Services - Invasive Species and Plant Disease

Cardiff County Council has suggested that over 800 hectares of Wales had been invaded by Japanese Knotweed (JPW) by the end of 2010. JPW remains a significant issue for local authorities due to the costs associated with controlling and stopping its spread. Estimates place the cost of controlling Japanese Knotweed at around £1.56 billion with total annual costs of JKW to the British economy estimated at around £180 million (Defra 2010).

Building on innovative work initially developed for Caerphilly County Council, Environment Systems has now undertaken a classification of the invasive Japanese Knotweed species, across 13 areas in Wales, encompassing some of the main urban and fluvial zones of the country. This work was carried out on behalf of the Environment Agency of Wales.

In this project colour-infrared imagery (CIR) and LiDAR dating from 2006 were used as the principle inputs to map the distribution of JKW. The principle is that the two data sources should be contemporaneous; however, this basic assumption is unlikely to be valid given that the CIR imagery alone is a patchwork of imagery acquired on different dates. In addition there was no LiDAR data available for some parts of the survey areas.

Despite these data limitations work in these test areas has proved the efficacy of the methodology developed to automatically classify JPW using advanced remote sensing techniques.

The availability of LiDAR data, both a first surface model (DSM) and a Digital Terrain Model (DTM) was crucial to the classification. By subtracting the DTM from the DSM it is possible to ascertain the height of objects on the surface, both natural and man-made (buildings etc.).

The LiDAR and CIR imagery approach to processing enables the data to be broken down into homogenous units which were then classified using a sophisticated rules approach, based on the data itself and knowledge of the local ecology. Numerically derived thresholds were established based on noted differences and changes within the imagery, and thus used progressively to produce a classification.

The variability of the available data meant that it was necessary to develop a hierarchical classification scheme where the best areas determined by the system were classified as having a ‘high probability’ of JKW presence, with a further two categories of ‘medium’ and ‘low’ probability.