Counting urban trees by satellite more accurate than manual method
University of Twente researchers develop technique for monitoring “urban forests”
Using satellite data, researchers at the University of Twente’s Faculty of Geo-Information Science and Earth Observation (ITC) have developed a highly automated method for counting the number of trees in a city and for pinpointing any changes. This has a range of potential uses, such as estimating the cost of maintaining green areas, and calculating a city’s CO2balance. This method of counting trees involves considerable time savings. The researchers are also convinced that it is much more accurate than traditional methods. One commented that “I trust this technique more than I do manual counting, which is the method currently used to monitor so-called urban forests.”
Local authorities in the Netherlands maintain a database containing details of the location and type of trees growing on public land. They do this for a variety of reasons, such as being able to make accurate estimates of the cost of maintaining green areas, and calculating a city’s CO2 balance (in connection with national reporting under the terms of the Kyoto Protocol). It is also useful for detecting the removal (legally or otherwise) of trees, and for monitoring the quality of the public space. At present, the location of trees and changes over time are determined manually. This approach suffers from the drawback that it is very time consuming, and therefore expensive. Moreover, the accuracy of such counts is not always consistent.
Researchers at the University of Twente have developed a highly automated method for counting the number of trees in a city and for spotting any changes in their number and crown diameter. This method is based on commercially available, high resolution satellite images, which are available for most parts of the planet. The technique is calibrated to identify trees growing on common land whose crowns are more than two metres in diameter. The method translates the data into a map showing the location of these tree crowns. This also shows changes over time (whether trees are still standing, whether new trees have been added, and whether there have been any significant changes in the size of the tree crowns). This is then double checked by a human operator. This double checking is relatively simple, as the technique also displays any inaccuracies in the detection of each tree. By this means, the map can be monitored and improved in a results-oriented way.
According to Prof. Alfred Stein, one of the researchers involved, not only does this method involve considerable time savings, it is also more accurate in practice. He commented that “I trust this technique more than I do manual counting, which is the method currently used to monitor so-called urban forests.” The method can be used in just about any urban area in the world, although it may be of more limited use in cities with large numbers of skyscrapers.”
In the future this method could also be used to monitor specific forests designated for the production of FSC timber, to check that the number of trees being felled does not exceed permitted limits.
Using this method, it currently takes about one working week to accurately interpret the data for a single city. The next step is to refine the method still further. To this end, the NEO company is working to make the system more automated, so that less manpower will be needed to interpret data. The system could be further improved by linking it to other geographic information resources and to laser altimetry maps.
This work was part of Juan Pablo Ardila Lopez’s doctoral research, which was carried out under the supervision of Prof. Alfred Stein, Dr Wietske Bijker, and Dr Valentyn Tolpekin of the department of Earth Observation Science (EOS) at the University of Twente’s Faculty of Geo-Information Science and Earth Observation (ITC). The research, which was carried out in the context of the Boom en Beeld (Tree and Image) project (a joint venture with the NEO company, the Alterra research institute, and the Faculty of Geo-Information Science and Earth Observation), is being funded by the Netherlands Space Office.
Note to the press: For further details, or an electronic version of the thesis entitled Object-based methods for mapping and monitoring of urban trees with multitemporal image analysis, please contact the University of Twente Science Information Officer, Joost Bruysters, +31 (0)53 4892773 / +31 (0)6 10488228.