Øystein B. Dick

Landslide susceptibility mapping at Hoa Binh province (Vietnam) using an adaptive neuro-fuzzy inference system and GIS

The objective of this study is to investigate a potential application of the Adaptive Neuro-Fuzzy Inference System (ANFIS) and the Geographic Information System (GIS) as a relatively new approach for landslide susceptibility mapping in the Hoa Binh province of Vietnam. Firstly, a landslide inventory map with a total of 118 landslide locations was constructed from various sources. Then the landslide inventory was randomly split into a testing dataset 70% (82 landslide locations) for training the models and the remaining 30% (36 landslides locations) was used for validation purpose. Ten landslide conditioning factors such as slope, aspect, curvature, lithology, land use, soil type, rainfall, distance to roads, distance to rivers, and distance to faults were considered in the analysis. The hybrid learning algorithm and six different membership functions (Gaussmf, Gauss2mf, Gbellmf, Sigmf, Dsigmf, Psigmf) were applied to generate the landslide susceptibility maps. The validation dataset, which was not considered in the ANFIS modeling process, was used to validate the landslide susceptibility maps using the prediction rate method. The validation results showed that the area under the curve (AUC) for six ANFIS models vary from 0.739 to 0.848. It indicates that the prediction capability depends on the membership functions used in the ANFIS. The models with Sigmf (0.848) and Gaussmf (0.825) have shown the highest prediction capability. The results of this study show that landslide susceptibility mapping in the Hoa Binh province of Vietnam using the ANFIS approach is viable. As far as the performance of the ANFIS approach is concerned, the results appeared to be quite satisfactory, the zones determined on the map being zones of relative susceptibility.

Land-use changes and their impacts on soil degradation and surface runoff of two catchments of Northern Ethiopia

Abstract Land-use/land-cover changes and their associated impact on environment in the period from 1964 to 2006 was studied in two catchments located in the highland of Tigray using geographic information system and remote-sensing approaches supplemented with field measurements. Results show that, for all periods, cultivated land constitutes the most prevalent (>60%) land-use type and shows a total increase of 1.7 ha y−1 at Gum Selassa and a decrease of 5.5 ha y−1 at Maileba. Forest and woodland suffered more damage in both areas losing 32.8 ha (100%) and 53 ha (100%), respectively at Gum Selassa; and 1.74 ha (96.7%) and 52.7 ha (100%) at Maileba over four decades. At Gum Selassa, shrubland decreased by 1.26 ha y−1 while at Maileba it showed a slight positive increment of 0.38 ha y−1. Area under settlement increased by a greater magnitude at Maileba (6.3 ha y−1) and a slight increase at Gum Selassa (1.4 ha y−1) in response to the rapid population increase. These changes in land uses/cover brought significant deleterious impacts on land degradation and surface runoff. The cumulative degradation index (DI) was negative for all land uses, with a higher value under Eucalyptus plantation (DI=−282) followed by cultivated land (DI=−260) at Maileba. Changes in land use/cover also decreased the water-storage capacity of soils by 1.63 and 1.09 mm y−1 at Gum Selassa and Maileba, respectively, with a corresponding increase in surface runoff by 2.7 and 2.3 mm y−1. Generally, the observed changes in land degradation and surface runoff are highly linked to the change in land use/land cover.

Effects of land-use change on carbon dynamics assessed by multi-temporal satellite imagery in a mountain watershed of Nepal

Abstract Satellite imagery from 1976, 1989, and 2003 was analysed to assess land-use change and its effect on carbon (C) dynamics in the Pokhare Khola watershed of Nepal. The soil organic carbon (SOC) pools and forest vegetation C pools were estimated using field data and Intergovernmental Panel on Climate Change (IPCC) tier 1 method. The analysis of a temporal series of satellite imagery was found to be an effective tool to determine the land-use-change dynamics in the mountain watershed. During the period 1976–2003, overall forest area decreased by 24%, with significant increase in area under cultivated lands (82%). While taking into account the individual land-use types, the areas under rainfed upland, managed dense forest, and Schima–Castanopsis forest increased by 200, 174, and 58%, respectively, with substantial decreases in both degraded forest/shrub lands (35%) and pine mixed forest (92%). The effects of land-use change on SOC pool varied according to the type of land-use change. There was a net gain in the SOC pool in the larger part of the watershed during 1976–1989, but in the period 1989–2003 a net loss was observed. Conversion from forest to cultivated lands has resulted in a severe loss of vegetation carbon. However, in the future, there will be higher vegetation carbon pools in the watershed, because a significant land area under degraded forest/shrub land and pine mixed forest has been converted to managed-dense forest and Schima–Castanopsis forest, which were shown to contain significantly higher amounts of vegetation carbon.

