Image classification based on remotely sensed data is the primary domain of automatic mapping research. With the increasing of urban development, keeping geographic database updating is imminently needed. Automatic mapping of land cover types in urban area is one of the most challenging problems in remote sensing. Traditional database updating is time consuming and costly. It has usually been performed by manual observation and visual interpretation, In order to improve the efficiency as well as the accuracy, new technique in the data collection and extraction becomes increasingly necessary. This paper studied an object-based decision tree classification based on orthophoto and lidar data, both alone and integrated. Four land cover types i.e. Forest, Water, Openland as well as Building were successfully extracted. Promising results were obtained with the 89.2% accuracy of orthophoto based classification and 88.6% accuracy of lidar data based classification. Both lidar data and orthophoto showed enough capacity to classify general land cover types alone. Meanwhile, the combination of orthophoto and lidar data demonstrated a prominent classification results with 95.2% accuracy. The results of integrated data revealed a very high agreement. Comparing the process of using orthophoto or lidar data alone, it reduced the complexity of land cover type discrimination. In addition, another classification algorithm, support vector machines (SVM) classification was preformed. Comparing to the decision tree classification, it obtained the same accuracy level as decision tree classification in orthophoto dataset (89.2%) and integration dataset (97.3%). However, the SVM results of lidar dataset was not satisfactory. Its overall accuracy only reached 77.1%. In brief, object-based land cover classification demonstrated its effectiveness in land cover map generation. It could exploit spectral and spatial features from input data efficiently and classifying image with high accuracy.

The Bearded Vulture, Gypaetus barbatus, was previously distributed across southern and western South Africa, as far south as Cape Town. Today, the entire population in southern Africa is restricted to the Lesotho highlands and the Drakensberg escarpment and nearby mountains in South Africa, where the population continues to decline. Research has shown that territorial abandonment is apparently associated with the density of human settlements within 10km of a territory. This pattern was assumed to be linked to the increased risk of unnatural mortalities in areas with higher human presence. However, habitat use and habitat selection, especially whether the species actively avoids human settlements has not yet been explored and could contribute to this pattern. In this study, we used data from nine adult Bearded Vultures fitted with Global Positioning System (GPS) satellite tags to determine which habitat is being selected. We modelled habitat selection in relation to various topographic and habitat variables, including information on built-up areas (i.e., areas with a high density of buildings). We predicted that birds would select areas of grassland and avoid areas with high building density and adjoining areas. We found that Bearded Vultures selected areas closer to their nest sites and supplementary feeding sites, with steeper slopes, and highly rugged terrain. In terms of habitat, they selected areas with grassland and avoided areas with forest, while cropland was neither selected nor avoided. As predicted Bearded Vultures avoided built-up areas and their vicinity. These results suggest that the Bearded Vulture may be sensitive to the negative impacts of a changing landscape in its distribution range. These results can help conservation managers in guiding development (e.g., human settlement expansion and afforestation) and protecting priority habitats (e.g., grassland) within the breeding range of adult Bearded Vultures.

Maximum sustained and peak gust winds from eighteen European windstorms over the last 25 years were analyzed previously to develop surface-level wind predictions across a large and topographically varied landscape based on an anisotropic kriging interpolation methodology for meteorological station data. Results suggested that coastal and mountainous areas experience the highest wind speeds and highest variability over short distances, resulting in the highest errors across concurrent interpolated surfaces. This study utilizes covariates in conjunction with cokriging to investigate the use of cokriging as a method of improvement through the interpolation of five windstorms that impacted both the Alps region and the topographically-varied coastal regions of Western Europe. Results show that cokriging improves isotach interpolation for windstorms in 8 out of 10 models by reducing root mean square error and the total number of high-error stations, primarily in coastal and mountainous areas. Land cover alone contributed to the greatest model improvement in a majority of the models, while aspect and elevation (singularly and collectively) also improved models when compared to original kriging models. Improved surface interpolation is critical for improved understanding of macro-scale windstorm patterns and resulting damage, thus improving risk and vulnerability estimates.

