Home Articles Land use and cover change detection and modelling for North Ningxia, China

Land use and cover change detection and modelling for North Ningxia, China

Weicheng Wu*+,Eric F. Lambin* and Marie-Francoise Courel+
+Department of Geography, University of Louvain,
Place Louis Pasteur, 3 1348 Louvain-la-Neuve, Belgium
*PRODIG-UMR8586/CNRS, 191 Rue St-Jacques, 75005 Paris, France
E-mail of the corresponding author: https://gis.globtierlabs.com/mailto:[email protected]

North Ningxia, extending from longitude 105°45’E to 107°00’E and latitude 38°20’N to 39°30’N, including 50% of the Helan Mountains and 80% of the Yinchuan Plain and surrounded by Inner Mongolia on the east, north and west (figure 1), is an arid and semi-arid region in Northwest China. The annual precipitation ranges from 78 to 295mm (the maximum, 430 mm, appears in the Helan Mts), annual evaporation from 1473 to 2318mm and annual average temperature from 8.2°C to 9.6°C in the recent decades (Ningxia Statistical Yearbook, 1988, 1990, 1992, 1997 and 2000). The analysis on meteorological data in the past half century indicates that the annual temperature has been increasing and precipitation decreasing in Yinchuan region. It is probably a local indicator of the global warming. The climate has been getting dryer and warmer and the natural conditions more and more difficult.

Land use, especially, in agriculture, has a long history in this region. As early as 35,000 to 25,000 years B.P., people already begun their activities (Geng et al, 1992). In the dynasty Qin (221 to 207 B.C.), the first irrigation canal came into existence. This signalled the launch of an agricultural era in the Yinchuan Plain. In the successive dynasties Han (206 B.C to 220 A.D.) and Tang (618 to 907 A.D.), the irrigation system was further developed and improved. But due to the frequent wars and large scale of emigrations in history (Geng et al, 1992), it did not have a stable development until recent decades. With the execution of the ‘reform-and-open’ policy in 1980s in China, land use and cover change has taken place rapidly owing to the agricultural, industrial developments and population growth. Recently, with more and more development programmes of the Chinese central government focused on Northwest China, North Ningxia becomes one of the hotspots. It is therefore of first importance to undertake a synthetic land use and cover investigation, change rate measurement, driving force analysis and develop a dynamic monitoring system in order to generate fundamental land use data, a management prototype and some useful references for the local governments in their sustainable land use planning and decision making. It is the objective of the research and one of the tasks scheduled in the Sino-Belgian co-operation project on Northwest China.

In correspondence to the implementation of the new land use policy in 1980s in China, the period from late-1980s to late-1990s was chosen to conduct a land use and cover change detection, monitoring and modelling utilising the multi-temporal remotely sensed data (Landsat TM dated 1987, 1989 and ETM 1999) and county-level socio-economic and meteorological data.

A multidimensional synthetic method from space to ground and from human activities to environmental changes was applied in this research.

  1. Changes observed from space
    Since the macroscopic and multi-temporal observation advantages, remotely sensed data are undoubtedly the most ideal data for extracting the land cover change information. There exist mainly two approaches to realise this procedure. The traditional one is the post-classification comparison, aiming to find out the difference between the classified images of two different dates (Weismiller et al.1977, Gorden,1980, Singh, 1989). Some authors, however, proposed to perform this detection by image differencing (Jensen et al. 1982, Quarmby et al. 1989, Singh 1989 and

    Lambin 1994, 1997). The latter has been followed to produce the general change maps in our study. Complemented with visual comparison, the detailed county-level land use changes were distinguished. The procedures adopted in this change detection are shown as follows:

    Figure.1:Location map of the study area, North Ningxia
    • Image-to-image registration of the remote sensing data (TM1987, 1989 and ETM1999) with a RMS error of 0.53 and 0.58 pixel using polynomial function (third model) and bilinear re-sampling by the topographic maps on a scale of 1/200,000 to 1/300,000 geocoded in the datum WGS84 and projection UTM (48).
    • Tasseled Cap transformation (Crist et al. 1984) on the ETM and TM images to convert the land cover information included in 7 bands into 3 indicators: brightness, greenness and wetness, which respectively means the land bareness, vegetation vigor and soil moisture.
    • Indicator differencing (e.g., greenness or brightness) between 1999 and 1987 or 1989.
    • Thresholding to acquire the changed areas and produce the general change maps which contain three classes: negative change, no change and positive change. For example, as to the greenness indicator, ‘negative change’ means vegetation degradation and ‘positive change’ vegetation increase.
    • Visual comparison to identify the change types (e.g., farmland extension, urban extension, land degradation, land to water depression, etc.) and create a detailed county-level land use and cover change map.
    • Quantification of the county-level land cover changes.
  2. Exploration on the mechanism from human activity to environmental changes
    A panel analysis (Lambin, 1994) conducted by a multivariate linear regression modelling was carried on to find out the land use and cover change driving forces and its decisive factors by linking the change detection results with the corresponding changes in human socio-economic and meteorological data.

1.This paper is a part of the outcomes of the Sino-Belgian cooperation project on Northwest China in the University of Louvain, B1348, Belgium. The research was funded by the OSTC of the Belgian government under contract BL/10/C15.
2.The dataset of Landsat TM image (129/33, Sept.20, 1987) belongs to the data archive of Fukui Research Group, Keio University, Japan. We use it with their permission.

Change detection results
According to the above processing procedures, the county-level land use and cover change detection was achieved and thematically mapped in figure 2. The change data are illustrated in table 1. It is recognised that 11.74% of the total territory had changed in the period 1987-1999, among which the predominant change is the farmland extension (471km2 in area, 49.44% of the total change). Urban and rural built-up increases are apparent (35.8 and 68.33km2 respectively in surface area, 3.58% and 7.34% of the total change). Water-body has an increase of about 49km2. However, the fastest change here is the Yellow River narrowing. The running course had decreased by 83.83km2 at an reduction rate of 6.10%

. Table 1:County-level land use and cover change data from 1987 to 1999 in North Ningxia

County County area Total farmland extension Natural vegetation increase Urban extension Rural built-up Land degradation Land to water body Water body to land Water body increase River course into land
Huinong 942.01 14.8 5.09 0.39 7.14 5.30 7.34 16.95 -9.61 10.81
Pingluo 2115.77 124.9 48.73 5.19 21.86 9.74 32.26 23.63 8.63 22.58
Taole 906.98 33.4 9.76 0.08 2.73 7.09 7.75 0.00 7.75 28.66
Shizuishan 575.65 1.01 0.07 9.39 0.00 12.33 0.89 1.11 -0.22 0.00
Helan 1229.45 55.73 12.02 0.56 15.39 3.54 14.93 3.38 11.55 5.52
Yinchuan 1321.53 117.9 14.77 19.13 11.61 13.49 16.65 2.17 14.48 10.95
Ongning 1028.78 123.8 4.29 1.10 9.6 3.97 16.61 0.4 16.21 5.31
North Ningxia 8120.18 471.54 94.73 35.84 68.33 55.46 96.43 47.64 48.79 83.83
Proportion in the total territory (%) 5.81 1.17 0.44 0.84 0.68 1.19 0.58   1.03
Annual change rate km2/yr 39.29 7.89 2.99 5.69 4.62 8.04 3.97 4.07 6.99
% 1.53   2.16 4.23       0.998 6.10

Figure 2: County-level land use and cover change map of North Ningxia from 1987 to 1999

Panel analysis, a multivariate regression modelling
Land use and cover changes can not take place independently but have certain linkages with the human activities and mutations in natural conditions (e.g., climate change). Understanding the dynamics of land use and cover change has increasingly been recognised as one of the key research imperatives in global environmental change research (Lambin et al, 1999; Geist et al, 2001). The monitoring of such changes would be most relevant and useful when it is accompanied by the understanding of the forces driving change processes. This task could be calibrated by a statistical modelling, ‘Panel analysis’, which links the changes in dependent variables (e.g., land use changes) during a certain interval of time with the changes in independent variables (e.g., human activities) in the corresponding interval of time and across a large number of localities (Lambin, 1994). This analysis postulates a linear relationship between the dependent and independent variables and can be mathematically expressed as follows (Kleinbaum et al, 1976 and 1998; Lambin, 1994):

Y = ß0 + ß1X1 + ß2X2+ ß3X3. + ß4X4 + ? + ßn Xn + E ………….……………………………………(1)
where, Y is the dependent variable, i.e., land use change(s), Xn are the independent variables, i.e., driving forces, or rather, human activities, ß0 is a constant (or intercept) and ßn are regression coefficients and E a random error component.

