Drivers Of Land Use Change Detection

 
Drivers Of Land Use Change Detection 3,5/5 8784 votes

BackgroundThe demand for meeting local food production has caused farmlands to expand at the cost of natural forests and grasslands in the Ethiopian highlands. However, empirical evidences on rate and patterns of LULC dynamics, and major driving forces in highlands of Ethiopia at catchment level were rare to contribute to design effective land management options. This study was to analyze the rate and patterns of LULC dynamics, and identify major driving forces in the Gelda catchment. ResultsSix different LULC maps derived from aerial photographs and Landsat images were produced, and comparisons were made.

  1. Land Use Change Analysis
  2. Drivers Of Land Use Change Detection Software

The results indicated that the study catchment has undergone significant LULC alterations and transformations since late 1950s. Farmlands and settlement were expanded by 57.7% while shrubs, forests and grasslands were declined by 18.6, 83.8 and 53.5% over the entire study period, respectively. The magnitude of initial grasslands and farmlands converted into degraded land seems small; however these can significantly cause an irreversible damage to the soil resources.

Drivers Of Land Use Change Detection

Windows xp sp3 jpn isotonic version. The combinations of land reform of 1975, forest development and villagization program 1980s, civil war, frequent changes in political structure, and population pressure were the major driving forces of LULC change. ConclusionTherefore, the GIS and remote sensing based change detection matrix analysis technique could provide useful baseline information to understand the spatiotemporal patterns of land use transitions caused by the major driving forces thereby sustainable land management planning is possible. Study area descriptionThe Gelda catchment is located between 11°38′14″ and 11°46′15″N latitude, and 37°25′54″ and 37°41′29″E longitude. It has an area of 26,264 hectares, covering about 2.2% of the Lake Tana watershed. It is drained by a stream named as Gelda (from where the name of watershed is given) flowing into Lake Tana from west direction (Fig. ). The landform of the catchment reflects its geological history where uplifting force created an initial elevated landmass and the subsequent outpouring of basaltic lava provided a thick protective cap and added on to the elevation (Eleni et al.

According to the report of Geological Survey of Ethiopia (GSE ), the catchment generally comprises materials ranges from alkaline to transitional basalts that often forming shield volcanoes, with minor trachyte and phenolite flows called “Tarmaber Gussa Formation” in southeast and eastern parts of the catchment formed during the Oligocene to Miocene epochs of the Tertiary period. The western and northwestern parts of the catchment consist of alluvial and lacustrine deposits of the Quaternary period. The altitude ranges from 1780 to 2481 meters above sea level. The slope gradient is dominated by gentle slope (0–7.6%) covering about 48.5% (12,682 ha) and moderately steep (7.7–16%) with 37.13% (9714 ha) of the catchment.

The steep (16.2–30.2%) and very steep (30.3%) slope gradients cover about 11.22% (2936 ha) and 3% (787 ha) respectively, commonly found in the southeastern and southern corners of the catchment. Gleysols (54.9%) and Nitisols (30.5%) form the major soil types of the catchment (FAO ). Gleyosols are poorly drained with seasonal water accumulation (Driessen and Deckers ). The other soils are commonly found on the sloping lands.The type of climate is generally sub-tropical with average total annual rainfall amount of 1453 mm showing a high inter-annual variability (Hurni ). The available rainfall records for the period between 1961 and 2014 records showed that the pattern is predominantly uni-modal. The main rainy season category, June to October, accounts for about 90% of the total annual rainfall (Fig. ).

The highest mean monthly temperature (30 °C) is recorded in April whereas the minimum (8 °C) in December with mean monthly range of 5 °C (Fig. ). As evidenced from protected areas like churches and grave yards, climatic–climax vegetation found in the area include Juniperus procera (locally tid), Hagenia abyssinica (locally called kosso); Albizia gummifera ( Sassa); Podocarpus falcatus (Zigba); Cordia africana ( wanza); and Ficus vasta ( Warka). Field observation also indicated that non-indigenous tree species like Cupressocyparis leylandii ( yeferenj Tid), Acacia sieberiana ( Yefereng Girar) and Eucalyptus spp. ( Bahir zaf) are expanding.Subsistence rainfed crop production with supplementary traditional irrigation and livestock husbandry are the main sources of livelihood. Dera wereda ARD office showed that the commonly grown crops in the catchment include barley ( Hordeum vulgare L.), maize ( Zea mays L.), teff ( Eragrotis teff), noug ( Guizotia abyssinica), and finger millet ( Eleusin coracana). Data sources and methods of analysisTime series black-and-white photographic and Landsat satellite images were the main source of input data for the LULC analysis in this study.

Drivers Of Land Use Change Detection

The 1957 aerial photographic images and 1973 Landsat image were the only oldest remote sensing data available for the study area, but there were no any reference images available between 1957 and 1973. Thus, further subdivisions for LULC analysis between 1957 and 1973 were not possible. Eleven 1 m × 1 m black-and-white scanned photographic images of 1957 by 600 dots per inch (DPI) were obtained from Ethiopian Mapping Agency (EMA). These were used for photogrammetric processing, as well as visual photo-interpretation activities. Materials such as mirror and pocket stereoscopes were employed for stereo viewing during visual interpretations of features on hardcopy pairs of aerial photos. The visual interpretations process using stereoscopes were helpful to substantiate onscreen feature classifications using GIS environment.