The Miombo woodland vegetation of southern Africa has been subjected to anthropogenic pressures in recent years, resulting in a change in its cover and floristic-structural composition, and the population status of important tree species. The recent land use land cover changes (LULCC) is of concern due to the negative impacts on the Miombo woodland ecological functions. Understanding the overall dynamics of the vegetation that include LULCC, forest cover change, tree species composition, and diversity, population structure (PS) in relation to forest degradation, loss of endangered/rare species, is pivotal in influencing policy and sustainable woodland management. The Malawi Government instituted policies that allowed for improved forest management. However, the impact of forest management regimes on forest cover, tree species diversity, and structure is not well established. The study aimed to determine LULCC and forest cover changes and the associated drivers and how co-management (CM) and government-management (GM) impact on the florist-structural composition, diversity and the population structure of the important tree species in Malawian Miombo landscape. Firstly, the study analysed LULCC and the comparative impact of CM and GM on the forest cover in Malawi, for the period 1999-2018. CM involves a contractual agreement between communities (with a forest management plan who have been empowered to manage forest resources) and the Forest Department (representing the Government) in managing forest reserves. GM is the protection of forest resources by the government through the Forest Department. Since the introduction of participatory forest management (PFM), such as CM, its impact has not been established. Google Earth images (Landsat mosaics) for 1999 and 2018 for Malawi were acquired, registered, and pre-processed in Environment for Visualizing Images (ENVI 4.7) Harris Geospatial Solutions. LULCC estimation using the Inter-Governmental Panel on Climate Change (IPCC) classes was determined using the differences in error-adjusted areas between 1999 and 2018. Overall accuracies were >90%. Woodland net losses of 8.4% were to Plantation, Grassland, and Agriculture transition intensities. Agriculture net gains of 9.6% were from Grassland, Settlement, and Woodland transitions for Malawi. Forest cover within CM and GM indicated losses. Participatory land use plans and monitoring for diversified management in Malawian Miombo woodlands are required to mitigate anticipated irreversible impacts in the landscape. The second study investigated the factors that influence changes in CM and GM forest reserves between 1999 and 2018. CM and GM regimes in Miombo woodlands are possible interventions to mitigate forest degradation and deforestation in southern Africa. However, few studies have investigated the direct and indirect drivers of LULCC using socioeconomic characteristics and Remote sensing data in CM and GM regimes. The drivers of LULCC in forest reserves, and management challenges were identified using participatory assessments in both management regimes. The changes in woodland were observed with varying extent. Communities' perceptions in the factors contributing to changes in CM and GM forest reserves were similar and mostly due to the conversion of woodlands to agriculture while extraction of woody products led to forest degradation. In both management strategies, population pressure, youthful age, poverty, and poor education were associated with forest-based livelihood activities and therefore the woodland changes. The overall woodland cover loss to grassland is attributed to its importance as a source of energy. There is thus a need to harmonize policies for sustainable use and management of woodlands in order to address local, national, and regional ecosystem services. Future studies will need to link Remote sensing and socioeconomic data as part of a monitoring tool that could assist to sustainably adapt to changes in the woodlands and surrounding communities. The third study compared the Miombo Tree species composition and diversity between CM and GM regimes in Malawi. Tree species composition and diversity information is limited between CM and GM regimes. Two CM and two GM forest reserves were purposively selected to act as representatives of management regimes in the northern and southern regions of Malawi. Forest inventory data from 80 randomly selected nested circular plots were used. Two plot sizes: a large plot (0.16 ha; radius 22.6 m) to record stems ≥30 cm DBH, and the main plot (0.04 ha; radius 11.28 m) to record stems 5.0-29.9 cm DBH and species names. In total, 109 tree species belonging to 38 families, 87 species in GM FRs (Kaning’ina 58, Thambani 52), and 69 in CM FRs (Perekezi 45, Liwonde 43) were recorded. The largest families (number of species between brackets) were Fabaceae (34, with 3 subfamilies, Caesalpinioideae (17), Papilionoideae (12), and Mimosoideae (5), an indication of their adaptive potential in the area. Other important families were Combretaceae (7), Rubiaceae (7), and Clusiaceae (4). Species similarity between management regimes was low and was attributed to site factors, species characteristics and intensity of disturbances. TWINSPAN classification results were related to differences in site conditions and disturbances caused by historical and current resource use in management regimes. The eigenvalues ≥0.3 across CM and GM sub-communities indicated high stability. Brachystegia and Julbernardia species were highly important in CM and GM sub-communities. Uapaca species were highly important in agriculture and settlement abandoned areas in GM forest reserve. The study recommends selective harvesting to allow for dominant (Brachystegia and Julbernardia) and associated Miombo species to regenerate. Species richness and evenness (diversity) was high in more disturbed CM and GM sub-communities compared to intact areas. The high diversity was related to tree species high abundances of smaller stems with few scattered big trees. These results call for a Forest policy review to allow planned harvesting in GM forest reserves. Law enforcement is also required in both management regimes to mitigate unsustainable harvesting in sensitive areas. Future studies should include zonation to improve differentiation between site factors and wood extraction in stand development stages in management regimes. The fourth study compared the Miombo population structure (PS) between CM and GM regimes. Such information is limited in CM and GM regimes since the introduction of PFM in Malawi. The size class distribution (SCD) of sub-canopy/canopy species showed a reversed J-shaped profile in CM and GM forest reserves (South), when compared to CM and GM, in northern Malawi. These findings may reflect differences in the historical woodland utilization in the two regions. The bell-shaped SCDs in CM sub-communities with high stem density of Brachystegia and Julbernardia species suggest strong demand for light for successful recruitment from regeneration to adult trees and could be related to wood utilization. The reversed J-shape SCDs in northern GM sub-communities with high regeneration stem density of Pittosporum viridiflorum suggest an increase in shade-tolerant evergreen tree species under a low-level disturbance. Timber species showed interrupted SCDs with few to no stems, indicating challenges in regeneration. Pioneer species were associated with disturbances under CM indicating woodland recovery. The patterns in SCD showed similarities and differences between CM and GM sub-communities between the two management strategies. In Community 2, there were significant differences (p = 0.002) between management strategies with low canopy densities in CM, which could be attributed to unsustainable harvesting. Furthermore, saplings showed significant differences with a higher stems ha-1 in CM compared to GM. Trees and regeneration SCDs suggest a thorough analysis of the PS of varied species associations to guide sustainable resource use. An adaptive management approach that uses silvicultural systems to promote sustainable forest management is recommended. Additionally, selective harvesting in recovery stages would reduce intense competition in the dense, even-aged stands. However, there is need for instituting enabling policies and to monitor changes in both management regimes to promote biodiversity conservation, resource use, and diverse ecosystem services at all levels. / Thesis (PhD (Forest Science))--University of Pretoria, 2020. / Malawi Government Scholarship Program and the African Forest Forum / Plant Production and Soil Science / PhD (Forest Science) / Unrestricted