Such modelling can discriminate the causes or driving forces governing land cover changes (Lambin,1994; Lambin et al, 2000; Merten, 1997; Serneel et al, 2001; Wu et al, 2001).

The county-level land cover change data (table 1) and the corresponding changes in county-level socio-economic and meteorological data from 1988 to 1999 (from the Ningxia Statistical Yearbook, 1989, 2000) were incorporated and inputted into SYSTAT, a software for multivariate analysis. Taking into account the land use changes as dependent and socio-economic and meteorological data as independents, within a confidence level of 0.05, the modelling results are shown in table 2.

Table 2: Land use and cover change driving forces in North Ningxia

Dependent Final entered independent Constant Parameter estimate Std error Std coeff. Df F Pr>F R2
Farmland extension Agricultural output increase – 0.3627 0.0240 0.0060 0.855 1 13.61 0.0140 0.731
Natural vegetation increase Meat product increase – 3.3314 3.9910 0.5570 0.955 1 51.357 0.0010 0.911
Land degradation increase Industry output increase 5.0306 0.0020 0.0010 0.841 1 12.112 0.0180 0.708
Rural built-up increase Rural labour force growth – 1.0098 -1.0305 0.2105 -1.594 1 23.960 0.0160 0.996
Food product increase 0.2885 0.0417 1.949 1 47.860 0.0060
Agricultural output increase 0.0281 0.0032 0.721 1 76.020 0.0030
Urban extension Urban population increase – 0.1519 0.1340 0.0100 0.986 1 170.066 0.0002 0.971
Water to land Sown area increase -2.0059 1.6716 0.3620 0.932 1 22.249 0.0050 0.867
Land to water Agriculture output increase 0.3965 0.0448 0.0120 0.862 1 14.461 0.0130 0.743


Discussion and conclusions
Based on the above change detection and panel analysis, a discussion on land use change mechanism and its future situation and its conclusions are demonstrated as follows:

  1. The farmland extension is the prevailing rural environment modification in North Ningxia in the period 1987-1999, however, the agricultural output increase plays a relatively weak part in the GDP growth (11.61%). Regression modelling shows that this extension is strongly associated with the agricultural output increase (R2 = 0.731), that is to say, the activities of the rural population are the decisive factors in this kind of land use change:

    [Farmland extension] = – 0.3627 + 0.024 [Agricultural output increase] ………………….(2)
    Developing at a rate of 39.3km2/yr or 1.53%, the farmland will be extended by about 432km2 in 2010. In fact, to compensate the loss in urbanisation and village extension (142km2), the farmland should have an increase of 574km2 in 2010 to meet the needs of the population growth. However, soil resources investigation (Wu et al, 2002) suggests that there is little arable land left for future reclamation. The Yinchuan Plain has been fully exploited in its long development history in agriculture, even the infertile sand land in Yongning, Yinchuan and Pingluo counties has been converted into cultivated land since 1987. Then where to find such a large uncultivated land for reclamation in the future?

  2. The urban extension (34km2 in area), especially in Yinchuan (17.43km2) and Shizuishan (9.39km2), constituting 3.58% of the total change, but its related production increase takes up 88.39% of the total GDP growth. Panel analysis discovers that the urban population growth is the key factors dominating this extension (R2 = 0.971), that is:

    [Urban extension] = -0.1519 + 0.1340 [Urban population growth] ………………………….(3)
    This equation indicates that the socio-economic activities of the increased urban population in industry, commerce and service enterprises play a preponderant role in the urban extension. At the same time, the urban population growth is positively correlated with the industrial output increase (R2 = 0.958) and GDP growth (R2 = 0.992). This means that the urban extension or recent urbanisation is attributed to the activities of the newly incremented urban population in the industry development and other GDP formation domains such as commerce and service enterprises.

