Land Use Dynamics and Landscape Change Pattern in a Mountain Watershed in Nepal
4.2 Classification of satellite images
We used supervised Maximum Likelihood classification method for the classification of all the images. Training areas corresponding to each classification item (hereafter, land use class), in case of IRS image, were chosen from among the training samples collected from the field and in case of MSS and TM images they were generated from the interpretation of aerial photographs of the study area from 1978 and 1992 respectively. Although the dates of the aerial photographs used as reference information in classification do not exactly match with the dates of the satellite images, they were used with the assumption that land use in the watershed, particularly forestry land use, was not substantially changed between the time of aerial photography and satellite observation dates. Moreover, this was the best feasible option that could be used in this research.
For producing land use maps for 1976, 1989 and 2000 and to investigate changes that occurred between these periods, the following six land use classes were considered in image classification: broadleaf forest, conifer forest, shrublands, grasslands, lowland agriculture, and upland agriculture and other. Choice of these land use classes was guided by: i) the objective of the research, ii) expected certain degree of accuracy in image classification, and iii) the easiness of identifying classes on aerial photographs. A brief description of each of the land use classes is given in Table 2.
Table 2: Land use classes considered in image classification and change detection
| Land use class | General description |
| Broadleaf forest | Forest areas with estimated 75 percent or more of the existing crown covered by broadleaf trees. The predominant species are: Castanopsis spp. and Schima wallichii in most part and Quercus spp. in higher elevations. |
| Conifer forest | Forest areas with estimated 75 percent or more of the existing crown covered by planted or naturally growing conifer trees. Pinus roxburghii, Pinus patula and Pinus wallichiana are common species. |
| Shrublands | Land covered by shrubs, bushes and young broadleaf regeneration. Degraded forest areas with estimated <10% tree crown cover are also included. |
| Grasslands | Non-cultivated areas dominated by herbal vegetation. |
| Lowland agriculture | Irrigated, level-terraced agricultural lands in river valleys, used for multiple cropping including winter crops. Wheat and potato are two major winter crops cultivated in these lands after the harvest of paddy rice in November-December. |
| Upland agriculture and other | Non-irrigated agricultural lands with or without slopping terraces, barren lands, settlements, roads, construction sites and other built-up areas. |
Among all the land use classes, “upland agriculture and other” (hereafter, upland agriculture) is the most complex class. In fact, it includes all other combinations of land uses, which are not included in the rest five classes. During winter, uplands in the study area, like most of the Middle Hills, are mostly barren and have spectral values similar to those of barren lands such as non-vegetative hills and riverbeds (Tokola et al., 2001). Moreover, during the time the satellite imageries were taken (particularly IRS image) many upland terraces had exposed soil due to fresh plowing by farmers as a preparation for the next summer crop. This condition of the cultivated uplands made it impossible to distinguish them from rough roads, new construction sites and other built-up areas. This justifies combining settlements, barren lands and built-up areas (estimated around 7 percent of this class) with upland agricultural lands in this study, which may not be acceptable at any other time of the year.
Presence of shadow in parts of all the images and cloud in parts of the TM image were other major problems encountered during image classification. Both of these areas were classified as separate classes and latter combined to the respective classes with the help of “ground-truth” information.
4.3 Post classification
After selectively combining classes, classified images were sieved, clumped and filtered before producing final output. Sieving removes isolated classified pixels using blob grouping, while clumping helps maintain spatial coherency by removing unclassified black pixels (speckle or holes) in classified images (Richards, 1994). Finally a 3x3 median filter was applied to smoothen the classified images. All activities related to image processing were performed in Environment for Visualizing Images (ENVI) Version 3.2 (Research Systems Inc., Colorado, USA).
Classified images were then exported to Arc View GIS Version 3.1 (ESRI, Redlands, USA) from ENVI and rest of the analyses was performed in GIS environments. The images were first converted to grid in Arc View and then to shape format. The polygon themes so generated, were exported to Arc Info GIS Version 3.5.1 (ESRI, Redlands, USA) and polygons of <0.5 ha in size were “eliminated” in Arc Info. This elimination was necessary to minimize the effects of classification errors arising from resolution differences among the three satellite images while at the same time without significantly altering the area under each land use class. The resultant polygon themes were used in further analyses.
4.4 Detection of land use changes
The land use polygon themes for 1976, 1989 and 2000, obtained from the digital classification of satellite data and subsequent GIS analyses using the method described above were overlaid two at a time in Arc View GIS and the area converted from each of the classes to any of the other classes was computed.