Enabling the public sector to save money, innovate and make more effective policy decisions by using space technology and data

Case Study: Rezatec – Space Applications for Precision Plant Health Information, Response & Evaluation (SAPPHIRE) (Phase 2)

General Information

  • Provider: Rezatec
  • Technology utilised: Earth observation (EO) Satellite Navigation
  • Thematic area: Environment Local Authorities and Devolved Administrations
  • End user(s): Central Government, Defra, Forestry Commission
  • Website: www.rezatec.com


  • 2017

This project is a phase 2  follow on project from Rezatec’s phase 1 project.

Forest Research is the UK’s primary research organisation for forestry / woodlands and is acutely aware of the difficulties associated with obtaining accurate tree distribution by species, especially broadleaf species. Its tree Health management teams lack contemporary data of tree species distribution and abundance needed to predict the progression of forest pathogens. This problem is particularly acute outside of managed forest estates (private land, network corridors etc.).

The primary objective was to demonstrate Rezatec’s ability to map the spatial distribution and abundance of 3 target deciduous species, Ash (AH), Oak (OK) and Sweet Chestnut (SC), selected for their susceptibility to recent prominent pests and pathogens in the UK e.g Phytophthora Ramorum, Sudden Oak Death and SC Blight. The methodological improvement in Phase 2 was to incorporate data not captured in the Area of Interest and apply it to a managed woodland, and the surrounding area, to map the known extent (based on managed woodland
for validation purposes) and the deciduous species that occur naturally outside of these well documented areas, but play a significant role in the potential spread of pests and pathogens.

Rezatec addressed this need with EO-derived analytics for tree species classification across both public (managed), and private woodlands, as well as phenology characterisation for stress detection. The solution derived utilised national ground-based data, and freely available (open-source) EO datasets, to allow for a scalable approach for consistency and repeatability (e.g. annual map updates of species distribution or monthly updates of stress monitoring). 

Satellite Enabled Solution:

  • Species distribution mapping utilised Sentinel-1 C-band SAR data and optical multi-spectral data
  • Ground survey data from the UK (specifically tree plot data) was collated and utilised to train a machine learning classification algorithm, which is trained on open-source freely available EO datasets, specifically optical (Sentinel-2) and radar (Sentinel-1)
  • The trained model is then applied to an area of interest, on the available EO data for the region, and a tree species classification at a 10x10 metre resolution is produced which depicts the distribution and abundance of each species (specifically each pixel contains the modelled number of each target tree species) present within the pixel. Therefore, the spatial distribution and abundance is outputted, and delivered via the online portal for end-user analysis.
  • The data utilised allows for blanket coverage within the imagery extents (excluding cloud-cover), so all trees present within the imagery can be classified (if they are one of the target species), meaning the method can be used to classify trees within managed woodlands, but also private trees in privately managed woodland or alone network corridors for example.
  • Extraction of baseline phenological behaviour patterns from Sentinel-1 time-series for key broadleaved species and development of heath monitoring system
  • The 6 day revisit period of Sentinel-1 C-band SAR backscatter is utilised to extract the mean annual phenological behaviour of ‘healthy’ trees of a target species e.g. Sweet Chestnut, and then this annual  ‘baseline‘ pattern is compared with subsequent observations to identify any significant and anomalous deviations which may be indicative of stress (potentially associated with pest / pathogen). This is conducted at a 20x20m pixel resolution, and the analysis can be conducted monthly, as part of an on-going monitoring system.

Forest Research and Defra have both actively engaged with Rezatec during the project, and are very encouraged by the method and results, due to the significant implications of the success of both products with regards to current known threats to UK forests from pests and pathogens, and known limitations of current survey methods for detecting and monitoring disease spread. Rezatec has already started discussions with both end-users as to how these results could be rolled out into both a larger extent for single baseline assessment of species distribution across the UK, but also the integration of continuous monitoring products into existing procedures. Rezatec will be continuing these engagements with public sector bodies, as well as the current success of similar methods applied in commercial coniferous forests with private sector end-users (e.g. John Clegg), where methods developed through SSGP Phase 1 resulted in a range of commercial products being developed in both the species mapping and health monitoring fields (pest / pathogen, windblow damage, and disturbance mapping being key).