    At an extension rate of 2.99km2/yr or 2.16%, the urban area will be increased by 41km2 under the contribution of the urban population till 2010.

  3. Rural built-up area has augmented by 70km2 at a change rate of 5.69km2/yr or 4.23% in the past 12 years. This augmentation is linked with the rural labour force growth, food product increase and agricultural output increase (R2 = 0.9960). Their relationship can be expressed as:

    [Rural built-up increase] = -1.0098 + 0.0281 [Agricultural output increase] + 0.2885 [Food product increase] – 1.0305 [Rural labour force increase] …………………………………..………………….(4)
    Rural construction, principally housing in village, is accomplished by the rural labour force to improve the living conditions of the rural people and stock the agricultural production. Driven by these factors, the village will be enlarged by 101km2 in 2010. This means that another 101km2 of the green land will be lost till that time.

  4. From our detection, land desertification stated by Zhu (1995) was not found in this study area, however, land degradation in a total area of 55.62km2, including coal residue increase (36.20%), vegetation degradation and salinisation (52.25%) and stone pit increase (11.53%), was apparently observed in Yinchuan, Shizuishan and Pinglguo counties. The degradation rate was measured as 4.62km2/yr. Panel analysis reveals that such degradation is linked with the industrial output increase (R2 = 0.708):

    [Land degradation] = 5.0306 + 0.0020 [Industrial output increase] ………….…………..….(5)
    This is because industry output contains coal mining and coal industry product. Furthermore, industry development needs more and more stones and sands for factory construction. Therefore, industry development promotes the GDP growth and urban extension but at the same time brings about a degradation of land. In line with the degradation tendency in the last decade, another 50.82km2 of land will be degraded till 2010, which will be spatially related with the coal mining and urbanisation.

  1. Two kinds of mutation of water-body were distinguished in the Yinchuan Plain. One is ‘land to water-body’, meaning that land changed into waterlogged depressions or fish ponds. This mutation is linked with the agriculture output increase (R2 = 0.743). That is to say, the activities of rural people in food production, fish farming, etc, which compose partially the agricultural output, are the main cause of this change. Their relationship is presented below:

    [Land to water-body] = 0.3965 + 0.0448 [Agriculture output increase] ………….……….…(6)
    The other change is ‘water-body to land’, representing water-body drying or having converted into land, partially farmland. Panel analysis shows that this change is associated with the sown area increase (R2 = 0.867). It is reasonable that people need more water for irrigation with the increase of the cultivated land. If the water-body has been poorly recharged, it can be easily dried. Additionally, the dried water depression can be cultivated and become farmland (e.g., Gaomiao Lake in Huinong county). These might be the reasons that the ‘water-body to land’ is related to the sown area increase:

    [Water-body to land] = -2.0059 + 1.6716 [Sown area increase] ……………….………………(7)
    Therefore, water-body changes seem to have some transversal mobility which is related to the agricultural activities: in some places, waterlogged depression extended or land turned into water depressions or fish ponds while such depressions converted to land elsewhere. But totally, the surface of the water-body in these counties has an increase of about 49 km2 at an increment rate of 4.07km2/yr or 0.998%. Then where does the water come from when the climate gets dryer and warmer ?

  2. The Yellow River has been largely narrowed in the extent of North Ningxia (by 83.83 km2). This decrease would be probably due to the following reasons: (1) local climate warming and precipitation decrease and (2) overuse, even waste of water in agriculture in the Yinchuan Plain.

    At a reduction rate of 6.99km2/yr or 6.10%, the present Yellow River (with a surface of 81.3km2, Wu et al, 2002) would be wholly dry in 2010. This means that the river, which has been the cradle of the 7000-years’ Chinese civilisation, would not exist any longer in ten years!

