Geospatial Technology For Acreage Estimation Of Natural Rubber

Natural rubber (Hevea brasiliensis) is a critical industrial raw material for the rubber goods manufacturing industry that is central to economic growth. Natural rubber (NR) being an industrial raw material, its consumption increases with growth in industrial activity and economic growth of the country. India is produces almost enough NR to meet its entire industrial requirement. Presently, the production of NR and its consumption almost match in the country, but, it cannot be assumed that this balance will be maintained in the future. There is a strong likelihood that in near future, NR may become a scarce commodity, both domestically and internationally. It is in this context, new countries in Asia, Africa and South America are now coming forward to cultivate NR in large extents. Out of the nearly 0.74 Mha of natural rubber cultivated in the country as of 2012, about 0.53 Mha is present in the traditional NR cultivating region, which includes entire Kerala and Kanniyakumari district of Tamil Nadu.

Considering the deciduous nature of natural rubber, temporal LISS-III data prior to defoliation, during defoliation and refoliation stages were procured in the study for better spectral discrimination. About thirty seven RS-1 LISS-IV datasets have been acquired covering relatively wider acquisition period of 2008 to 2010 owing to smaller swath coverage of 24 km of the sensor. The satellite data analysis consisted of data preparation, ortho-rectification, image fusion, data normalization, NDVI generation, digital classification, visual interpretation, accuracy assessment and acreage estimation. Temporal LISS-III images were acquired at different periods under variable atmospheric conditions, solar illumination and view angles and thus, require radiometric normalization to remove radiometric distortions and make the images comparable. Initially the digital numbers were converted to Top of the Atmospheric radiance (TOA) and subsequently, relative radiometric normalization was carried out using Pseudo-Invariant Features (PIF). PIF corrected images were used for generating of NDVI. The Normalized Difference Vegetation Index (NDVI) is a numerical indicator that uses the reflectance in red and near-infrared bands of the electromagnetic spectrum which is sensitive to vegetation cover, vigour, biomass and condition. Temporal NDVI generated using Resourcesat-1 LISS-III data were used in hierarchical decision rule based classification for delineation of natural rubber plantations.

Satellite data acquisition period in relation to tree phenological stages plays a critical role in spectral discrimination of NR plantations from other vegetation types. Similarly, natural rubber showed typical NDVI profile was coinciding with the tree phenology. High NDVI values of NR plantations were noticed during October month which reduced significantly during February month due to defoliation and increased subsequently due to the refoliation. Mature and old NR plantations exhibited this trend while, younger NR plantations of 2-3 year age showed NDVI profile similar to fallow lands / scrublands due to poor canopy coverage and hence could not be delineated using LISS-III data alone. Hierarchical decision rule classifier using temporal NDVI has been used for preliminary classification of natural rubber plantations. FCC of LISS-III data along with the classified data showing spatial distribution of natural rubber in Tripura state. In the present study, final interpretation of natural rubber and potential wastelands was carried out using cartosat-1 and LISS-IV merged data. The image interpretation keys such as the tone, texture, terrain, shape and size were used for distinguishing of the NR plantations and potential wastelands with other land cover classes. Main focus was given for digitizing only these two classes, namely existing NR areas and cultivable / NR potential wastelands in the study area. Vegetation classes such as NR plantations, tea gardens and forests were the conflicting classes for interpretation.

The old NR plantations consisting of matured plantations of more than 10-15 years manifested as bright red colour with clear field boundaries while the tea gardens were characterized by light reddish green tone due to wider spacing exposing the underneath soil. Further, shade trees in tea gardens were clearly visible on the merged data giving rough appearance. The younger plantations of 4-5 years and above manifested as bright red color with smooth texture due lush green canopy which was completely covering the underground soil background. These two types of rubber plantations could be easily delineated using merged data based on the typical tone, texture and terrain conditions. It was observed that 2-3 years old rubber plantations in the plain areas with good canopy cover could be identified using merged data; however, interpretation of similar young plantations in the hilly and steep slopes was still difficult. The food crops cultivated in the study area were harvested during October – November and hence, appeared as fallow signatures. Typical and representative spectral signatures of wastelands noticed in the North, East and South Tripura which have been verified during the field visits. In the Western plain lands, most of the wastelands are characterized by undulating terrain with low to medium slope.

This type of land manifested in cyan or reddish tone with smooth texture depending on the land cover. The lands generally covered with lush green weeds/scrub showed pinkish appearance while land without vegetation appeared in cyan color. Generally, these lands were denuded forest without any large tree cover. Typical jhum lands were observed in the central and Eastern region of Tripura. In general, these jhum lands were occupied with lush green, medium tall weeds and sparsely distributed tree / cut tree trunks giving pinkish tone, smoother texture with dots of individual trees / tree trunks. This type of typical signature is wide spread especially in the central and Eastern regions of the state. Presently, these types of jhum lands are being converted to rubber plantations. Third type of wasteland category predominantly observed in the Eastern part of the state are large denuded lands with or without scattered trees. Typically, some of the reserved forest areas consisted of this category of wastelands. During the field visit it was observed that a significant expansion of rubber plantation area has occurred in the recent 1-2 years and most of the young rubber could have been interpreted as wastelands. This was mainly due to data acquisition period of RS-1 LISS-IV data which could not capture the recently cultivated rubber plantations. Hence, the thematic map was refined using RS-2 LISS-IV data to cover the recently cultivated rubber plantations. In spite of acquisition of latest data, discrimination of very young plantations at early stages of 1-2 years was limited.

This is one of the limitations of application of remote sensing data for perennial crops such as natural rubber which takes more than 2-3 years for significant establishment of the canopy covering the soil background. The mapping accuracy for the old matured rubber was 96.15 per cent indicating the potential of RS data for accurate delineation of matured rubber plantations. In contrast to this, the mapping accuracy for the younger rubber was 86.5 per cent. The lower mapping accuracy for the young rubber could be attributed to the non-clear manifestation of this class and often mixed with scrublands. Total natural rubber area estimated through remote sensing for the state was 48,037 ha. However, this area does not include the recently planted rubber plantations of about 1-2 years i.e. planting done during the planting seasons of June 2009-2011 and therefore, the total acreage under rubber plantations estimated through satellite data is an underestimate to this extent. It was observed that some of the potential wastelands were found inside the forest areas. Hence, a reserved forest area mask was generated from map obtained from Tripura State Forest Department as provided by RRII. About 20 per cent of the potential wastelands were observed within the reserved forest area which Rubber Board may not consider for promoting NR cultivation. About 22,947 ha area available outside the reserved forest area could be utilized for natural rubber cultivation in the state depending on the suitability. It may be noted that, it is highly likely that a part of the recently cultivated young rubber plantations (mostly up to 2 years old, and possibly to a smaller extent up to 3 years old) could have been inadvertently included in the category of wastelands, because the spectral signatures of very young rubber holdings and wastelands were not easily differentiated from each other. To this extent, available wasteland for NR cultivation is over-estimated in the study

Geospatial Technology For Acreage Estimation Of Natural Rubber