Monitoring of North West New Territories Development -
Using Normalized Difference Vegetation Index (NDVI)

By YUEN Sze Chun, Anthony


Normalized difference vegetation index (NDVI) derived from SPOT data can be a tool for urbanization monitoring. Today, most of the Hong Kongˇ¦s areas have been highly urbanized. Her urbanization process has, however, not been commenced until 1970s when the Hong Kong Government has implemented plans for developments of a number of new towns in the New Territories. With the tremendous economic growth in coastal regions of inland China particularly in the areas of the Pearl River Delta and a number of extensive road projects such as Deep Bay Link for strengthening the linkage between inland China and the North West New Territories (NWNT) in Hong Kong, a significant leap in urbanization in NWNT can be expected.

Monitoring of urbanization in NWNT is required for socioeconomic planning. In this class project, simple image differencing between NDVIs computed from 1993 and 2004 SPOT data is conducted for observing changes in NDVI in NWNT. It has been observed that the decrease in NDVI from 1993 to 2004 can be related to conversion of land into man-made structures such as road, railway, drainage channels, buildings and village houses which are the phenomenon of urbanization. Apart from urbanization of the various town centers, separated decrease in NDVI at various spots can be observed which reflects development in villages by construction of Small Houses. It can be concluded that remotely sensed images and application of remote sensing methodologies would have provided some effective means for observing and monitoring the urbanization process and outcomes in the NWNT.


There is a general trend of urbanization, depending on different countries, around the world in the last hundred years.1 As larger portion of human lives in the urbanized areas, change in land-cover, environmental and socioeconomic conditions induced by urbanization has become an important study. Hong Kong has been developed into an international financial city for over twenty years. According to the Census and Statistics Department of the Hong Kong Special Administrative Region Government, the Gross Domestic Product of Hong Kong in the 4th quarter of 2004 is HK$404,909 million (at year 2000 constant market prices) compared to HK$266,594 million (at year 2000 constant market prices) in the 4th quarter of 1993 2.

Since 1970s, the new towns such as Tsuen Wan, Shatin, Tuen Mun, Yuen Long and Tin Shui Wai, Tai Po, Sheung Shui, Fanling and Tseung Kwan O in the New Territories have been developed successfully. In addition, the unique New Territories Small House Policy since 1972 has also helped to certain extent, the successful development of the New Town projects. Under the New Territories Small House Policy, any eligible indigenous villager of the recognized villages in the New Territories can apply to the Hong Kong Special Administration Region Government for building a Small House with maximum roofed-over area of 65.03 m2 once in a lifetime 3 in the village environs of the village. Apart from the Policy, the prosperous property market in the 1990s has also helped causing many village houses proposed and built. There have been significant land cover changes in the New Territories in the past decades as a result of rapid urbanization in the New Territories. North West New Territories (NWNT) comprising of Yuen Long with the Tin Shui Wai sub-district as its hinterland, is one of the major recent developing focus. Several infrastructure projects have been recently completed or under construction such as Tai Lam Tunnel, KCRC West Rail, Deep Bay Link4, etc. As NWNT is in close proximity to the Pearl River Delta 5, it is expected that there will be significant land cover changes in NWNT and means of monitoring will help policy makers to formulate socioeconomic and infrastructure planning.

In this study, I would try to apply the Remote Sensing techniques to land management projects for a better realization of its application and limitation in studying sub-urban development in a small city.

2. Normalized Difference Vegetation Index (NDVI)

NDVI is a commonly used vegetation index aiming at measuring biomass or evaluate the physiological conditions and patterns of green vegetation distribution.6 In equation form, NDVI = (NIR ˇV R)/(NIR + R), whereas NIR represents the spectral reflectance in near infrared band while R represents red band. Healthy photosynthesizing green vegetation normally has characteristics of strong absorption in blue and red band and strong reflectance in near-infrared band. Urbanization often displaces green vegetation despite landscaping or urban greening. It has become a significant consideration in designing urban development or urban planning in recent years. As a result, urbanization may be evaluated by measuring the change in NDVI across time.

