By Aleksandar Dujakovic | Research Assistant | BOKU
Remote sensing of vegetation is the process of monitoring the physical characteristics of vegetation by measuring its reflected and emitted radiation at a distance. By utilizing data from satellite, airplanes, and unmanned aerial vehicles (UAVs), remote sensing enables the collection of spectral, spatial, and temporal information to study vegetation health, productivity, and changes over large areas.
The STELLA Project
One innovative application of remote sensing is the detection and monitoring of crop diseases. A pest detection system will be part of the STELLA Pest Surveillance System (PSS) that will aim to timely and accurately detect pests for both early (not visible to the human eye) and visible symptoms. This system will leverage remotely piloted aerial systems (RPAS or UAV), remote and proximal sensing, citizen science, and traps.
STELLA PSS is being deployed in 6 Use Case Pilots (UCPs) (1 per crop) in 5 countries. UCPs have been selected to match the challenges of monitoring and suggesting phytosanitary measures for regulated pests, and they represent real-life environments with various contexts in terms of climate, location, disease type and farming system/forestry. Quantitative and qualitative information on pest and disease occurrence will be collected in the UCPs and used to train Machine Learning models. These models can use data from various sources to timely detect pest infestations in crops, orchards or forests.
How Satellites Detect Plant Diseases
Satellites, for instance, capture data in multiple parts of the electromagnetic spectrum, including visible, infrared, and thermal wavelengths. These data are based on the reflectance of energy from Earth’s surface in specific spectral bands. Reflectance depends on a material’s properties. Therefore, each material has a unique spectral signature. Changes in the infested plants (e.g., plant’s structure, water content, pigments) affect their spectral signatures. Satellites can be used in detection and monitoring of diseases in case a disease induces spectral changes that can be detected by a specific sensor or systems of sensors. This technology provides non-contact, spatially continuous monitoring of plant health, efficiently identifying stress indicators caused by pests.
Mapping Pest Outbreaks with Satellite Models
One of the key advantages of satellite technology is its ability to cover vast areas efficiently. This capability enables growers to assess pest distribution across different regions, providing crucial insights into the extent of infestations and informing targeted pest control strategies. For instance, satellites can monitor multiple fields simultaneously, identifying hotspots of pest activity that require immediate attention. For example, satellite technology enables calculation of various vegetation indices, such as the Normalized Difference Vegetation Index (NDVI), which indicates crop health and helps in identifying stress factors like pests or diseases.
Another advantage of satellites is the ability to capture images over fixed intervals (e.g., every 5, 10 days) of a given area, providing continuous monitoring of vegetation health and changes. Satellite data archives provide access to historical Earth observation imagery, which enables us to analyse changes over time and recognize trends and anomalies. With the knowledge of previous pest outbreaks, this can be useful to identify factors and changes in crops that may precede outbreaks.
In the STELLA project, satellite data will be used to develop pest detection models, analyse vegetation indices to identify areas of stress that may indicate pest infestations, track changes over time, analyse historical reflectance data and identify hotspots and assess the spread of infestations.
While satellite images may lack the precision for detailed pest and disease monitoring, they facilitate the quick detection of affected areas and the assessment of their spread. Overall, integration of remote sensing and machine learning promotes resource conservation by enabling precise application of pesticides and fertilizers, reducing waste and environmental impact.
Machine Learning Application
Machine learning (ML) and deep learning (DL) algorithms have enhanced the precision of pest detection through reliance on modern, technology-driven approaches. ML models have been used to analyse data on diseases and relationship between the spectral reflectance and disease occurrence and intensity. These algorithms automatically learn complex feature representations from large datasets. ML and DL models are trained to recognize patterns associated with healthy and diseased plants. Supervised learning techniques, where the model is trained on labelled data, are commonly used.
Types of Satellite Data
Different types of satellite data come with various spatial, spectral, and temporal resolutions, enabling monitoring tailored to specific needs.
- Multispectral satellites can capture data across multiple spectral bands, typically spanning the visible, near infrared, and shortwave infrared regions of the electromagnetic spectrum. Sentinel-2 satellites, for example, have been extensively used for crop monitoring. With two satellites covering the entire Earth every five days, they offer 13 spectral bands of 10 to 60 m spatial resolution, providing data that can be utilized in pest detection.
- Hyperspectral satellites, such as PRISMA or EnMAP, consisting of numerous narrow spectral bands, have the potential to detect diseases by capturing subtle changes in the spectral profile of plants that may be caused by pests and diseases.
- High spatial resolution satellites (e.g., WorldView, Pleiades Neo) provide potential of more precise identification of individual infested plants and mapping of affected areas. With pixel sizes of 30 cm or smaller, these satellites can offer detailed imagery that is crucial for pinpointing specific problem areas within a field.
Satellite remote sensing and its integration with machine learning give an opportunity to analyse large datasets and patterns that support pest and disease management strategies. As agriculture faces increasing challenges from climate change and evolving pest dynamics, the use of satellite data in systems like STELLA PSS can help farmers and researchers manage pests more effectively, promoting healthier crops and sustainable farming practices.