Monitoring the Urban Leaf Index From Space.

The United Nations estimates that almost three billion of the earth’s 6.5 illion human inhabitants live in urban settlements (UNIS 2004) Natural increase, changing economics, cultural practices, political policy, and environmental degradation are largely responsible for the ongoing population movement from rural regions to cities. This shift is not just a phenomenon of nations with developing economies. Between 1900 and 2000, the percentage of North American population living in urbanized areas increased from 50% to 75%. It is clear that cities are increasingly important as human habitat. Like any habitat, urban areas have both nonliving and living components that characterize them. A principal component of the urban environment is the urban forest (Turner, 2005). Urban forests are typically defined as woody vegetation in an urban area, usually limited to the trees in the city in contrast to grass and shrubs. The effect of the urban forest on urban heating, urban cooling, carbon sequestration, air and water pollution remediation, flooding abatement, noise reduction, human mental health, wildlife habitat, and pollution reduction have all been documented formany years (e.g., Cavanagh&Clemons, 2006; Deng, Song, Chen, & Rong, 2008; Er, Innes, Martin, & Klinkenberg, 2005; Fernandez-Juricic, 2000; Huang, Akbari, Taha, & Rosenfeld, 1987; Jensen, Boulton, & Harper, 2003; Jensen, Gatrell, Boulton, & Harper, 2004; Jensen & Hardin, 2005; Kielbaso & Cotrone, 1989; Lichtenberg & Hardie, 2007; Lo, Quattrochi, & Luvall,1997; McPherson,1998; Quattrochi & idd,1998; Ulrich, Simons, Losito, & Fiorito, 1991). The economic benefits of the urban forest are so documented. For example, the density of urban forest influences potential homebuyers/renterecisions and impacts real estate market structure (Anderson & Cordell,1985; Getz, Karow, & Kielbaso, 1982; Laverne & Winston-Geideman, 2003; Sydor, Newman, Bowker, & Cordell, 2003). Other benefits of the urban forest are less tangible but nonetheless contribute to the quality of life for urban residents (Michopoulos et al., 2007). These include the provision of aesthetically pleasing landscapes, resident screening and privacy, and recreation opportunities (Hull, 1992; Kennard, Putz, & Neiderhofer, 1996; Sheets & Manzer, 1991; Summit & Sommer, 1998; Tyrvainen & Vaananen, 1998). Unfortunately, urban expansion often leads to considerable loss of forests and these associated benefits (Jain & Jain, 2006).

Study area

Terre Haute is located in West-Central, Indiana, USA along the Wabash River (N 39 250, W 87250; Fig. 2). The city recorded a population of 69,614 in 2000 (US Census, 2000). Currently the city of Terre Haute has over 18,400 street trees, creating an average density of 53 trees per street mile (Hardin & Jensen, 2005). Terre Haute’s elected and civic leaders have made a conscious effort to effectively manage their urban forest. This is evident through a comprehensive city tree ordinance that restricts removal of trees in the city’s right-of-way. The ordinance is administered by a Tree Advisory Board that is staffed by both city residents and officials. In addition, Terre Haute enjoys a local nonprofit volunteer organization (Trees Incorporated) that actively supports the urban forest by organizing tree-planting programs and staffing education programs to promote the benefits of the urban forest. Trees Incorporated has been very successful n planting relatively large ‘‘street trees’’ throughout the city – especially in areas previously lacking adequate tree cover.

This study was conducted in a mid-size city in the Midwestern United States. However, this study’s methods should be applicable and its results should be similar in other urban areas throughout the world. Results generated from this study lead to the following specific conclusions: Overall, we think that urban LAI can be accurately estimated using AISAþ
hyperspectral data. We also think that otherhyperspectral sensors will enjoy similar successes. Model regression R2 values ranging from 0.27 to 0.73 were obtained from the various models.Features appearing most
frequently in the models included radiance at 0.727, 0.753, 0.848, 0.870, 0.900 and 0.917 mm.
 The best single variable predictor of LAI was the absolute difference in radiance between 0.777 and 0.673 mm. This represents the difference between the near infrared plateau and the red absorption band.  The best models performed well at low and intermediate LAI levels, but were unable to accurately model differences in LAI between 5.0 and 8.0. In these high LAI cases, themodel usually underestimated the true LAI of the canopy.  Underlying substrates have the potential to confound canopy LAI estimated values. This is especially a concern when the
shoulder-height and above LAI value approaches zero.  Although LAI can be estimated using these ethods and data, noise exists in the models. This noise is not easy to eliminate,
and is one of the limiting factors of our models and estimations.