Land Use Change Analysis

Subsequently, all photographic images were geometrically corrected based on geo-referenced and 1:50,000 scale topographic map to produce orthorectified images using “ Geo Correction Tools” of ArcGIS. Finally, the mosaic image was produced from eleven orthorectified images using “ Mosaic Tool” and later “ subset image” into the required study area AOI of ERDAS EMAGINE 9.2.Landsat images were downloaded from Global Land Cover Facility (GLCF) in the USGS archives at Glovis. Five images were downloaded at about ten years’ interval to easily visualize changes in spatiotemporal LULC patterns. However, some discrepancy ±1 year was considered due to the availability and quality of Landsat images from USGS archives for the study area. Therefore, the Landsat MSS of 1973 (60 m × 60 m), Landsat TM of 1984 and 1995 (30 m × 30 m), Landsat ETM+ of 2004 (30 m × 30 m) and Landsat OLI of 2014 (30 m × 30 m) at path (169 and 170p), and row (52r) images were used. These were preprocessed such as layer stacking, sub-sampling the study area by AOI file, re-sampling of all time-series images into similar ground resolution (30 m × 30 m), gap filling for Landsat 7 SCL-off (2004 and 2014 images), and spectrally enhancing the images before actual image classification process.The major LULC classes considered in the classification are as given in Table. The farm and settlement areas were included in the same land cover as it was difficult to separate these two on the employed images.

Likewise, wetland class has been excluded in the classification process as an independent LULC class because this class largely covered with grassland, cultivated fields and forest adjacent to Lake Tana. Consequently, it was difficult to analyze wetlands separately while the area under study is occupied by different LULC classes. About 100 ground control points (GCP) representative of the different LULC classes were taken by a GPS receiver to improve accuracy of classification and to produce thematic land cover maps representative to the entire study period. Field observation was also conducted to substantiate the image classification and analysis.LULC classification was based on quantitative method that requires an independent classification of time-series images for the same geographic location followed by a comparison of the corresponding pixels to identify and quantify areas of change (Lillesand and Kiefer ).

Based on the correlation between the collected field data and the preliminary visual interpretation, the entire area was delineated from the orthorectified and geometrically corrected historical aerial photographs of 1957 using on-screen digitization by “ Spatial Analysis Tool” of ArcGIS 10. The digitized vector data also have undergone editing, overlay analysis and topological checks. Finally, cartographically finished historical LULC maps of 1957 were produced. Conversely, because of low resolution of Landsat images, only major LULC types were considered. Accordingly, supervised classification was done using training areas obtained from ground truth data combined with spectral signatures in false color composite images by maximum likelihood classifier. Subsequently, the accuracy assessment was conducted for all the classified Landsat images (maps) using accuracy assessment tool in ERDAS Imagine 9.2 to evaluate the user’s and the producer’s accuracy.

Thus, the accuracy assessment result indicated that most of the maps met the required minimum 85% accuracy in LULC analysis as reported in Table. This showed a strong agreement between the classified LULC classes and the geographical data (ground truths) and made it possible to use the output maps for change detection analysis. Where A t1 is the area of one type of land use in t 1 time; A t2 is the area of the same type in t 2 time and C is the rate of change in percent.

Drivers Of Land Use Change Detection Software

This formula is a simple and effective method of weighting the relative rate of change of one LULC class over time where the change directions are determined. Moreover, matrix analyses (image differencing) were employed for images between 1973 and 1995, 1995 and 2004, 2004 and 2014 to determine the LULC change detection using change detection tool in ERDAS Imagine 9.2. However, change detection between 1957 and 1973 was not possible due to incompatibility of 1957 LULC map in the change detection analysis of ERDAS Imagine 9.2. Accordingly, areas that are converted from each LULC class to any of the other over time were computed. The change detection matrix represented the simultaneous gross loss and gain for each LULC class where net losses or gains could be calculated in a given period of analysis. However, field observation and informal interview with the local elders using observation checklist and interview guide were also undertaken in the study area to substantiate the information obtained from GIS and remote sensing based LULC analysis.

Therefore, LULC analysis was done using ERDAS Imagine 9.2 and ArcGIS 10 version involved key steps (Fig. ). LULC change in the study catchment between 1957 and 1973Table and Fig. Indicated an existence of spatiotemporal transformations in the identified LULC classes. Shrinkage of forest cover was observed between 1957 and 1973, while the other LULC classes showed expansions of their original extent in this early period of analysis.

Forest cover was declined by 71.5% (265.2 ha/year) over the time span of 16 years. The shrinkage could be attributed to destruction of natural forests in search for additional farm plots, construction materials and domestic fuel consumption in the study catchment. However, farmland and settlement cover, which accounted the lowest coverage ever in the entire 57 years, but has been extended by 12.9% (73.3 ha/year).

