1858 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 44, NO. 7, JULY 2006 Leaf Area Index Retrieval Using IRS LISS-III Sensor Data and Validation of the MODIS LAI Product Over Central India Mehul R. Pandya, Raghavendra P. Singh, Karshan N. Chaudhari, Govind D. Bairagi, Rajesh Sharma, Vinay K. Dadhwal, and Jai Singh Parihar Abstract—This paper reports results on the LAI Retrieval and Validation Experiment (LRVE) that was conducted for two agricultural areas in Central India during the winter season of 2001–2002. The study aimed at relating field measurements of leaf area index (LAI) to spaceborne Indian Remote Sensing Satellite (IRS) Linear Imaging Self Scanning Sensor-III (LISS-III) data, preparation of site-level LAI maps, and validation of Moderate Resolution Imaging Spectroradiometer (MODIS) 1-km LAI global fields. Measurements of field-level LAI, aerosol optical thickness and water vapor were carried out on the day of LISS-III overpasses. Empirical models based on the site-specific LAI–veg- etation index relation were developed and used to generate 23-m resolution LAI maps for two sites (Indore and Bhopal) covering 30 km 30 km. These LAI images were degraded to 1-km spatial resolution and used for validation of the version 3 and 4 MODIS LAI products (MOD15A2). The results indicate a positive cor- relation between LAI derived from LISS-III data and MODIS data. However an overestimate by a factor of 1.6 to 2.5 in the version 3 MODIS product is observed with root mean square error (RMSE) ranging from 0.20 to 1.26. The factor of overestimation reduces significantly by 50% and RMSE by 40% when version 4 MODIS LAI was analyzed. The improvement in accuracy was observed to be associated with the change in algorithm path adopted for retrieving version 3 and 4 MODIS LAI. Analysis of the MODIS land cover product that is an input in the MODIS LAI retrieval algorithm indicated errors in assigning land cover classes for the study sites, which could be one of the sources of error in MODIS LAI product. Index Terms—Aerosol optical thickness (AOT), crops, Indian Remote Sensing satellite (IRS), leaf area index (LAI), Linear Imaging Self-Scanning Sensor (LISS), Moderate Resolution Imaging Spectroradiometer (MODIS), normalized difference vegetation index (NDVI), remote sensing, retrieval, validation. I. INTRODUCTION L EAF area index (LAI) is one of the most important at- tributes characterizing a canopy. LAI is a dimensionless biophysical variable used to quantify the single-sided vegeta- Manuscript received October 1, 2004; revised January 16, 2006. M. R. Pandya, R. P. Singh, K. N. Chaudhari, and J. S. Parihar are with the Space Applications Centre, Indian Space Research Organization, Ahmedabad 380015, India (e-mail: mrpandya@sac.isro.gov.in; rpsingh@sac.isro.gov.in; kishan@sac.isro.gov.in; jsparihar@sac.isro.gov.in). G. D. Bairagi is with the MP Remote Sensing Applications Centre, MPCOST, Bhopal 462003, India (e-mail: bairagigd@indiatimes.com). R. Sharma is the with Chhattisgarh Remote Sensing Applications Centre, Council of Science and Technology, Raipur 4920001, India (e-mail: rajesh_sharma1061@rediffmail.com). V. K. Dadhwal is with the Indian Institute of Remote Sensing, Dehradun- 248001, India (e-mail: vkdadhwal@iirs.gov.in). Digital Object Identifier 10.1109/TGRS.2006.876028 tion leaf area per unit of ground area in broadleaf canopies. It is a biophysical variable determining vegetation photosyn- thesis, transpiration, and the energy balance of canopies [1]. LAI and the fraction of absorbed photosynthetically active ra- diation (0.4–0.7 m) (fAPAR) are important surface attributes controlling water, carbon, and energy exchange between vege- tation and atmosphere [2]. LAI is not only an important driver of most ecosystem productivity models operating at landscape to global scales [3], but also an interacting component of gen- eral circulation models [4]. For effective use in such large-scale models, regional and global LAI must be available over a period of time. Field measurements of LAI, however, are cumbersome, time consuming, and impossible to obtain at the global scale, and in that respect, satellite remote sensing is the most effective means of estimating LAI global fields on a regular basis. As part of the Earth Observing System (EOS), the Terra (launched in December 1999) and Aqua (launched in May 2002) satellites carry the Moderate Resolution Imaging Spectrora- diometer (MODIS) along with a host of other advanced sensors. Algorithms have been developed to generate a number of land products from MODIS, including LAI/fAPAR [5] that have been made available by the Land Processes—Distributed Ac- tive Archive Center (LP-DAAC) for evaluation/validation and utilization. The MOD15 LAI and fAPAR are 1-km products pro- vided on a daily and eight-day basis. The validation of LAI global fields, i.e., assessment of uncertainty of remote-sensing-derived products by analytical comparison with reference data assumed to represent the target values [6], is a strong requirement and has been carried out for many sites in the U.S., Africa, Finland, France [7]–[10], and elsewhere.However, no results are available for India. An LAI Retrieval and Validation Experiment (LRVE) aiming at the development of remote-sensing-based site-specific vegetation index–LAI relationships and the validation of the MODIS LAI product was conducted at two sites of Central India during the wheat-growing season of 2001–2002. The experi- ment had three components: 1) measurements of field LAI and atmospheric properties (aerosol optical depth and water vapor); 2) generation of a high-resolution (23 m) LAI map from the Indian Remote Sensing Satellite (IRS) Linear Imaging Self Scanning Sensor-III (LISS-III) and field data; and 3) the generation of 1-km LAI maps and their comparison with the MODIS LAI product. II. MATERIALS AND METHODS This section describes: 1) field site description; 2) satellite data used in the study; 3) experimental measurements of LAI 0196-2892/$20.00 © 2006 IEEE