In the New Territories, agricultural land grown with green vegetation has been converted into building land with buildings, village houses. Theoretically, NDVI may be applied to evaluate land cover change and this project would select NWNT as a sample area to evaluate whether the application of NDVI and change detection techniques of remotely sensed images can detect changes.

3. Datasets, data pre-processing and methodology

3.1. Datasets

For the purpose of this study, only two SPOT images have been chosen. They were acquired on 26 December 1993 and 11 December 2004. They were both acquired in winter of dry season and could be considered as near anniversary images because both images were taken in December.

3.2.Image Pre-processing

The SPOT image in 2004 available to this study project had been geo-referenced after checking against Hong Kong DEM data and overlay with road and coast vector files. By contrast, geometric registration had not been conducted on the 1993 image.

In essence, the 1993 image was registered against the 2004 image utilizing GCPWorks of PCI Geomaticac V.9.0 (Figure 1). The RMS error of this registration process was achieved at 0.25. Although the 1993 image covered only part of the Hong Kong territory, it did not cause problems in this project because the study area focused in NWNT.

Figure 1 geometric registration of 1993 SPOT

Pixel digital number (DN) may vary between two remotely sensed images of different dates even though the object has no change in reflectance. In this project, a relative radiometric correction was done with the application of linear regression analysis tool in Microsoftc Excel. A total of thirty sample points were selected, which were believed to have static or minimal change such as airport, runways, concrete surface, fresh water body such as reservoirs etc.

The respective DNs in red (R) band and near-infrared (NIR) band for these sample points in both SPOT images in 1993 and 2004 were recorded in the Microsoftc Excel spreadsheet and linear regression analysis was conducted to work out a model in the form of y=a+bx (Figure 2 and Figure 3). The respective sets of a and b values (R band and NIR band) were then applied on respective DNs of each band of 1993 SPOT image.

Figure 2 Linear Regression of NIR band

Figure 3 Linear Regression of R band

Since this project area focuses on NWNT, only the interested area of images were exported to reduce file size for saving processing time.

Figure 4 The 1993 NDVI image extracted from SPOT data

NDVI for each pixel of the study area was computed after the above image pre-processing. As shown on the 1993 and 2004 NDVI images (Figure 4 and Figure 5), water bodies including seawater and fishponds clearly have the lowest NDVI values and hence they, followed by urban developments including buildings, roads.

Figure 5 The 2004 NDVI image extracted from SPOT data

3.3. Methodology

Image Differencing is a relatively simple and effective tool amongst common change detection techniques. It involves subtracting the imagery of one date from that of another. If there is an increase in reflectance, the image appears light tone while the image appears dark tone if there is a decrease in reflectance. If the image appears gray, it may imply that the area has no or minimal change between two dates. Notwithstanding with its simple algebra, it is a fast and effective tool to detect changes over two remotely sensed images of different dates despite it has limitation of not showing information about the nature of change.

Image differencing technique can apply not only to images of two different dates. It can apply to comparison of vegetation index information derived from multiple dates of imagery.7 Change detection based on differencing multiple-date NDVI in Hong Kong has also been conducted.8 In this project, only the change of computed NDVI in NWNT area derived from NDVI images in 1993 SPOT and 2004 SPOT was conducted.

In computing NDVI and NDVI differencing using built-in arithmetic module within XSpace of PCI Geomaticac V.9.0, consideration should be given whether scaling would be applied. The NDVI real values, by definition, would be between -1 and 1 but XSpace could allow scaling up for display purpose, depending whether 8-bit, 16-bit unsigned or signed or 32 bit real data channel being chosen. For NDVI differencing, either 16-bit signed or 32-bit real was chosen in this project to reflect the increase in NDVI and decrease in NDVI by positive values and negative values respectively.

