Detection and mapping of Japanese knotweed with the use of satellite images
Satellites are our silent assistants and are, among other things, helping us to better understand the events visible on the surface of the Earth. Remote sensing, as we call the technology of observing objects on the Earth’s surface, is already an established approach in forestry, agriculture, ecological applications etc., as remote sensing is a more cost-effective technique than field-survey methods.
Satellites are observing and detecting electromagnetic waves - or emitted solar radiation- from targets on the ground, which are then processed and analysed by experts. Different materials on the Earth’s surface reflect and absorb differently at particular wavelengths. Thus, the targets can be differentiated by their spectral reflectance signatures from remotely sensed images. Japanese knotweed also has its own spectral signature, which is very similar to the rest of green vegetation (it greatly absorbs waves in the red and blue part of the spectrum, while it reflects green and near infrared light), but it is still so different that it can be detected from other plant species with the use of complicated algorithms. Larger Japanese knotweed stands have a fairly typical occurrence pattern – they are usually present on the riverbanks, along roads, at construction sites and abandoned areas. On satellite images, the stands appear differently in winter or summer time, as the chlorophyll activity in the leaves varies greatly during the growing season. Furthermore, different stands might be in different growth stages during the same period; this depends on groundwater level, exposition to sunshine, etc. For easier detection of Japanese knotweed and inclusion of all previously mentioned time variables we use images acquired at different time points. The analysts call this time series satellite data analysis, which basically means that we observe the appearance of Japanese knotweed over time.
In addition to the time variable, the size of the unit that we detect is also important. A single image element - or a pixel – may not necessarily carry enough important information to determine where exactly invasive alien plant species are located. For a relevant analysis we therefore have to observe a sufficiently large area of Japanese knotweed stands. Most types of invasive alien species are present individually or on smaller land plots and are therefore not always visible at larger spatial scales on satellite images. Conversely, Japanese knotweed creates monoculture stands, which makes it easier to detect compared to other alien species. An alternative data source to satellite images are aerial photographs which have a much higher spatial resolution, but a much lower temporal resolution, as they are acquired approximately once every three years.
Once Japanese knotweed stands are detected from the image, the accuracy of the results is estimated. The correctness of the detected Japanese knotweed appearance sites is assessed using field reference data. From the invasive alien species management and control point of view, a large number of falsely detected polygons is a more desirable situation than a greater number of overlooked polygons.
Author of text and images: Urša Kanjir, Slovenian Centre of Excellence for Space Sciences and Technologies Space-SI