Retrieval of Leaf Area Index in mountain grasslands in the Alps from MODIS satellite imageryRemote Sensing of Environment

About

Authors
Luca Pasolli, Sarah Asam, Mariapina Castelli, Lorenzo Bruzzone, Georg Wohlfahrt, Marc Zebisch, Claudia Notarnicola
Year
2015
DOI
10.1016/j.rse.2015.04.027
Subject
Computers in Earth Sciences / Soil Science / Geology

Text

ra zo mar

Received in revised form 12 April 2015

Accepted 21 April 2015

Available online 24 May 2015 1. Introduction time consuming and unfeasible for large scale monitoring over time. parametric and nontral vegetation indices elatively good perfore-, and biome-specific

Remote Sensing of Environment 165 (2015) 159–174

Contents lists available at ScienceDirect

Remote Sensing o j ourna l homepage: www.eAlthough several methods for ground based measurements of LAI exist (Weiss, Baret, Smith, Jonckheere, & Coppin, 2004), these are typically and their application is limited by the representativeness of the set of reference samples used for calibration (Meroni, Colombo, & Panigada, 2004;

Colombo, Bellingeri, Fasolini, & Marino, 2003). Another category of ap-water and energy fluxes (Running & Coughlan, 1988; Turner, Ollinger, &

Kimball, 2004), and which use the patterns and trends of LAI to force or adjust model predictions by means of assimilation techniques (Dorigo et al., 2007; Moulin, Bondeau, & Delecolle, 1998; Quaife et al., 2008). definition of empirical relationships by applying parametric regression methods on LAI and spec has been widely used due to its simplicity and r mance. However, such relationships are site-, timThe green Leaf Area Index (LAI), defined as half the surface area of green leaves per unit of ground area (Watson, 1947), is a requirement in a variety of ecological and agricultural applications (Sellers et al., 1997). The availability of spatially and temporally distributed information about LAI is of crucial importance for climate change studies (GCOS, 2006;

Sellers et al., 1996), as well as for ecosystemmodels that quantify carbon,

The exploitation of satellite earth observation imagery can overcome this limitation.

The retrieval of biophysical vegetation variables such as LAI from remote sensing data takes advantage of the causal relationship that exists between canopy characteristics and the electromagnetic radiation it reflects. To this end several algorithms have been proposed in the literature, which broadly fall into two main categories of approaches. The⁎ Corresponding author.

E-mail addresses: luca.pasolli@yahoo.it (L. Pasolli), sar mariapina.castelli@eurac.edu (M. Castelli), lorenzo.bruzzo georg.wohlfahrt@eurac.edu, georg.wohlfahrt@uibk.ac.at ( marc.zebisch@eurac.edu (M. Zebisch), claudia.notarnicola http://dx.doi.org/10.1016/j.rse.2015.04.027 0034-4257/© 2015 Elsevier Inc. All rights reserved.Keywords:

Leaf Area Index (LAI)

Biophysical parameter retrieval

Radiation transfer modeling

Moderate Resolution Imaging

Spectroradiometer (MODIS)

Mountain Grassland

AlpsImaging Spectroradiometer (MODIS) satellite imagery that has been specifically customized for mountain grasslands in the Alps. The main features of the proposed algorithm, which is based on the inversion of a radiative transfer model, are: i) a higher spatial resolution (250 m) with respect to the corresponding standard MODIS product and ii) tuning the model to the spectral characteristics of mountain grasslands.

To quantify the effects of the features of the proposed algorithm, the approach is first applied to a MODIS reflectance data time series from 2007 up-scaled to a 1 km spatial resolution for better comparison with the standard

MODIS LAI product. In the next step, the benefit of the higher spatial resolution is assessed by applying the algorithm to a series of MODIS satellite images with a spatial resolution of 250 m acquired over the central

Alps in the period 2005–2007. LAI estimates were validated for both temporal consistency and accuracy using ground measurement time series collected at three different study sites in the investigated area. The results obtained demonstrate the capability of the proposed algorithm to follow the expected temporal and range dynamics of LAI in this challenging environment, showing an overall RMSE accuracy of 1.68 (m2/m2). This approach thus opens a promising avenue for the exploitation of moderate resolution satellite data for novel and more accurate monitoring studies at a regional scale in mountain environments. © 2015 Elsevier Inc. All rights reserved.Article history:

Received 15 January 2015

This paper presents an improved algorithm for the retrieval of Leaf Area Index (LAI) from Moderate Resolutiona b s t r a c ta r t i c l e i n f oRetrieval of Leaf Area Index in mountain g

MODIS satellite imagery

Luca Pasolli a, Sarah Asam b, Mariapina Castelli b, Loren

Marc Zebisch b, Claudia Notarnicola b,⁎ a Informatica Trentina, via G. Gilli 2, 38121 Trento, Italy b Institute for Applied Remote Sensing, Eurac Research, Viale Druso, 1, 39100 Bolzano Italy c Department of Information Engineering and Computer Science, University of Trento, Via Som d Institute of Ecology, University of Innsbruck, Sternwartestr. 15, 6020 Innsbruck Austriaah.asam@eurac.edu (S. Asam), ne@ing.unitn.it (L. Bruzzone),

G. Wohlfahrt), @eurac.edu (C. Notarnicola).sslands in the Alps from

Bruzzone c, Georg Wohlfahrt b,d, ive, 14, 38121 Trento Italy f Environment l sev ie r .com/ locate / rseproaches is the inversion of canopy radiative transfer models. Radiative transfer models are based on a rigorous physical description of the interactions between electromagnetic radiation, canopy elements and the underlying soil surface. They can simulate a great variety of conditions in terms of vegetation type and characteristics as well as sensor acquisition 160 L. Pasolli et al. / Remote Sensing of Environment 165 (2015) 159–174geometry, thus typically ensuring a higher generalization and portability with respect to empiricalmodels. However, the inversion process is computationally demanding and ill-posed which may lead to instability in the results (Combal et al., 2002a). To cope with these issues, prior information on the investigated area can be exploited to better constrain the inversion process (Combal, Baret, &Weiss, 2002b; Lavergne et al., 2007).

In the domain of remotely sensed images, medium to coarse resolution satellite imagery with their high temporal frequency and global coverage offer a valuable database for modeling and monitoring purposes. The Moderate Resolution Imaging Spectroradiometer (MODIS) sensor on board the AQUA and TERRA satellites is probably the most used medium resolution systemwithin the remote sensing community (Justice et al., 2002). A suite of high-level standard products, including global LAI maps (the MODIS 15A2 and 15A3 products), has been developed by theMODIS Land Discipline Groups and is freely available to the user community.