4. Results and discussion

Figure 6 Differenced image: NDVI2004-NDVI1993

The NDVI differencing result (NDVI2004 ˇV NDVI1993) is at Figure 6. There are near black regions or blocks showing decrease in NDVI. They are newly developed features such as Route 3 road networks, West Rail, new buildings and houses, newly constructed drainage channel built under river training projects 9. These areas are basically at the outskirt of the original Yuen Long Town and situated at village areas near the town. Apart from normal urbanization from the fringe areas from town centers, we can observe smaller spots distributed in rural areas with decrease in NDVI from 1993 to 2004. The result is due to development of villages by construction of Small Houses under the New Territories Small House policy. Also the villages near towns having a faster development in the period between 1993 and 2004 than other villages farther away from towns. The major reasons are due to these areas having better transportation network and infrastructure support.

Despite there are not many bright areas reflecting increase in NDVI since 1993, there are some areas near north-eastern side of Yuen Long town. These areas are converted from fishponds to vegetation land cover. Due to strong competition from inland market, the local fisheries have been diminishing and the fishponds have been converted to other uses by landowners. Apart from Mai Po Conservation Area, such land cover changes are restricted by the enforcement of the Town Planning Ordinance (Chapter 131) because filling of fishponds or ponds may seek prior approval from the Town Planning Board 10.

Apart from normal urbanization from town developments in NWNT, the development in villages driven by N.T. Small House Policy is another factor because the Small Houses built would give rise to environmental and socioeconomic problems 11. Infrastructure planning has to take these factors into account during development of NWNT in the 21st century.

5. Conclusion

With archives of remotely sensed images as time goes by, remote sensing techniques provide an efficient and effective tool for monitoring rapid changes in sub-region area like in Hong Kong. As technology advances with more high spatial resolution remote sensing satellites, the application of conventional change detection methods such as NDVI image differencing in this project is able to detect land surface change from vegetated land to built-up areas as a means in monitoring urbanization in NWNT.

6. Further Studies

This study aims at giving a perspective view of applying remotely sensed images and remote sensing methodologies in Hong Kong for urbanization monitoring. The NDVI values computed can be further analyzed quantitatively about the general magnitude of change in NDVI between 1993 and 2004 in NWNT. Also, addition of remotely sensed SPOT data acquired between 1993 and 2004 would surely enable the observation of trend of NDVI change of these 11 years.

7. Acknowledgement

Before ending the study report, I would like to say my special thanks to Professor YANG Limin and Ms. WANG Dan. They have consistently steered me in conducting the study and have given a great number of concrete advices and effective guidance. It is with their valuable supports that this study can have been materialized as above.

  1. D. MAKTAV, F.S. ERBEK and C.JURGENS, 2005, Remote Sensing of urban areas, International Journal of Remote Sensing Vol. 26, No.4, p655-659
  2. Census and Statistics Department of the Hong Kong Special Administrative Region Government, the GDP Statistics can be downloaded in Microsoftc Excel format at
  3. Pamphlet issued by Lands Department, the Government of HKSAR, ˇ§The New Territories Small House Policy ˇV How to Apply For a Small House Grantˇ¨ which can be accessed on-line through
  4. Deep Bay Link, upon completion, will provide a new link between Hong Kong and the inland through the Hong Kong ˇV Shenzhen Western Corridor
  5. Planning Study on the Co-ordinated Development of the Greater Pearl River Delta Townerhsip
  6. JOHN R. JENSEN, 2005 Introductory Digital Image Processing A Remote Sensing Perspective 3rd Ed. 310-314 p
  7. JOHN R. JENSEN, 2005 Introductory Digital Image Processing A Remote Sensing Perspective 3rd Ed. 478-482 p
  8. T. FUNG and W. SIU, 2000, Environmental quality and its changes, an analysis using NDVI International Journal of Remote Sensing, 2000, Vol. 21, No. 5, 1011-1024
  9. (accessed in April 2006)
  11. LISA HOPKINSON and MANDY LAO MAN LEI, 2003 Civic Exchange, Rethinking the Small House Policy