Costs and Benefits:

The partners (Rezatec and Forest Research) are aiming to develop the application(s) further to reach a wider audience. The partners would envisage further investments in the service.

Considering the benefits outlines above in more detail, the models need more data and a wider range of issues to be modelled into the platform. For each new data set that is modelled, there are the associated analytics and also the associated human costs to build up a reliable database that can be rolled out elsewhere. Rezatec would like to increase its broadleaf footprint in the UK to make what is a fairly comprehensive conifer footprint in the UK. Once modelled and inputted, the benefits are the ability of offering a wider range of solutions that are specific to different areas of the country and being able to offer a fuller UK wide coverage for monitoring purposes.

It is not limited to the UK. Whilst the species and diseases may change in new territories, the methodologies will remain consistent and as long as we can receive the requisite data inputs to train the models, we will be able to increase the coverage footprint.

The savings for the client are clearly defined – early detection saving costs of mitigation. For Defra for example, as we build on the relationship, we will be able to provide them with an increasing baseline of data and as things are detected from a species perspective and then a disease perspective, Defra will be able to monitor more frequently. This will mean they detect issues sooner, and mitigation takes place earlier, saving a great deal of time, effort and money, as compared to a later detection.

The Sentinel datasets are freely-available, as is the software utilised in the method, making this developed method scalable and repeatable in a manner that significantly reduces the costs, time and effort currently required to achieve large scale mapping and monitoring using aerial and ground-based survey techniques. Current forest inventory techniques average at £44 per hectare, so the ability to roll this product out and provide a national baseline, as well as timely monthly updates on condition, would have a huge economic potential for the public sector teams who currently conduct these activities, increasing efficiency and the ability to predict pathogen spread, and mitigate loss, thereby creating a greater economic and environmental / social saving to forested landscapes.

Lessons Learned:

The ability to work with the End User, Forest Research and thus assist UK public sector woodland managers in more effectively monitoring and controlling the spread of tree pathogens.  Direct engagement with Forest Research has led to associated networking and engagement with the wider community, including Defra, allowing for relevant and constructive positive future collaborations for all parties. 

How Sentinel Earth Observation and standard survey ground-based data layers can be combined to create a landscape intelligence tool, focusing on depicting tree species distributions and their condition. The ability to do this across landscapes (managed and private) with freely available data has been a significant success of this project.

Next Steps:

Extension of tree species mapping to a broader scale to explore scalability and repeatability e.g. county-level or nation-wide maps, and monthly stress monitoring:

  • Given the open-source nature of the data and method, there is a significant potential to increase the spatial extent of the outputs to a much larger scale, outside of the current project boundaries. The project successfully demonstrated that the method can be applied across the imagery, not requiring independent methods for managed woodlands and private trees (e.g. within hedgerows, along network corridors etc). Ultimately this means the method can be applied nationally, on the prerequisite that sufficient training data is available to parameterise the classification model for the required (core) broadleaf species so as to provide a more accurate and complete baseline assessment.

Engaging with existing interested end-users (Defra, Natural England etc.) to roll-out the products nationally:

  • Both Forest Research and Defra have expressed interest in a national baseline map of certain broadleaf tree species, as this does not currently exists, and is essential for their tree / plant health teams who require timely data on the location of susceptible trees within known possible infection zones to allow appropriate mitigation actions. Rezatec hopes to continue further development with these partners, alongside other commercial activities with commercial forestry end-users who are already utilising Rezatec’s coniferous tree species mapping and health monitoring products.

Click on the following links to see Rezatec's presentation and video from Phase 1 of the project.