USE OF SATELLITE IMAGERY FOR ESTIMATING SOWING DATES OF WHEAT IN UTTARANCHAL, INDIA Ajeet S Nain 1 , K. Christian Kersebaum 1 and Vinay K Dadhwal 2 1 Centre for Agricultural Landscape and Land Use Research Müncheberg, Germany 2 Indian Institute of Remote Sensing (IIRS) Dehradun, India nain_ajeet@rediffmail.com , ajeet@zalf.de ABSTRACT The information on sowing dates of crop is an important step for the forecast of the crop yield. Remotely sensed data can be a valuable tool when crop models are applied in a GIS environment for simulating regional behavior of crop. An attempt has been made in the present study to derive sowing information using satellite-based imageries. LISS-III (Linear Imaging Self-Scanning) imageries of Indian Remote sensing Satellite (IRS) containing information on Agriculture-farm of GBPUA&T, Pantnagar, India, have been acquired for the period of 1996-97 to 1999-2000. A spectral profile of active vegetative growth stage of wheat was generated and regressed with DAS (days after sowing). A Normalized Difference Vegetation Index (NDVI) based model was developed for estimating sowing dates of wheat. The data of 24 plots were used for obtaining coefficients for exponential based model, while performance of the model was tested for independent 12 plots. Analysis showed that model could estimate sowing date with a minimum difference of 1 day and maximum difference of 24 days with an average deviation of 8 days. The Standard Error of the Estimate was 9.22 for the estimates made after an average of 78 DAS for the all plots, while it was 10.24 in the calibration period and 6.42 in the prediction period. KEY WORDS Modelling, Remote Sensing, Agriculture, Wheat 1. Introduction Remote Sensing is the science of acquiring and interpreting information about the earth’s environment from measurements made without physical contact with the object being studied [1]. The vegetation indices are specific combinations of spectral responses in different wavelength bands, which emphasize a particular feature of the vegetation. The vegetation indices are quantitative measurements indicating the vigor of vegetation [2]. They show better sensitivity than individual spectral bands for the detection of biomass [3]. The application of remote sensing for crop identification and crop yields assessment is being done for quite long time. In last two decades some efforts have been made to link space borne information with process oriented dynamic crop simulation model. Moran et al. [4] evaluated the application of remote sensing in precision crop management and recommended that future work should be focused on assimilating remotely sensed information into existing decision support systems (DSS). Barnes et al. [5] has listed three major applications of remotely sensed data for crop models: a) Aid in the definition of zones that require simulation as independent areas; b) Provide validation data on crop development, canopy density, and evapotranspiration; and c) Serve as a direct input to the model. There are several options for integration of remotely sensed data and crop model predictions, which include: a) Direct comparison of model predictions and remotely sensed estimates [6]; b) Iterative adjustment of the crop model's genetic parameters and initial conditions by comparing the model's predictions to remotely sensed estimates of ET and LAI [7]; c) Forcing the crop model's predictions to match remotely sensed estimates of actual field conditions [8]; and d) Using radiative transfer models so that satellite reflectance data can be directly compared to a crop model's predictions [9]. The large area (regional) crop yield estimation requires integration of crop simulation model in GIS (Geographical Information System) environment [10, 11, & 12]. The major limitation of this procedure is the large amount of information needed on crop distribution and management practices. Efforts have been made in the past to account for the spatial variability in crop by using remote sensing data. The most efforts were focused on iterative adjustment of the crop models genetic parameters and initial conditions that affect leaf growth and senescence so that simulated LAI matches estimates of LAI determined with remotely sensed observations of crop canopy reflectance [13]. The major spatial variability in crop at any given time is mainly due to the varying time of sowing. Thus, iterative adjustment of crop model parameters and initial conditions, some time miss the important information of sowing and lead to wrong estimates of crop yield. The information on sowing can be generated using remote 432-051 139