|
|
|
Agriculture / Soil
|
Effect of Time of Data Acquisition on Crop Identification in North -Eastern India using NOAA -AVHRR Data
S. Panigarhy S.S. Sharma and J.S. Parihar
Space Applicatiion Centre Ahmedabad 380053
Abstract
Performance of data acquired from morning pass of NOAA 12 and afternoon pass of NOAA 11 for land cover classification was evaluated for north each region of India lying beyond 850E longitude. It was observed that entropy of all the 3 bands viz. Visible NIR and the MIR were higher in NOAA 23 data. The total number of spectral vectors obtained from band 1, 2 and 3 of NOAA 12 was double the number of total vectors from NOAA 11 data. The mean digital number of various known sites showed a reduction of 10 - 35 % in band 1 and 2 from morning to afternoon pass. The computation of percent change in isolation showed that there was significant reduction in morning to afternoon pass time of NOAA in this region, particularly from 86E onwards in the month of October and November. The separability of various land cover classes in general and crop classes in particular were poor in afternoon pass data. The results strongly suggest that morning pass data of NOAA 12 is preferable for such studies particularly in kharif season for this region.
1. Introduction
Among the existing satellite borne earth observations sensors, the Advanced very High Resolution Radiometer (AVHRR) on board the NOAA/TIROS satellite has been found suitable for vegetation monitoring at regional level (Belward 1992). The advantages over the high spatial resolution systems like IRS, LANDSAT are the daily coverage of an area, which largely overcomes of data availability due to cloud cover, a major problem in tropical regions (Currey et al. 1987). This is particularly more useful for studies in Indian subcontinent where number of cloudy days are very high during monsoon season (Rao & Lalitha a1980, Panigrahy et al., 1992). In India AVHRR data has been used to derive useful information on crop area, yield and vegetation condition (Dubey et al. 1991 Panigrahy et al., 1992, Potdar, 1990. However, all these studies in India and elsewhere have used AVHHR data acquired during afternoon pass of NOAA 7, 9 and 11. (Justice 1990, Malingreau et al. 1989, Quaramby et al., 1992). Off numbered satellite are preferred since their overpass are during noon or early noon. The even number satellites 8, 10, 12 are morning polar orbiter with nominal equator crossing time of 7.30 IST, node descending (day time coverage). Over India, the day time passes are around 15.30 hrs. IST for NOAA 11 and 07.30 hrs for NOAA 12.
The present work analyses the comparative performance of morning and afternoon pass data for land cover classification in general and agriculture crop classification particularly in north eastern region of India.
2. Study Area
The study area lies from 820 E to 970 E and 220 N to 280 N comprising states of Assam, West Bengal, Meghalaya and Sikkim. This region harbors a wide range of forests on undulating and hilly terrain's. Rice is the signal most dominant crop which occupies more than 90% of the agricultural area. On the basis of cultural Practices prevailing in this area, there are different rice crops whose growth characteristics differ significantly. The irrigated rice of Burdwan, Hughli are very different rice crops whose growth characteristics differ significantly. The irrigated rice of Burdwan , Hughli are very different from the low land rice of Murshidabad, Noida, the rained rice of cooch Bihar and dryland rice of Purulia in West Bengal. Similarly in Assam there are low land rice to floating bao rice grown in waterlogged areas. The average size of fields in this area is very small (<0.2 ha). However, since rice is grown in large number of contiguous fields, training sites in AVHRR data are available. The crop is grown from July to November month, thus mid October data corresponds to anthesis to grainfilling stage of the crop. The other major land cover found in this area are mangroves (Sunder-bans), tea gardens and water logged areas, besides a range of forest sub classes.
3. Data Based and Analysis Approach
NOAA-AVHRR data of visible (0.55 -0.7m m) NIR (0.71 - 0.98mm) and thermal MIR(3.55 - 3.93mm) were used in the study, as there are the channels extensively used for vegetation studies. The orbital calendar of N?OAA 11 and 12 satellites were the overhead (nadir view) pass of the area was available same day for morning and afternoon passes. Data from two cloud free dates e,g October 6 and October 23 for both the satellites NOAA 11 ad 12 were used in this study. The morning pass was at 0725 hrs IST in descending code and afternoon pass was at 1540 hrs IST in ascending node. The data received at receiving station at the space Applications Centre, Ahmedabad was used. Analysis was done using DIPIX Aries III image processing system.
4. Analysis Approach
The analysis was done in following steps :
-
Extraction of study area and geometric correction using map-to-image transformation model developed using ground control points and overlaying of state boundaries on the image data.
- Image-to-image registration of morning and afternoon data using a third degree polynomial transformation and cubic convolution resampling.
- Selection of prominent surface features viz. dense forest, mangrove, sea, river, urban (Calcutta) and extraction of windows romcommon location in NOAA 11 and NOAA 12 images.
- Selection of training class sites for various land cover classes and crop sub-classes.
- Generation of training statistics and confusion matrix of training site pixels, classifieds using he maximum likelihood classifier (MLC).
- Generation of normalised difference vegetation inex of vegetation classes.
- Computing solar zenith and azimuth angles as a function of position on the Earth's surface, month, day, year and time of satellite pass.
|
|
|
|
|
|
|