Spatially distributed modelling and mapping of soil organic carbon and total nitrogen stocks in the Eastern Mau Forest Reserve, Kenya

Detailed knowledge about the estimates and spatial patterns of soil organic carbon (SOC) and total nitrogen (TN) stocks is fundamental for sustainable land management and climate change mitigation. This study aimed at: (1) mapping the spatial patterns, and (2) quantifying SOC and TN stocks to 30 cm depth in the Eastern Mau Forest Reserve using field, remote sensing, geographical information systems (GIS), and statistical modelling approaches. This is a critical ecosystem offering essential services, but its sustainability is threatened by deforestation and degradation. Results revealed that elevation, silt content, TN concentration, and Landsat 8 Operational Land Imager band 11 explained 72% of the variability in SOC stocks, while the same factors (except silt content) explained 71% of the variability in TN stocks. The results further showed that soil properties, particularly TN and SOC concentrations, were more important than that other environmental factors in controlling the observed patterns of SOC and TN stocks, respectively. Forests stored the highest amounts of SOC and TN (3.78 Tg C and 0.38 Tg N) followed by croplands (2.46 Tg C and 0.25 Tg N) and grasslands (0.57 Tg C and 0.06 Tg N). Overall, the Eastern Mau Forest Reserve stored approximately 6.81 Tg C and 0.69 Tg N. The highest estimates of SOC and TN stocks (hotspots) occurred on the western and northwestern parts where forests dominated, while the lowest estimates (coldspots) occurred on the eastern side where croplands had been established. Therefore, the hotspots need policies that promote conservation, while the coldspots need those that support accumulation of SOC and TN stocks.

Analysis of X-SAR SRTM elevation data to estimate surface cover heights over land areas

We present the first results of the SRTM AO-038 project in Norway. The SRTM X-SAR elevation data are evaluated with respect to digital reference maps and surface cover types like agricultural fields, forest, lakes and infrastructure. Optical satellite images, aerial photos and field observations are used to aid in the analysis of the SRTM elevation data. The X-SAR SRTM DEM gave better elevation accuracies over many open agricultural areas as compared to the 1:50000 topographic DEM commonly used in Norway. The SRTM DEM gave surface cover elevations rather than ground elevations in areas of dense forest. These SRTM elevation offsets in forested areas may be used further to categorize the forest cover type.

Effects of Land Cover Changes on Soil Organic Carbon and Total Nitrogen Stocks in the Eastern Mau Forest Reserve, Kenya

This study analysed the variations of soil organic carbon (SOC) and total nitrogen (TN) stocks under natural forests (NF), plantation forests (PF), bamboo forests (BF), and croplands that had been converted from such forests (i.e., NF2C, PF2C, and BF2C) in the Eastern Mau Forest Reserve using field, laboratory, spatial, and statistical techniques. The results displayed significant differences in SOC and TN stocks between NF and NF2C (p < 0.0001), and between PF and PF2C (p < 0.0001). For instance, the surface soils (0–15 cm) of NF had the highest SOC and TN stocks (71.6 and 7.1 Mg ha−1, respectively), while NF2C had the lowest (35.4 and 3.5 Mg ha−1). Similarly, the subsurface soils (15–30 cm) of NF had the highest stocks (55.7 and 5.6 Mg ha−1), while NF2C had the lowest (32.5 and 3.2 Mg ha−1). This reflects a decline in both SOC and TN stocks by about 51 % in the surface and about 42 % in the subsurface soils after NF conversion. There were also significant differences in SOC and TN stocks (p < 0.05) between the surface and subsurface soils of different land cover types. The stocks decreased as soil depth increased. This trend suggests that (i) forest-to-cropland conversions are undermining the ecosystem’s capacity for carbon sequestration, and (ii) subsurface soils have potential for carbon sequestration. SOC and TN losses in the croplands may be mitigated by adopting best management practices (BMPs), especially agro-forestry. These findings are useful for designing sustainable land management (SLM) and carbon sequestration projects.

A Novel Evolutionary Genetic Optimization-Based Adaptive Neuro-Fuzzy Inference System and Geographical Information Systems Predict and Map Soil Organic Carbon Stocks Across an Afromontane Landscape

Abstract Soil organic carbon (SOC) pool has the potential to mitigate or enhance climate change by either acting as a sink, or a source of atmospheric carbon dioxide (CO2) and also plays a fundamental role in the health and proper functioning of soils to sustain life on Earth. As such, the objective of this study was to investigate the applicability of a novel evolutionary genetic optimization-based adaptive neuro-fuzzy inference system (ANFIS-EG) in predicting and mapping the spatial patterns of SOC stocks in the Eastern Mau Forest Reserve, Kenya. Field measurements and auxiliary data reflecting the soil-forming factors were used to design an ANFIS-EG model, which was then implemented to predict and map the areal differentiation of SOC stocks in the Eastern Mau Forest Reserve. This was achieved with a reasonable level of uncertainty (i.e., root mean square error of 15.07 Mg C ha−1), hence demonstrating the applicability of the ANFIS-EG in SOC mapping studies. There is potential for improving the model performance, as indicated by the current ratio of performance to deviation (1.6). The mapping also revealed marginally higher SOC stocks in the forested ecosystems (i.e., an average of 109.78 Mg C ha−1) than in the agro-ecosystems (i.e., an average of 95.9 Mg C ha−1).

Georeferencing, feature extraction and change detection in Harbour and city areas using high-resolution satelite SAR images acquired from multiple sensors

The quality and usefulness of feature extraction and change detection will be governed by the geoposition accuracy of the SAR image. We show results from satellite SAR image pixel geolocation accuracy using deployed radar corner reflectors as reference targets. The best accuracy (latitude, longitude) is below one meter. We have also developed algorithms that are using stereo acquisitions to obtain decimeter geolocation accuracies in XYZ of the high-resolution satellite SAR images. We demonstrate how high-resolution satellite SAR image modes can be used to extract man-made features in harbors and city areas.