Instrumentation and remote sensing
Core project 3 – The so-called core projects of the Biodiversity Exploratories (BE) emerged from the on-site project selection and the establishment of the exploratories (2006-2008), and have been providing the infrastructure since 2008.
The systematic and continuous recording of environmental data is an important element of the research infrastructure in the Biodiversity Exploratories.
Core Project 3 “Instrumentation and Remote Sensing” is responsible for (i) the systematic and large-area measurement and recording of meteorological and soil environmental variables and (ii) the provision of remote sensing and geospatial data on land cover and land use.
For this purpose, Core Project 3 develops and operates an important research infrastructure consisting of a very large network of climate stations, climate towers and remote sensing platforms, sensors and imagery. It regularly collects climatic and remote sensing data according to standardized procedures to produce coherent time series. These are made available through special database applications being developed in the project.
Instrumentation has therefore been a key component of the research infrastructure since the beginning of the Biodiversity Exploratories. Due to strong demand for remote sensing / geospatial data and services, the remote sensing component was added to the project in 2014. This will allow the establishment of an archive of geospatial data and remote sensing products accessible to all research groups, and coordinate the acquisition of new data, e.g., with drones and aircraft. The services offered by Core Project 3 do also contribute to making the geodata in the Biodiversity Exploratories standardized and harmonized.
The goal of the core project is to establish and support the measurement and sensor technology and to provide environmental data as processed basic products (“ready to be used be the Exploratories ´ researchers”) to address a wide range of research questions in the Biodiversity Exploratories. These data sets include:
- Climate variables: Climate monitoring stations on grassland and forest stations, TreeTalker on forest experiments (FOX), climate towers for vertical profiles.
- Vegetation structure variables: Indices of vegetation spatial structure from laser scanning and photogrammetric point clouds.
- Audio environment: (ultrasound) recordings
- Specific vegetation parameters: Tree sap flow measurement, LAI
- Provision and processing of land use and land cover maps
- Landscape structure: recording and quantification of landscape structure
- Coordination and execution of drone surveys
- Processing of Copernicus and Landsat satellite data to produce continuous time series and derived products (productivity, land use intensity, change analysis)
In addition to these products, the core project would like to foster the use of climate and geodata in biodiversity research. To this end, the core project offers workshops on the evaluation and use of climate and geodata.
Environmental measurement equipment
Meteorological and soil variables are among the main factors influencing the organismic diversity and functionality of ecosystem cycles. In order to carry out systematic and large-areameasurement and recording of these environmental variables, more than 400 permanent measuring devices are installed and maintained in the three Exploratories Schorfheide-Chorin, Hainich-Dün and on the Swabian Alb.
In each Exploratory, >140 measuring stations were set up, of which 50 each were installed in grassland and 91-111 each in the respective forest areas. Due to the very large number of stations, not all of them were equipped identically, but with graded sensor groups. Currently, the following station types are in use:
- 279 Core Environmental Monitoring Units (CEMU)
- 21 Enhanced Environmental Monitoring Stations (EEMU)
- 148 Fox Environmental Monitoring Units (FEMU) a la Treetalker
Type CEMU is used on all experimental sample plots (EPs) of the Exploratories in the forest areas and grassland. In addition to general meteorological variables such as air temperature, surface temperature, and relative humidity, measurements are also taken in the soil. Five probes are deployed in the topsoil to continuously record soil temperature and moisture.
Type EEMU is deployed on selected intensive sample plots (VIPs) and topographically important sample areas, and also has complementary sensor technology to measure precipitation, wind speed and direction, incoming and outgoing radiation components of shortwave and long-wave radiation, photosynthetically active radiation (PAR), and atmospheric pressure.
Type FEMU/ Treetalker is used on the forest experiment plots (FOX). In addition to general meteorological data such as air temperature and humidity, soil moisture and temperature measurements are also taken in the soil and solar radiation is recorded under the canopy in 12 spectral bands (450 – 860nm).
