Qiusheng Wu

An object-based conceptual framework and computational method for representing and analyzing coastal morphological changes

This article presents an object-based conceptual framework and numerical algorithms for representing and analyzing coastal morphological and volumetric changes based on repeat airborne light detection and ranging (LiDAR) surveys. This method identifies and delineates individual zones of erosion and deposition as discrete objects. The explicit object representation of erosion and deposition zones is consistent with the perception and cognition of human analysts and geomorphologists. The extracted objects provide ontological and epistemological foundation to localize, represent, and interpret erosion and deposition patches for better coastal resource management and erosion control. The discrete objects are much better information carriers than the grid cells in the field-based representation of source data. A set of spatial and volumetric attributes are derived to characterize and quantify location, area, shape, orientation, depth, volume, and other properties of erosion and deposition objects. Compared with the conventional cell-by-cell differencing approaches, our object-based method gives a concise and high-level representation of information and knowledge about coastal morphological dynamics. The derived attributes enable the discrimination of true morphological changes from artifacts caused by data noise and processing errors. Furthermore, the concise object representation of erosion and deposition zones facilitates overlay analysis in conjunction with other GIS data layers for understanding the causes and impacts of morphological and volumetric changes. We have implemented a software tool for our object-based morphological analysis, which will be freely available for the public. An example is used to demonstrate the utility and effectiveness of this new method.

Delineation and Quantification of Wetland Depressions in the Prairie Pothole Region of North Dakota

The Prairie Pothole Region of North America is characterized by numerous, small, wetland depressions that perform important ecological and hydrological functions. Recent studies have shown that total wetland area in the region is decreasing due to cumulative impacts related to natural and anthropogenic changes. The impact of wetland losses on landscape hydrology is an active area of research and management. Various spatially distributed hydrologic models have been developed to simulate effects of wetland depression storage on peak river flows, frequently using dated geospatial wetland inventories. We describe an innovative method for identifying wetland depressions and quantifying their nested hierarchical bathymetric/topographic structure using high-resolution light detection and ranging (LiDAR) data. This contour tree method allows identified wetland depressions to be quantified based on their dynamic filling-spilling-merging hydrological processes. In addition, wetland depression properties, such as surface area, maximum depth, mean depth, storage volume, etc., can be computed for each component of a depression as well as the compound depression. We successfully applied the proposed method to map wetland depressions in the Little Pipestem Creek watershed in North Dakota. The methods described in this study will provide more realistic and higher resolution data layers for hydrologic modeling and other studies requiring characterization of simple and complex wetland depressions, and help prioritize conservation planning efforts for wetland resources.

Automated extraction of ground surface along urban roads from mobile laser scanning point clouds

ABSTRACT Extracting ground surface from high-density point clouds collected by Mobile Laser Scanning (MLS) systems is of vital importance in urban planning and digital city mapping. This article proposes a novel approach for automated extraction of ground surface along urban roads from MLS point clouds. The approach, which was designed to handle both ordered and unordered MLS point clouds, consists of three key steps: constructing vertical profile from MLS point clouds along the vehicle trajectory; extracting candidate ground points using an adaptive alpha shapes algorithm; refining the candidate ground points with an elevation variance filter. To evaluate the performance of the proposed method, experiments were conducted using two types of urban street-scene point clouds. The results reveal that the ground points can be detected with an error rate of as low as 1.9%, proving that our proposed method offers a promising solution for automated extraction of ground surface from MLS point clouds.

A localized contour tree method for deriving geometric and topological properties of complex surface depressions based on high-resolution topographical data

Surface depressions are abundant in topographically complex landscapes, and they exert significant influences on hydrological, ecological, and biogeochemical processes at local and regional scales. The increasing availability of high-resolution topographical data makes it possible to resolve small surface depressions. By analogy with the reasoning process of a human interpreter to visually recognize surface depressions from a topographic map, we developed a localized contour tree method that is able to fully exploit high-resolution topographical data for detecting, delineating, and characterizing surface depressions across scales with a multitude of geometric and topological properties. In this research, we introduce a new concept ‘pour contour’ and a graph theory-based contour tree representation for the first time to tackle the surface depression detection and delineation problem. Beyond the depression detection and filling addressed in the previous raster-based methods, our localized contour tree method derives the location, perimeter, surface area, depth, spill elevation, storage volume, shape index, and other geometric properties for all individual surface depressions, as well as the nested topological structures for complex surface depressions. The combination of various geometric properties and nested topological descriptions provides comprehensive and essential information about surface depressions across scales for various environmental applications, such as fine-scale ecohydrological modeling, limnological analyses, and wetland studies. Our application example demonstrated that our localized contour tree method is functionally effective and computationally efficient.

