This is a dispatch from Theodore Endreny’s sabbatical in Italy….
The urban areas of our planet are an extremely popular living environment, and the simple act of maintaining or increasing tree cover can profoundly improve urban sustainability . The global urban area covers only 4% of our land, yet it contains 60% of our population. The metabolism of these areas is enormous, with each person needing between 1 and 10 hectares of non-urban area to support their resource consumption and waste generation . Urban trees can help reduce the ecological footprint of this metabolism and improve ecosystem carrying capacity by delivering an array of ecosystem services. These services include production and regulation, such as growing nutritious foods and maintaining a livable climate, as well as supporting and cultural services such as biodiversity and peace of mind . With urban areas containing such a high density of residents, an urban tree has the potential to improve the well-being of a large number of people. Our i-Tree research team develops tools for measuring the benefits of urban tree cover in order to help communities manage their sustainable well-being. In January 2016 we initiated a collaborative urban metabolism research effort with Italian scholars (led by Professor Sergio Ulgliati of Parthenope University) to collect data on tree cover and potential tree cover in a set of global urban areas, predict the associated ecosystem services, and investigate whether trends in tree cover and their services scale geographically or demographically. Our urban areas include several in Italy, such as Naples were the group is stationed, as well as the global megacities (Tokyo, Beijing, Istanbul, Cairo, London, New York, Manila, etc), defined as areas with at least 10 million human inhabitants.
This report presents the first step in our urban metabolism research, which was to select a method for determining the percent of tree cover, and potential tree cover, in our set of global urban areas. Although there is no international standard for land cover classification, most land cover maps limit classes to landscape units and fail to explicitly include trees in urban landscape units, limiting them to forested units . Ecological engineers will often use such landscape units, and make inferences about associated ecosystem structures (e.g., trees) and services (e.g., wood and fuel products, climate regulation) they need in their project designs (see Figure 1). However, in urban landscapes there is no explicit estimate of the tree cover and structure, and the assumption of zero tree cover ignores the substantial value contributed by existing urban trees .Our research method involved testing several products to estimate tree cover in Naples, Italy, defined by its political boundary to have an area of 118 km2. By testing several land cover classification products we could determine if there were differences in the estimated area between products, and then identify which product would be best for our classification of trees in the set of global urban areas. We considered the following products, NLCD, CORINE, MODIS, MAGLC, i-Tree Canopy, each explained below: In the US, the 30 m raster National Land Cover Dataset (NLCD) from LandSAT is a common land cover product that classifies urban areas as 21 – 24 (developed areas of low to high density), and forested areas as 41-43 (deciduous, evergreen, and mixed forests). In Europe, the polygon CORdination of INformation on the Environment (CORINE) land cover dataset is a common land cover product that classifies uses the class of artificial areas, and sub-classes of continuous or discontinuous urban fabric, which can include many sub-classes of residential cover, as well as several forest area classes such as agro-forest, broadleaf, coniferous, and mixed forests. Global datasets include the 500 m raster Moderate Resolution Imaging Spectroradiometer (MODIS) Land Cover product (MCD12Q1) that has a single urban class and five forest classes (evergreen needle and broad, deciduous needle and broad, and mixed). The Millennium Assessment global land cover (MAGLC) used similar data to create a 1000 m raster land cover map that inventoried landscape elements for estimation of associated ecosystem services, and it used 1 urban class called artificial cover, and several forest classes (e.g., broad leaf, needle leaf, mixed). An alternative method for land cover inventories involves a random survey, using photo-interpretation with i-Tree Canopy, to identify the fraction of inventoried points in discrete land cover classes, such as deciduous tree and evergreen tree to identify tree canopy, and other classes to discriminate between non-plantable and plantable areas, such as impervious area that is not-plantable, and impervious are that is potentially plantable. Outside of the United States, in place of NLCD we used the LandSAT based maps of forest cover, created by M.C. Hansen et al., and published in Science.
Our results clearly identified the i-Tree Canopy photo-interpretation product as the best estimate of tree cover, and the product we will use for future urban land cover characterization. The Naples area has a mixture of landscape units, including urban and forest, clearly seen in aerial photographs (Figure 2). Using i-Tree Canopy with a 2014 photo dataset the tree cover of Naples was estimated as 24.2% of the urban area, and potentially plantable urban area, such as sidewalks and plazas, could contribute another 20% of the total urban area to canopy cover. This estimate is based on a survey of 500 points, which takes approximately 2.5 hours to complete, and it had an uncertainty of 2%; 500 points is a recommended minimum for controlling the uncertainty in the estimate.
No other land cover product approached this i-Tree Canopy estimate of 24.2% forest cover. LandSAT (i.e., Hansen et al. equivalent of NLCD) estimated tree cover in Naples for 2014 as 6.3% of the total area, with this number accounting for all detected loss and gain in cover between 2000 and 2014. CORINE estimated forest cover from 2006 data at 7% of the Naples area (Figure 3),the MAGLC estimated forest cover as 4.5% from 2000 data (Figure 4), and MODIS estimated forest cover from 2013 data as 1.8% of the Naples area (Figure 5). Although the dates of each land cover product were different (2014 to 2000), this is not expected to explain the range in tree cover (24.2 to 1.8%). The MODIS and i-Tree Canopy products are closest in date (2013 and 2014), yet they capture the 22.4% range in variation between the maximum and minimum estimates. In general, based on land cover analysis by the Naples government, the vegetated area in Naples has decreased by 1.2% between 2011 and 2015, suggesting the 2014 data product used by i-Tree Canopy is a conservative estimate of tree cover. Given the uncertainty in this estimate is less than 2%, it is also considered the best estimate, and spatially it is the most precise estimate by providing a value for tree cover in the urban fabric.
Future work in this research area will involve applying the i-Tree Canopy tool to nearly 30 global cities. We are interested in having volunteers contribute to this work, and if you are interested please contact us (see below). We will then apply the i-Tree Canopy surveys of land cover types to estimate the existing and potential ecosystem services in these urban areas. This will include using the i-Tree Hydro tool to examine stormwater runoff and how trees reduce volumes and pollutant loads. For the i-Tree Hydro applications, the i-Tree Canopy photo-interpretation product was able to sub-classify each tree cover area by the type of land cover below the canopy, as either impervious or pervious. This sub-classification is important for simulation of urban water balances, in order to allow precipitation passing below the canopy to partition into soil infiltration or overland runoff. The i-Tree Canopy product identified shrub and herbaceous cover in the urban environment, as well as bare soil areas, and of course the impervious areas as potentially plantable or not plantable. The i-Tree Canopy tool could be used to provide data for regression models that estimate the tree cover for each urban class used in CORINE and NLCD (e.g., LandSAT product by Hansen et al.), perhaps implementing multiple-regression with additional explanatory variables such as geographic region or urban density. This would allow users of these CORINE and NLCD data products the opportunity to benefit from our estimates of urban tree cover.
Contact information: Dr. Theodore Endreny, firstname.lastname@example.org
Acknowledgement: The scholarly collaboration for this project has been supported by the U.S. – Italy Fulbright Commission and Parthenope University through a Fulbright Scholar grant to Theodore Endreny to serve as Distinguished Chair in Environmental Science at Parthenope University in Naples, Italy, and by the State University of New York College of Environmental Science and Forestry through a sabbatical leave to Theodore Endreny.