Land cover and climate change with their associated impacts on runoff are among the pressing areas of research within the western United States. In the first paper of this dissertation, we identified a total of 39 watersheds draining to U.S. Geological Survey (USGS) streamflow gauges, chosen either from the USGS Hydroclimatic Climatic Data Network of gauges that are minimally impacted by anthropogenic alterations, or because they have long, relatively continuous streamflow records and are representative of large areas within the study region in Utah. In each of these watersheds we examined trends in precipitation, temperature, snow, streamflow and runoff ratio as well as land use and land cover information. In addition, we developed a water balance model to quantify the sensitivity of runoff to changes in vegetation based on differences in evapotranspiration from different land cover types. The second paper addressed runoff sensitivity to land cover changes in a spatially explicit way by performing detailed simulations using a Regional Hydro Ecological Simulation System (RHESSys) model applied to the Weber River near Oakley watershed (USGS gauge # 10128500). Our runoff sensitivity results suggest that during winter reduced Leaf Area Index (LAI) decreases canopy interception, which tends to increase snow accumulations, and hence snow available for runoff during the early spring melt season. Increased LAI during spring melt season tends to delay the snow melting process due to reduced radiation beneath high LAI surfaces relative to low LAI surfaces. The last paper examined the sensitivity of the Great Salt Lake level to changes in streamflow input or changes in climate that manifest as changes in air temperatures over the lake. We quantified this sensitivity by examining an elasticity measure defined as the ratio of the variability of streamflow, precipitation, evaporation, area and salinity to the variability in historic volume changes. We also developed a mass balance model to simulate lake level and volume driven by stochastic precipitation, streamflow and climate inputs. We showed that fluctuation in streamflow is the dominant factor in lake level fluctuations, but that fluctuations in lake area, which modulates evaporation and precipitation directly on the lake, are also important.

