Land cover classification in Thailand using : a combination of vegetation index and day and night time thermal bands of NOAA-AVHRR | |
| Author | Chada Narongrit |
| Call Number | AIT Diss. no.SR-00-4 |
| Subject(s) | Land use--Thailand--Classification Advanced very high resolution radiometers |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Technical Science, School of Advanced Technologies |
| Publisher | Asian Institute of Technology |
| Abstract | The A VHRR sensor providing coarser spatial resolution and daily temporal resolution both in day and night time of global surfaces was initially designed to observe the Earth's weather in form of cloud pattern. However, this sensor was subsequently designed to measure other phenomena such as sea surface temperature (SST), tenestrial vegetation, forest fire, and so on. For land covering mapping and land processing NDVI index computed from the ratio of A VHRR channels 1 and 2 was typically used in classifying land cover types. Later, brightness temperature of channel 4 was promising for classifying land cover types. However, only daytime A VHRR channels were used for land cover type classification. Hence, in this study, the combination of day and night time A VHRR data was investigated for classifying land cover types in Thailand. The aim is to examine the usefulness of the nighttime A VHRR data in the classification with respect to improve accuracy and to increase numbers of classes of land cover types derived from A VHRR coarse resolution data. The experiment was canied out in various band combinations including, NDVI, channels 3 and 4 of A VHRR composite images acquired in cool (December, 1997) and hot (March, 1998) seasons. Moreover, land surface temperature (LST) images retrieved from the adjusted LST model were also used in the classification. Two classification algorithms were implemented: Maximum Likelihood and Decision Tree. The study reveals that among the bands derived from A VHRR data, NDVI gave the most consistent result with high classification accuracy. However, the combination between NDVI and other bands such as band 3 and LST gave a better accuracy than those obtained from using NDVI alone. The classification results in both cool and hot season showed that by using the combination of day and night time data, the overall accuracies were better than those obtained from using only daytime data. The LST and the A VHRR channel 3 gave better accuracy compared with ordinary channel 4. A combination of three-bands NDVI, daytime LST, and nighttime LST gave the best accuracy, 84.14% in cool season and 79.40% in hot season; whereas only daytime input ofNDVI and LSTd provided the accuracy of 81.46% in cool season and of 75.35% in hot season. Although the overall accuracy of the combination between day and night time data was not much improved compared with the using only daytime data the combination between day and night can well classify built up areas and forest. The accuracy of these two classes was much improved. The results revealed that nighttime LST can well classify built up area and forestland, thus, this data is effective to detailed study on heat island and forest landscape. The study suggested an integrated approach involving day and night time data that were used model vegetation based on landscape or to classify land cover types associated with elevation or landscape at regional scale. The LST model adjusted in this study, LST = T4+ 1.11 (T4-T5), has different estimating temperature from actual ground temperature for 1.2 °C in daytime and for -0.4 °C in nighttime. The LST images derived from the adjusted model were used in the classification step. In part of LST investigation, the nighttime LST showed a stronger con-elation with ground temperature than that of daytime LST. The con-elation between LST and actual ground temperature increased when NDVI increased. This indicates that usmg LST m classifying land cover types is more precisely in dense-vegetation area. |
| Year | 2000 |
| Type | Dissertation |
| School | School of Advanced Technologies (SAT) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Space Technology Application and Research (SR) |
| Chairperson(s) | Mitsuharu, Tokunaga;Murai, Shunji |
| Examination Committee(s) | Kaew Nualchawee;Apisit Eiumnoh;Suphat Vongvisessomjai;Yasuoka, Y oshifumi |
| Scholarship Donor(s) | Naresuan University |
| Degree | Thesis (Ph. D.) - Asian Institute of Technology, 2000 |