A visualization of the measured temperature time series can be seen here:
In the biosphere reserve Schorfheide-Chorin and on the Swabian Alb, a measurement tower with a height of 44 m and 37 m, respectively, was installed. Both towers include the sensor technology of the EEMUs and additionally measure the air temperature, air humidity and the PAR in the profile. Two further towers, which were built by the Max Planck Institute for Biogeochemistry, can also be used in the Hainich-Dün National Park.
On selected plots, audio recorders capture soundscape in both audible and ultrasonic frequencies. Estimates of the occurrence and intensity of activity of birds, bats, and auditory ecosystem disturbances are recorded semi regularly. This provides the opportunity to systematically and simultaneously record and analyze differences in occurrence and activity between seasons and plots.
Remote sensing & geo data
The objective of the remote sensing component is to complement the large-area collection of meterological variables with area-wide remote sensing coverage in the three Exploratories. By providing and updating geospatial data and other products – derived from remote sensing data – as ready-processed Analysis Ready Data (ARD), it should be possible for non-remote sensing scientists to integrate spatial and temporal analyses based on remote sensing observations into their research and thus gain new insights into the functional relationships at different scale levels of biodiversity. For this purpose, a remote sensing infrastructure has been established in the core project and will be presented in the following.
Unmaned Aerial Systems (UAS) offer the possibility of temporally flexible high-resolution images of smaller areas (plots + environment). This is especially useful when image acquisition is to be synchronized with a particular field acquisition, e.g. to develop image classification models. Thus, both (1) the plot state can be documented during the field recordings and (2) training data can be collected for the development of classification models. The very high spatial resolution of the drone imagery allows the detection of very small features. Since 2017, the core project has been conducting drone aerial surveys of the Exploratories. For this purpose, the project has several drones (copters, fixed-wing aircraft) and different sensors (thermal, RGB, multi-spectral, hyperspectral) at its disposal. At the same time, Core Project 3 has established the necessary computer hardware and software and developed specialized data processing chains .
The use of drones, however, is subject to strict legal requirements which often require special permits, especially in nature conservation areas. In cooperation with the nature conservation authorities, the Core Project 3 was able to develop good solutions in many cases that allow the use of drones even in sensitive areas.
Satellite and aircraft data
To address the diversity of research questions and methods within the biodiversity exploratories, a wide range of remote sensing platforms are used in addition to drones. These range from optical satellite imagery (e.g. Pléiades, RapidEye, Planet, Sentinel-2) to airborne hyperspectral and LiDAR sensors as shown in Table 1.
|System||Plattform||Spatial Resolution (m)||Number of bands||Spectral resolution(μm)Spektrale Auflösung (μm)||Temporal resolution||Coverage|
|RapidEye||Satellit||5||5||0.44 – 0.85||2009-2015 ~ min. three phenological periods per year||area-wide|
|Planet||Satellit||3.5||4||0.44-0.88||2020-2021, saeveral recordings per year||area-wide|
|Landsat*||Satellit||15 – 120||4 - 8||0.43 – 12.5||1972 - 2014, ~ annually||area-wide|
|MODIS**||Satellit||250 – 1000||36||0.40 – 14.39||2001 - 2014, ~ daily||area-wide|
|Sentinel-2||Satellit||10-20||13||0.49-2.2||2016-,~ every five days||area-wide|
|Pléiades||Satellit||0.5– 2||5||0.43 – 9.40||unique (2015)||100 km² per Exploratory|
|Hyperspektral||Plane||1||> 200||0.40 – 2.40||unique (2015)||100 km² per Exploratory|
|LiDAR||Plane||~14 points/m²||Full wave form||---||unique (2015)||100 km² per Exploratory|
Databases and processing
The climate measurement facilities described above record >70,000 readings daily in each research area. These are automatically transmitted to Marburg via the mobile network and collected in the climate data time series database TubeDB (https://environmentalinformatics-marburg.github.io/tubedb/), which was specially developed for the Exploratories. The researchers can use these processed climate data directly via BExIS and visualize and export them according to individual requirements, such as quality correction, temporal aggregation and interpolation.
