Grassland occupies a large proportion of utilised agricultural area, especially in mountainous regions. Despite its importance current and reliable data on grassland yields and cutting frequencies with a sufficient spatial coverage are lacking. Both are essential for optimizing the use of grassland, nature conservation and policy consultation. Model approaches for the assessment of grassland yields take cutting dates and frequency into account despite environmental and cultivation factors. The European Earth Observation programme Copernicus provides large quantities of spatial and temporal high resolution data collected by a set of Sentinel satellites. The freely and openly accessible Sentinel-1 radar data form a valid basis for automated satellite and ground data processing methods to detect cutting events. These cutting frequencies are a fundamental information source for further analysis – the computation of grassland yields with different model approaches. In this study we like to present our overall approach integrating and analyzing data of different sources. A comparison between two different automated data processing methods to detect cutting frequencies from radar satellite data in three different regions in Bavaria is included. The common statistical detection represents a robust and reliable way by analysing time series of Sentinel-1 radar images of the same acquisition geometry with time intervals of 6 days. In contrast, machine learning techniques offer the opportunity to increase the accuracy and limit cutting dates to more precise time intervals.
Grasslands are among the largest ecosystems worldwide and according to the FAO they contribute to the livelihoods of more than 800 million people. Harvest dates and frequency can be utilised for an improved estimation of grassland yields.
In the presented project a highly automatised methodology for detecting harvest dates and frequency using SARamplitude data was developed based on an amplitude change detection techniques. This was achieved by evaluating spatial statistics over field boundaries provided by the European Integrated Administration and Control System (IACS) to identify changes between pre- and post-harvest acquisitions. The combination of this method with a grassland yield model will result in more reliable and regional-wide numbers of grassland yields. In our contribution we will focus on SAR-remote sensing for monitoring harvest frequencies, discuss the requirements concerning the acquisition system, present the technical approach and analyse the verified results.
In terms of the acquisition system a high temporal acquisition rate is required, which is generally met by using SARsatellite constellations providing a revisit time of few days. COSMO-SkyMed data were utilised for the pilot study for developing and prototyping a monitoring system. Subsequently the approach was adapted to the use of the C-Band system Sentinel-1A becoming fully operational with the availability of Sentinal-1B.
The study area is situated northeast of Munich, Germany, extending to an area of approx. 40km to 40km and covering major verification sites and in-situ data provided by research farms or continuously surveyed in-situ campaigns. An extended time series of SAR data was collected during the cultivation and vegetation cycles between March 2014 and March 2016. All data were processed and harmonised in a GIS database to be analysed and verified according to corresponding in-situ data.
SAR systems are recently used to generate robust and projectable information about maritime traffic, ice extent and geohazards. By utilising multiple SAR satellites dynamic information can be derived at variable temporal scales. Therefore acquisition systems and processing techniques become a key issue which is requested to work in a robust and efficient way. This paper will present generalized concepts for a monitoring approach that address unmatched or interferometric acquisitions. Its goal is to show the potential of increasing the acquisition rate but also to illustrate limitations resulting from the specific monitoring schemes and their combination. The paper will visualise practical examples derived from realized studies and projects. Finally we can conclude that an agile multi satellite and multi-mode SAR system, such as COSMO-SkyMed, is well suited to monitor to dynamic phenomena on the earth’s surface. The practicability needs to be discussed in detail case by case related to the real world requirements.