京都大学 / 0048 / 新制・論文博士 / 博士(地球環境学) / 乙第12828号 / 論地環博第8号 / 新制||地環||24(附属図書館) / 31315 / 京都大学大学院地球環境学舎地球環境学専攻 / (主査)准教授 西前 出, 教授 渡邉 紹裕, 教授 小方 登 / 学位規則第4条第2項該当 / Doctor of Global Environmental Studies / Kyoto University / DFAM

The Provo River watershed has experienced land cover change over the past several decades. Land cover influences water quality inasmuch as land cover determines the type and quantity of non-point source (NPS) pollutants that may enter the water. This study examines the historical impacts of land cover changes on water quality in the Provo River using remote sensing and statistical analysis. Statistical correlations and linear regressions were used to study the relationship between various land cover types and water quality variables for six years between 1975 and 2002. This thesis supports research finding myriad impacts of urban land cover on water quality. The study also revealed that increasing pH, alkalinity, and bicarbonate levels in the Provo River are likely related to increasing urbanization of the watershed.

Detailed urban land cover maps are increasingly useful and important applications of remote sensing. Municipal agencies and others use land cover maps and data for numerous critical local planning and monitoring functions and for urban geographical research studies. Because of this, there is a demand for accurate urban land cover maps that can be produced quickly and economically. The availability of very high resolution multispectral imagery is an important factor in enabling such production, as the judicious selection of source imagery has a large impact on the resulting map products. Likewise, the implementation of appropriate digital image processing methods is crucial for deriving urban land cover maps of acceptable accuracy and cost. This study compared two common image classification algorithms using 2006 NAIP 1-meter GSD CIR images of Orem and Provo, Utah. The two classification procedures – conventional per-pixel supervised classification coupled with post-classification filtering, and object-based feature extraction – were compared for resulting accuracy and, in general terms, for their cost-effectiveness. Results demonstrated that object-based feature extraction has the potential to produce maps with better accuracy, but at a somewhat higher cost than per-pixel supervised classification. Classification errors and their probable causes are discussed; also a number of options for improving the classification accuracy are presented together with considerations of the potential costs involved. Although the ultimate goal of economical production of accurate urban land cover maps was not fully realized, this study nevertheless has established a cost containment baseline upon which methodological improvements can be built.

This study involved a geographic information systems (GIS) approach to modeling ground water pollution potential in the Southern Edwards Aquifer Region in Texas. The DRASTIC ground water model was analyzed using two methods. First, the effects of adding land cover data to the drastic model were evaluated. In the second approach, the effects of the removal of DRASTIC variables were evaluated. Six, five, and four variable models were generated and analyzed.