We undertake the processing of wildlife camera images, primarily capturing the occurrence of mammals. The processing categorizes the images into images of animals, humans, and misrecorded images. The position of animals in the images is marked so that subsequent manual or machine species identification is simplified. Our image management software enables the review of the categorized images and the manual identification.
Our audio data management software is the central collection place for the recorded audio data. In the web interface recordings can be listened to, visualized as spectrum The machine learning (ML) based automatic labeling runs in cycles of initial manual labeling, ML training, manual evaluation of the automatically generated labels and renewed ML training.
Remote sensing data
The remote sensing database RSDB, (https://environmentalinformatics-marburg.github.io/rsdb/) manages the processed raster data, point clouds and vector data. The web interface provides functions for searching and exploring the stored geodata interactive visualizations, processing and data export. In addition, all data of the RSDB can be processed directly in R via the R package rsdb (https://github.com/environmentalinformatics-marburg/rsdb/tree/master/r-package).
Current phases (2020-2023)
UAV flights in the forest and grassland
For many research projects in the Exploratories, it is important to take high-resolution images of the plots at the same time as or at a very close distance from one’s own field recordings. The use of drones has proven to be particularly efficient here. We have organized, conducted, and evaluated UAV surveys in both grassland and forest from 2017 until 2021 and will continue so.
Landscape structure analyses
Ecosystem processes affecting biodiversity operate at very different scales. Ecologists frequently hypothesize that many of the processes observed on the intensively sampled EPs be significantly influenced by the neighborhood surrounding the plot. Current and historical land use and land cover maps can be used to examine the adjacent landscapes. This will allow the intensive observations on the EPs to be placed in the landscape context. Thematic classification keys adapted to the different questions will be developed. In the forest, for example, different tree species groups and other features are differentiated, in the grassland mainly agricultural use and woody structures (e.g. shrubs) serve as classification criteria. These thematic maps allow a quantitative description of the horizontal landscape structure of the three Exploratories. Elevation information e.g. from the Airborne Laserscanner (ALS) data extends this description by a vertical component, so that a 3D description of the vegetation and landscape structure is possible.
Satellite image time series
The interactions between ecosystem functions and land use intensity is a central research topic in the Biodiversity Exploratories ´ projects. The currently observed state is the result of processes acting over a long period of time. Therefore, describing the historical evolution can help to provide information about the current state. Remote sensing methods offer the possibility to describe the development of a landscape with the help of time series analyses. We provide such time series based on different satellite platforms. Through the publication of the Landsat-Archive a data source is available which allows a description of the development back to the 1980s. In addition, the data of the European Copernicus program are used to create current satellite image time series with high temporal resolution, which are used, for example, to record phenological processes.
The network of climate stations distributed across the Exploratories provides time series of point measurements of the local microclimate. Using machine learning methods and with the help of Sentinel-2 time series and digital terrain models, we generate areal microclimate maps. These microclimate maps are complemented by reliability estimates which are also generated as maps.
Smaller gaps in the climate time series caused by station failures are optionally interpolated automatically from our climate data time series database with high precision. For larger gaps or for periods before the start of the climate stations, we offer an algorithm for filling the climate data utilizing external DWD weather data.
Precipitation is measured at some of our climate stations. Based on the DWD RADOLAN product, we can create precipitation time series for all plots.
- Recordings the phenological changes of the forest structure
- Evaluation of historical satellite time series to capture disturbance intensity
- Combined hyperspectral and laser scanner survey in the three Exploratories in 2015
- Mapping forest structure metrics using Airborne Laser Scanner (ALS) data
- Creation of RapidEye satellite time series