The authors want to thank first the Federal Office for the Scientific, Technical and Cultural Affairs (OSTC) of Belgium Government for funding our research of the Sino-Belgian Co-operation Project on Northwest China. We would also like to thank our Chinese partner, the United Remote Sensing Centre of Northwest China, especially, Ningxia Remote Sensing Centre, for their reception during our field verification and for their provision of the necessary maps and socio-economic data so that our research could be achieved smoothly. We would too thank Prof. Hiroyuki Yoshida and Fukui Research Group, Keio University, Japan, for their permission to access the Landsat TM data (1987). Here a special thank will go to Mr. ZHANG Wenfeng for his first arrangement of the county-level socio-economic data during his stay in Belgium for the co-operation research.


  • CRIST, E. P., CICONE R. C., Application of the Tasseled Cap Concept to Simulated Thematic Mapper Data, Photogrammetric Engineering & Remote Sensing, Vol.50, p.343-352, 1984.
  • GEIST, H. and LAMBIN, E.F., What drives tropical deforestation? LUCC Report Series No.4, LUCC International Project office, University of Louvain, 2001.
  • GENG, K., SHAN, P. et al, The past, present and future of Yinchuan region: The evolution processes, characteristics and regularities of the late Quaternary geographical environment (in Chinese), Survey Press of China, 1992.
  • GORDON, S., Utilizing Landsat imagery to monitor land use change: A case study in Ohio, Remote Sensing of Environment, Vol. 9, p.189-196, 1980.
  • JENSEN, J.R. and TOLL, D.R., Detecting residential land use development at the urban fringe, Photogrammetric Engineering and Remote Sensing, Vol.48, p.629-643, 1982.
  • KLEINBAUM, D.G., KUPPER, L.L., MULLER, K.E. and NIZAM, A., Applied Regression Analysis and Other Multivariable Methods, p.111-119, Duxbury Press, 1998.
  • LAMBIN, E.F., Modelling deforestation processes (A Review), Tropical ecosystem environment observations by satellites, TREES series B: Research Report n°1, EUR15744EN, p.45-101, 1994.
  • LAMBIN, E.F., Modelling and monitoring land-cover changes processes in tropical regions, Progress in Physical Geography, Vol. 21, p.375-393, 1997.
  • LAMBIN, E.F. et al, Land use and land cover change (LUCC) Implementation Strategy, IGBP Report 48 and IHDP Report 10, IGBP, 1999.
  • LAMBIN, E.F., Land cover assessment and monitoring, in WILEY, J.: Encyclopedia of Analytical Chemistry’, 2000.
  • MERTENS, B., and LAMBIN, E.F., Spatial modelling of deforestation in Southern Cameroon: Spatial disaggregation of diverse deforestation processes, Applied Geography, Vol. 17, p.143-162, 1997.
  • QUARMBY, N.A., CUSHNIE, J.L., Monitoring urban land cover changes at the urban fringe from SPOT HRV imagery in south-east England, International Journal of Remote Sensing, Vol.10, p.951-963, 1989.
  • SERNEEL, S. and LAMBIN E.F., Proximate cause of land-use change in Narok District, Kenya: a spatial statistical model, Agriculture, Ecosystem and Environment, Vol.85, p.65-81,2001.
  • SINGH, A., Digital change detection techniques using remotely sensed data, International Journal of Remote Sensing, Vol.10, p.989-1003, 1989.
  • STATISTICAL BUREAU OF NINGXIA, Ningxia Statistical Yearbook, China Statistics Press, 1989, 1992, 1994, 1997 and 2000.
  • STATISTICAL BUREAU OF NINGXIA, Ningxia Economic Atlas, Xi’an Map Press, 1989, 1997.
  • WEISMILLER, R.A., KRISTOF, S.J., SCHOLZ, D.K., ANUTA, P.E., and MOMEN, S.A., Change detection in coastal zone environments, Photogrammetric Engineering and Remote Sensing, Vol.43, p.1533-1539.
  • WU, W. and LAMBIN, E. F., Land use and cover change dynamics in North Ningxia (an abstract), LUCCD, Aug. 2001, Beijing.
  • WU, W. and ZHANG, W., Present land use and cover patterns and their development potential in North Ningxia, China, Journal of Geographical Sciences (submitted), 2002.
  • ZHU, Z.D. and CUI, S.H., Desertification in China: Status, Trends and Strategies, Journal of Chinese Geography, Vol.5, p.32-42, 1995.