Conclusions and Recommendations
1.1.1. Research Questions
a. Can tea plants be realistically identified automatically from satellite images like Landsat, LISS III and ASTER?
The prime objective of this question is to identify the tea plants or the tea patches from the LANDSAT (of spatial resolution 30m), LISS III (of spatial resolution 23.5m) and ASTER data (of spatial resolution 15m). From these three datasets the tea patches were identified.
The LANDSAT image showed the diseased, moderately affected and healthy tea patches. Band 2 detects the green reflectance from the healthy tea patches. Band 3 detects the chlorophyll absorption in tea. Band 4 data is ideal for detecting high peaks in healthy green tea patches and for detecting water-land interfaces.
From the LISS III image, the tea patches could be easily identified. Both affected and unaffected patches could be observed from the LISS III image. The Band 2 (Green) is used for tea patches discrimination. Band 3 (Red) showed the chlorophyll absorption of the tea patches. Similarly Band 4 (NIR) shows high reflectance for healthy tea patches and is useful for green biomass estimation and crop vigour.
From the ASTER image, the tea patches could be very well distinguished between healthy, moderately affected and diseased patches. The affected and the unaffected tea patches were very prominently visible. As compared to the LANDSAT and LISS III image, the ASTER image gave much better results. The ASTER data gave the relative spectral reflectance and emissivity, surface radiance, temperature, brightness temperature-at-sensor, and digital elevation models. Band 1 (Green) distinctly discriminate the tea patches. Band 2 (Red) showed the chlorophyll absorption of the two patches. Band 3 (NIR) showed the peak reflectance for healthy tea, moderately affected tea and diseased tea patches. The three classes could be very well identified from the ASTER image.
Once the tea patches were identified, they were verified on the ground and the shift in the disease was observed.
b. Can we understand the features observed on remote sensing images from field conditions in contexts with the tea planters/farmers?
Once the features were identified in the images, they are then verified on the field. The managers were interviewed about the present status of their tea gardens. The information regarding the yield, management practices, as well as the diseases of the gardens were collected. Further information regarding the time of irrigation and frequency of irrigation were also collected. This information were further compared with the ground information. From the field the ground LAI values were collected using a plant canopy analyzer based on healthy, moderately healthy and diseased patches from each divisions of the garden. The GPS point at that particular LAI was also collected.
c. What is the role of texture and tone in determining tea bush health?
The textural analysis was carried out to identify the three patches of tea such as healthy, moderately healthy and diseased tea. A comparison was made between the classified maps and the texture images. In this study, Grey Level Co-occurrence Matrix (GLCM) was used for studying the texture of tea patches. Texture images were generated taking into account all the texture parameters. These parameters were then analyzed band wise and the parameters giving good texture result were selected. The mean provided better results than the other parameters. For contrast, dissimilarity and variance, however the boundary of the tea patches were distinctly visible. Remaining parameters like the second’s moment, homogeneity, correlation and entropy could not produce any good results and so these parameters had to be discarded.
Once the texture images were generated, then the thresholding of the images was done at different ranges to generate the classified images. In case of ASTER it was observed that the thresholding could be given to all the three patches of tea (Healthy, Moderately Healthy and Diseased Tea) and they could be well separated out. But in case of LANDSAT and LISS III it was observed that the tea patches could be thresholded at two levels only as healthy and moderately healthy because when the thresholding was carried out at three levels, inter class mixing occurred. Mixing was such that a healthy patch was thresholded as diseased patch or a diseased patch as healthy or moderately healthy patches. It was also observed that proper thresholding was possible only with the mean parameter. For the remaining parameters thresholding was a problem because of excessive mixing.
We could not conclude that texture analysis could be used for studying the bush health of tea plantations and whether the bush health is healthy, moderately healthy or diseased that could be well separated. It was observed that to a certain extent GLCM technique could separate out the affected and non-affected tea patches. But the best texture analysis could then be judged when all the texture techniques are applied to this study and their results are compared. Other then this, the analysis should be carried out in high resolution images like IKONOS, LISS IV, etc. The tonal variations also play an important role for visual interpretation of images into healthy or moderately healthy or diseased tea patches. The different bands give different reflectance values by means of which one can easily identify the different tea patches.
d. How is tea bush health affecting the production or yield of tea plantations?
Present tea scenario is that there is a declining trend in the production of the gardens. There are many factors regarding the declining of tea yield. The problems started in 1999 due to draught leading to sharp production cuts. There was slight improvement in 2000 but again the prices drop down. Many factors were responsible for such sharp cut in production. Biggest problem that most of the tea gardens were facing was the infestations from pests and diseases. But among all the pests and diseases it was observed that Red Spider Mite attack and Helopeltis was creating havoc in most of the gardens resulting in lower production of tea. These two diseases mostly occur during June when the temperature and the humidity are high and accompanied by heavy monsoon showers. This type of weather is highly favourable for the occurrence of these two diseases. The biggest problem with the two infestations was that they spread very rapidly and controlling them is a big concerned for most of the gardens. Though pest control measures have been adopted still the total control of the diseases could not be achieved till date. There were also instances where due to excessive application of pesticides the quality of tea has deteriorated. Production in Assam in 2001 was low as compared to the national average. During the year, prices further declined. Export also dropped by 27 million Kgs and Assam could export only 18 million Kgs of tea. During the field trip when the managers were interviewed it was seen that the severity of disease was more from 2003 and the management were finding it difficult to control the disease due to it’s rapid shift.
When the field was surveyed, certain tea patches could be seen with a severe infestation of the two diseases. The management had already stopped plucking from those areas. The only possibility is replanting of new tea plants. Though a garden needs to produce an average of 3000 ton annually, almost all gardens produced below the average yield. The yield trends for some of the gardens have been shown in the appendix. Using the texture analysis affected and unaffected tea patches were identified and analyzed. It was also observed that the LAI collected from the field gave lower values for the affected patches and higher for the healthy patches. When the LAI values were compared with the texture images it clearly showed the affected and unaffected tea patches. The results were further verified with the yield of the gardens.
e. How can the results obtained be helpful in overcoming the problems of conventional tea practicing methods?
Using the optical remotely sensed data it is possible to observe the affected tea patches. Onscreen visual interpretation is done to assess the affected and non-affected tea patches. The healthy patches were seen as bright red colour while the affected patches were seen as brownish red or dark brown in colour. Identification through the imageries will help the management to take necessary steps the earliest. Management can now identify their area of interest and then take the necessary measures accordingly. It will help the management to identify the spot and go there directly and assess the affect instead of surveying the whole field. Using these images the management can monitor the shift in the disease patches timely saving time and labour.
Given the present rate of affect on the tea plantations it has become very necessary to improve the planning and decision making process of the management so that the situation could be dealt with effectively. Occurrence of pest and diseases takes place frequently so continuous monitoring and assessment is very essential. In order to have long term solutions, efforts should be made to replant the affected tea areas by hybrid and resistant varieties. This will prevent the spreading of the diseases to further areas. Efforts should be made to remove the severely affected tea patches and applying the necessary control measures for moderately affected plants.
- Replanting young hybrid tea plants can help prevent spread of the diseases.
- Proper drainage channels to prevent water logging in the plantation area.
- Proper manuring and pesticide application in time to prevent the outbreak of diseases.