The widespread loss and degradation of habitat constitutes the largest threat to biodiversity in North America. While regulatory programs such as the Endangered Species Act of 1973 and wetland permitting under the Clean Water Act have addressed acute assaults on critical habitat, large areas of unprotected uplands have been lost. Urban development, particularly the advent of lower density suburban and rural sprawl, has greatly diminished the extent of contiguous patches of forest habitat and introduced a host of other undesirable effects on ecosystem function. This study sought to evaluate the extent of urban sprawl and its effects on ecological integrity in Florida using Landsat-derived land cover data collected by the Florida Fish and Wildlife Conservation Commission (FWC) circa 1987 and 2003. Chapter 1 described a novel GIS technique for correcting the systematic errors in the FWC 1987 and 2003 land cover data and converting those data to a common classification system so that they could be used in any ad hoc land cover change analysis. Comparison to ground-truth observations demonstrated a significant improvement in the accuracy of the land cover data following the Land Cover Correction Process (LCCP). Change detection between 1987 and 2003 using the correct land cover revealed trends in land cover conversion that were very different from previously published results derived from the original FWC land cover data. Conversion to urban uses in the corrected data was 47,293 ha lower, and conversion to agricultural uses was reduced by 196,773 ha, resulting in 244,067 ha less anthropogenic land conversion than had been previously estimated. Although the corrected land cover data showed that overall land conversion of natural areas was lower compared to the earlier estimate, the corrected data showed proportionally greater habitat losses for four important habitat types: Pinelands (-10.08% in the corrected land cover as compared to -5.90% in the original FWC data); upland forest (-9.46% versus 6.37%); sandhill (-13.90% versus 11.18%); and scrub (-15.52% versus -9.83%). Given the relatively small areal extent of some of these habitats, the larger percent loss estimates over the study period revealed by the corrected land cover data are cause for even greater concern by conservation planners and policymakers. Now that its utility has been demonstrated, the LCCP technique can be applied to any pair of roughly similar land cover mapping datasets provided that their original classification systems can be composed by a cross-walk into a single scheme, and that one or more ancillary data sets are available to serve in the tie-breaker role performed here by the land use data from Florida's Water Management Districts. The Soil Survey Geographic (SSURGO) and State Soil Geographic (STATSGO) soils data of the National Resource Conservation Service, the National Land Cover Dataset (NLCD) or the statewide habitat mapping of the USGS GAP Analysis Program could be adapted to provide the ancillary tie-breaker data required by the LCCP to conduct change detection between disparate land cover data sources heretofore considered too incompatible for that purpose. In Chapter 2, measures of urban sprawl, habitat loss and fragmentation in Florida were estimated using the corrected land cover data for 1987 and 2003. The Northwest and North regions of the state exhibited significantly higher indices of urban sprawl, habitat loss and habitat fragmentation via application of the Moran's I statistic. Reducing urban sprawl and habitat fragmentation spatial metrics to simple ordination variables through the use of non-metric multidimensional scaling produced new measures of urban sprawl and habitat fragmentation that correlated strongly with the original FRAGSTATS metrics, but could be more easily mapped and interpreted. Urban and Habitat ordination metrics were each spatially autocorrelated (Local Moran's I and K-means grouping analyses) but not correlated to each other using the Procrustes analysis PROTEST statistic (m2 = 0.952, p = 0.061). In contrast, individual urban sprawl metrics (CA, NP, LPI, ED, SHAPE_AM and DCAD) correlated with habitat fragmentation. NP and DCAD appeared to be particularly useful in predicting fragmentation, and county governments should take measures to reduce establishment of new urban patches to minimize NP and DCAD. Chapter 3 explored the relationship between environmental outcomes in habitat loss and fragmentation and the quality of county local government comprehensive plans. The use of NMS analysis provided a powerful technique for capturing the intrinsic variability of the Local Government Comprehensive Plan (LGCP) plan scoring systems of Brody (2003) and Pannozzo (2013) into a pair of variables each that could be used to explore associations with metrics of urban sprawl, habitat fragmentation and other county characteristics that influence urban growth and development. The geographic distribution of LGCP plan quality favored coastal counties with higher quality plans over inland counties, and there was some evidence that plans in Central and South regions of Peninsular Florida were superior to those in the North and Northwest Panhandle regions. Key factors in plan quality, specifically Coordination and Management, were strongly associated with urban sprawl or habitat fragmentation outcomes. The resources available to counties in the form of tax revenues, whether the county possessed a rural or urban economy, and the county's political makeup also appeared related to LGCP plan quality, urban sprawl or habitat fragmentation outcomes. More research will be needed to elucidate the specific causal mechanisms behind the implementation of local government planning that resulted in the observed environmental outcomes.