biomass mapping uses the Sentinel-2 Leaf Area Index (SeLI) which exploits Sentinel 2 B5 red-edge (705nm) and B8A near-infra red (865nm) bands.
Cloud-free Sentinel2 images (mainly 2A but occasionally those from the 2B satellite) are downloaded from ESA's Copernics Open Access Hub and the relevant bands are pan-sharpened to a resolution of 10 metres using the Windows 64-bit desktop version of the Sentinel Application Platform (SNAP) SuperResolution plugin. The SeLI is then calculated, and the generated geotiff image reprojected using QGIS from the EPSG 32631 projection to the EPSG 4326 projection. Map tiles are created using the QGIS gdal2tiles tool.
Some further comments on the results of the biomass mapping will be indicated later. For the several areas that have been mapped so far it is noted that the highest-level of biomass (designated using the lightest colour in the mapping) that arises in the spring before grazing starts is mainly removed between mid-July and mid-August. There follows relatively little change between mid-August and mid-September when grazing ceases. The maximum biomass level is quickly regained by mid-October.
There are of course varations within pastures for a given area and between areas. It is also possible to identfy the effect of closing off parts of a pasture using temporary fencing (see in the pasture map and in the image the difference between the north-western and south-eastern parts of area 7 in mid-September 2020 which are divided by a fence that is indicated as a red line).
In summary, the recently proposed and relatively simple multi-vegetation SeLI index that minimises saturation at high levels of biomass appears to provide a reliable and useful indicator for grazing dyanmics in Jura Mountain pastures. A complete mapping over several years is in progress.
The main aim is to establish the feature types that should be used in multi-sensor data extraction for unsupervised Local Climate Zone classification, possibly using the Dempster-Shafer Theory (see V. Saeidi et al).