The conversion of agricultural land in the peri-urban areas of Hanoi (Vietnam): patterns in space and time

Figures & data

Figure 1. Location of the study area within the former city of Hanoi.

Table 1. Population of the study area in 1993, 2000 and 2007.

	1993	2000	2007
	Number
Urban population	1,118,218	1,484,138	2,082,825
Rural population	979,711	980,793	970,929
Total population	2,097,929	2,464,931	3,053,754
Number of wards	100	113	133
Number of communes	102	93	73

Source: General Statistic Office, 2009.

Table 2. Classification rules.

Land cover class	Rule
Built-up area	Mean layer 1 ≥ 95
	Brightness ≥ 95
	153 ≤ X distance to scene left border ≤ 620272 ≤ Y distance to scene bottom border ≤ 668
	0.32 ≤ HSI transformation (R = layer 7, G = layer 5, B = layer 2) Hue ≤ 0.329
Agricultural land	NDVI ≥ 0.3
	HSI transformation (R = layer 6, G = layer 5, B = layer 2) Hue ≥ 0.38
	Mean layer 1 ≤ 88
	GLCM mean (all direction) ≥ 50
Water body	MNDWI = 0.01
	NDVI < 0
Bare land	Brightness ≥ 95
	HSI transformation intensity (R = 6, G = 5, B = 2) ≥ 0.4

Notes: Brightness: gray level; HSI transformation: hue saturation intensity transformation; GLCM mean: mean value of gray level co-occurrence matrices (for a detailed description of rules, readers may consult eCognition 8.64 (Definiens, 2009)).

NDVI (normalized difference vegetation index) and MNDWI (modified normalized difference water index) (Xu, 2006) are calculated as follows:

where G, NIR, MIR and R are the spectral values in the green, near-infrared, middle-infrared and red bands of the Landsat images, respectively.

Table 3. Description of the spatial metrics used in this study (McGarigal et al., 2002).

Metric	Descriptions	Unit	Range
TCA	TCA equals the sum of the core areas of each patch (m^2) of the corresponding patch type, divided by 10,000 (to convert to hectares). Core area is the area (m^2) within the patch that is further than the specified depth-of-edge distance from the patch perimeter. Core area is a compound measure of shape, area and edge depth, that is, core area is effected by all three of these components simultaneously	Hectares	TCA ≥ 0, no limit
LPI	LPI is equal to the percent of the total landscape that is made up by the largest patch. When the entire landscape is made up of a single patch, the LPI will equal 100. As the size of the largest patch decreases, the LPI approaches 0	Percent	0 < LPI ≤ 100
MPI	Measure of the degree of isolation and fragmentation. MPI uses the nearest neighbor statistic	None	MPI ≥ 0
MNN	Measure of patch isolation. The nearest neighbor distance of an individual patch is the shortest distance to a similar patch (edge to edge). The mean nearest neighbor distance is the average of these distances (meters) for individual classes at the class level and the mean of the class nearest neighbor distances at the landscape level	Meter	MNN > 0, no limit
LSI	LSI equals the total length of edge in the landscape, given in number of cell surfaces, divided by the minimum total length of edge possible, also given in number of cell surfaces, which is achieved when the landscape consists of a single patch	None	LSI ≥ 1, no limit

Notes: TCA, total core area; LPI, largest patch index; MPI, mean proximity index; MNN, mean nearest neighbor distance; LSI, landscape shape index.

Table 4. Confusion matrix conducted for the 2007 image classification (unit = ha).

	Classification data
Reference data	Built-up area	Agricultural land	Water body	Bare land	Row total	Producer accuracy
Built-up area	264.2	35.8	9.6	6.6	316.3	0.8
Agricultural land	48.6	309.2	26.1	16.0	399.9	0.8
Water body	15.9	18.5	112.4	4.5	151.8	0.7
Bare land	1.4	6.8	2.3	33.9	44.5	0.8
Column total	330.1	370.4	150.5	61.5	912.4
User accuracy	0.8	0.8	0.7	0.6		0.8

Figure 2. Final classification of land cover types in 1993, 2000 and 2007.

Note: Numbers in (b) and (c) represent districts with important agriculture conversion, see also their name in Figure 1. Raw image data available form U.S. Geological Survey.

Table 5. Trajectories of land cover change in 1993–2000 (above) and in 2000–2007 (below) (unit = ha).

	1993 Land cover
2000 Land cover	Built-up	Agriculture	Water	Bare land
Built-up	7055.7	3856.2	147.8	38.3
Agriculture	730.3	39,641.9	1309.1	379.3
Water	80.8	2826.8	4832.7	517.5
Bare land	21.2	348.0	246.9	201.2
	2000 Land cover
Built-up	10,690.7	8109.6	697.3	138.7
Agriculture	253.2	30,216.3	1468.4	205.7
Water	90.2	2047.6	5612.9	246.1
Bare land	58.0	1662.1	483.9	229.3

Table 6. Extent and annual change and annual rate of change of land cover types.

	Annual change (ha/year)		Annual rate of change (%)
Land cover class	1993–2000	2000–2007	1993–2000	2000–2007
Built-up	467.4	1205.9	0.05	0.08
Agriculture	–659.7	–1419.8	–0.01	–0.04

Figure 3. Changes in the five metrics over the two periods of observation.

Figure 4. The process of agricultural landscape transformation: agricultural patches in the peri-urban are fragmented in 1993 and become more irregular (2000) then isolated from each other (2007).

Table 7. Mean values of changes in area and metrics for each administrative group.

	Average changein 1993–2000			ANOVA		Average changein 2000–2007			ANOVA
	RR	RU	UU	F-value	R^2	RR	RU	UU	F-value	R^2
Area	–8.3	–23.1	–4.1	23.4	0.19	–18.1	–23.4	–4.8	46.5	0.32
LPI	–11.0	–27.5	–9.2	7.6	0.10	–24.1	–26.0	–14.1	4.3	0.07
TCA	–48.1	–49.7	–4.3	14.2	0.17	–87.2	–58.4	–25.7	12.7	0.17
MPI	–4.9	8.7	43.8	3.5	0.09	26.7	34.1	31.2	0.1	0.00
MNN	364.9	–81.5	–19.1	44.7	0.05	–36.3	–14.4	–8.0	1.3	0.02
LSI	0.9	0.2	–0.3	23.4	0.40	0.1	–1.1	–0.8	34.1	0.36

Note: Underline values: except these values, the others are significant at p < 0.01. Figure 1. Location of the study area within the former city of Hanoi.

Definiens. (2009). eCognition developer 8 – User guide. München: Definiens.

 Google Scholar

Xu, H. (2006). Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery. International Journal of Remote Sensing, 27, 3025–3033.

 Web of Science ®Google Scholar

McGarigal, K., Cushman, S. A., Neel, M. C., & Ene, E. (2002). FRAGSTATS: Spatial pattern analysis program for categorical maps. Amherst, MA: Computer software program produced by the authors at the University of Massachusetts.

 Google Scholar