Shrubs accounted for 27% of the total area of the catchment in 1957 and expanded by 10.2% (45.5 ha/year). Likewise, grasslands were around 15.8% of the study area and exceptionally increased by 51.7% (134.2 ha/year). These possibly reflect the impacts of deforestations on land use transitions where natural forests are ultimately converted into farmlands, shrubs and grasslands between 1957 and 1973. Actual transformed figures of LULC class to other were not given because change detection matrix analysis was not possible. LULC change in the study catchment between 1973 and 1984An expansion of farmland and settlement, forests and bare land were observed between 1973 and 1984 (Table ).

The change detection matrix also indicated similar results although some portions of their original extent were converted into other LULC classes (Table ). The farmland extended by about 19.24% (1961.51 ha), giving an annual average expansion of 178 ha/year. The expansion was largely contributed by shrubs by about 32.3% (3946.1 ha) and grassland by about 15.8% (1929.56 ha) despite some of the original extent was converted largely into grassland by 16.9% (1735.59 ha), and shrubs by 26.04% (2660.21 ha). This indicated farmlands were expanded, largely, at the expenses of shrubs and grasslands as the only means of increasing crop production. This expansion could also be attributed to the demand for attaining of food crops and more settlement land (Fig. ). This suggests that food crop production under the then period of Ethiopian situation could have been obtained through farmland expansions; which is a sign of poor land productivity as well as leading to destruction and farming of non-cultivated lands such as grazing, shrubs and wetlands. The exceptional increment of forests by 20.06% (338.87 ha) was largely due to additional areas obtained from grassland by 13.7% (277.56 ha), farmland and settlement by 18.7% (379.82 ha), and shrubs by 37.75% (765.94 ha) despite some of the original cover was lost largely into farmland and settlement by 52.98% (895.27 ha) and shrubs by 6.97% (117 ha).

This suggests that land acquisition from other LULC classes for forest conservation program was in response to initiatives of restoring indigenous trees and forests by the past socialist regime’s forest development program since the establishment of Gelda protected forest in 1980. Similarly, bare land cover also expanded due to major area gain from conversion of grasslands by about 59.2% (119.6 ha), and farmland and settlement by about 29.3% (59.35). The expansion of bare land reflected the impact of unsustainable utilizations of grasslands and farmlands due to overgrazing and land degradation.Shrinkages were observed in the extent of grasslands by 27.6% (158 ha/year) and shrubs 7.26% (52 ha/year) despite some gains were obtained from other LULC classes (Table and ). It is because some areas gained areas from different LULC classes could not bring a net expansion in grasslands and shrubs or did not compensate its decline in this second period of analysis. The substantial decline for grasslands were largely contributed by conversion of its original extent into farmland and settlement 30.63% (1929.56 ha) and shrubs by 33.92% (2136.44 ha) despite some gains were observed mainly from farmland and settlement by 38% (1735.59 ha) and shrubs by 19.2% (875.56 ha). This was attributed to land acquisition for farmlands following the downfall of the imperial regime in 1974 and the resulting policy changes at the expense of grazing lands and shrubs. This also showed the increasing demand for farmlands as well as forest development into areas once occupied by grassland.

In addition, shrinkage of shrubs were largely due to conversion of its original extent into farmlands and settlement by about 50.2% (3946.1 ha) despite some area gains were observed mainly from grasslands, and farmland and settlement by about 29.3% (2136.44 ha) and 36.48% (2660.21 ha), respectively. This clearly showed that about half of the original shrubland were transformed into farmland and settlement caused by population pressure and changes in land tenure system that transformed landless tenants into land owners. LULC change in the study catchment between 1984 and 1995LULC study between 1984 and 1995 showed expansion of grassland, farmland and settlement, and bare land despite some area were lost into other LULC classes (Tables, ). The area devoted to farmland and settlement showed a steady expansion by about 33.44% (370.3 ha/year) in this third period of analysis.

Change detection matrix also indicated that the most important contributors of farmlands and settlement expansions were grasslands by about 11.93% (1940.02 ha) and shrubs by about 24.15% (3925.9 ha) despite some of its initial extent was transformed into other LULC classes (Table ). This indicated that expansions were at the expenses of grasslands and shrubs cover.

A remarkable expansions were also observed in bare land by about 89% (16 ha/year) and grasslands by about 28.67% (118.9 ha/year). The expansion of bare land was due to area gain obtained largely from grasslands by about 48.27% (183.85 ha). This showed that unsustainable grassland management in the form of uncontrolled grazing attributed to the transformations of its initial grassland cover into bare soil areas. As a result, expansions in grasslands were observed largely due to farmland and settlement by about 28.6% (1680.4 ha), and shrubs by about 33.66% (1975.4 ha) despite some its initial extent was mainly converted into farmland and settlement.

Use

This possibly forced the local farmers to prepare grasslands meant for grazing converted largely from their farmlands and shrubs.