A Shortwave Infrared Modification to the Simple Ratio for LAI Retrieval in Boreal Forests: An Image and Model Analysis

In preparation for new satellite sensors, such as VEGEciated with vegetation, and as an input to ecosystem proTATION on SPOT-4 and the MODerate Resolution Imductivity models (Sellers et al., 1986; Bonan, 1993; Liu aging Spectrometer (MODIS), we investigate the potenet al., 1997). The need for LAI information over large tial of the shortwave infrared (SWIR) signal to improve areas has prompted investigations of the relationship beLeaf Area Index (LAI) retrieval in the boreal forests of tween ground-measured LAI and vegetation indices deCanada. Our study demonstrates that an empirical SWIR rived from satellite-measured reflectance (Franklin, 1986; modification to the simple ratio (SR) vegetation index, Spanner et al., 1990; Nemani et al., 1993; Chen and Cihlar, termed the reduced simple ratio (RSR), has the potential 1996; Fassnacht et al., 1997). The common approach has to unify deciduous and conifer species in LAI retrieval, been to correlate ground-measured LAI against the simshows increased sensitivity to LAI, and demonstrates an ple ratio (SR) (Jordan, 1969) or the normalized differimproved correlation with LAI in individual jack pine ence vegetation index (NDVI) (Deering, 1978). and black spruce canopies. The unification of deciduous In the boreal forests of Canada, studies investigating and conifer species suggests the possibility of not requirthe relationship between LAI and satellite-derived vegeing a cover type stratification prior to retrieving LAI intation indices have achieved r2 values between 0.38 and formation from remotely sensed data, and has impact 0.66 (Chen and Cihlar, 1996). However, there is often where no cover type information will be made or where considerable scatter evident in the relationship. Where a the mix of cover types within a pixel is unknown. We use pixel covers more than a single tree, the reflectance is a a geometric–optical canopy reflectance model to quantify combination of the forest canopy reflectance and the the potential variation in jack pine and black spruce cantype and amount of understory observed by the sensor opy reflectance caused by differences in background re(Robinove et al., 1981; Franklin, 1986). In this situation, flectance. The modeling study supports the results from there is no single LAI vs vegetation index curve, but the image analysis of the RSR showing increased sensitivrather a family of curves, each a function of canopy cloity to LAI and reducing background effects in these conisure and the reflectance characteristics of the backfer canopies. Elsevier Science Inc., 2000 ground (Spanner et al., 1990). The amount of background observed through a forest canopy is determined mainly by canopy closure INTRODUCTION (Franklin, 1986; Spanner et al., 1990). To compensate for differences in canopy closure and background reflecAccurate leaf area index (LAI) measurements are critical tance, studies have used shortwave infrared (SWIR) refor understanding biological and physical processes assoflectance to quantify canopy closure (Butera, 1986; Baret et al., 1988) and have used the SWIR reflectance as a modification to the NDVI (Nemani et al., 1993). Other * Environmental Monitoring Section, Canada Centre for Remote Sensing, Ottawa, Ontario, Canada vegetation indices including SAVI (Huete, 1988) and Address correspondence to Leonard Brown, Landcare Research, modifications of the SAVI (Qi et al., 1994), which have Private Bag 11 052, Palmerston North, New Zealand. E-mail: BrownL@ been developed to minimize the effect of background landcare.cri.nz Received 17 December 1998; revised 30 March 1999. observed may also be used; however, in the boreal forest