A New Approach for Detecting Urban Centers and Their Spatial Structure With Nighttime Light Remote Sensing

Urban spatial structure affects many aspects of urban functions and has implications for accessibility, environmental sustainability, and public expenditures. During the urbanization process, a careful and efficient examination of the urban spatial structure is crucial. Different from the traditional approach that relies on population or employment census data, this research exploits the nighttime light (NTL) intensity of the earth surface recorded by satellite sensors. The NTL intensity is represented as a continuous mathematical surface of human activities, and the elemental features of urban structures are identified by analogy with earth’s topography. We use a topographical metaphor of a mount to identify an urban center or subcenter and the surface slope to indicate an urban land-use intensity gradient. An urban center can be defined as a continuous area with higher concentration or density of employments and human activities. We successfully identified 33 urban centers, delimited their corresponding boundaries, and determined their spatial relations for Shanghai metropolitan area, by developing a localized contour tree method. In addition, several useful properties of the urban centers have been derived, such as 9% of Shanghai administrative area has become urban centers. We believe that this method is applicable to other metropolitan regions at different spatial scales.

An Extended Minimum Spanning Tree method for characterizing local urban patterns

ABSTRACT Detailed and precise information on urban building patterns is essential for urban design, landscape evaluation, social analyses and urban environmental studies. Although a broad range of studies on the extraction of urban building patterns has been conducted, few studies simultaneously considered the spatial proximity relations and morphological properties at a building-unit level. In this study, we present a simple and novel graph-theoretic approach, Extended Minimum Spanning Tree (EMST), to describe and characterize local building patterns at building-unit level for large urban areas. Building objects with abundant two-dimensional and three-dimensional building characteristics are first delineated and derived from building footprint data and high-resolution Light Detection and Ranging data. Then, we propose the EMST approach to represent and describe both the spatial proximity relations and building characteristics. Furthermore, the EMST groups the building objects into different locally connected subsets by applying the Gestalt theory-based graph partition method. Based on the graph partition results, our EMST method then assesses the characteristics of each building to discover local patterns by employing the spatial autocorrelation analysis and homogeneity index. We apply the proposed method to the Staten Island in New York City and successfully extracted and differentiated various local building patterns in the study area. The results demonstrate that the EMST is an effective data structure for understanding local building patterns from both geographic and perceptual perspectives. Our method holds great potential for identifying local urban patterns and provides comprehensive and essential information for urban planning and management.

Prediction of Water Depth From Multispectral Satellite Imagery—The Regression Kriging Alternative

Bathymetric information is crucial to the study and management of coastal zones. Passive remote sensing provides a cost-effective alternative to acoustic surveys and bathymetric LiDAR techniques. Most previous studies estimated water depth from multispectral imagery in shallow coastal and inland waters by establishing the relationship between image pixel spectral values and known water depth measurements, in which the log-linear inversion model is most widely used. Given a set of known water depth sample points, a bathymetric grid/map can be created by using a spatial interpolation technique. However, when a limited number of water depth sample points are available, the interpolation result is often unsatisfactory for portraying benthic morphology. In this letter, we propose to use the regression kriging (RK) approach to combine the optimal spatial interpolation of kriging with the high-resolution auxiliary information of multispectral imagery for a detailed bathymetric mapping. A case study has been performed to demonstrate and evaluate the performance of the RK method in comparison with ordinary kriging and log-linear inversion methods. It shows that the RK method can produce more accurate water depth estimations than the log-linear inversion method due to the account of the spatial pattern of the modeling residuals. The bathymetric grid created from the RK contains much more spatial details about the ocean floor morphology than that from the ordinary kriging owing to the incorporation of auxiliary information from multispectral satellite imagery.

Comparing Pixel- and Object-Based Approaches in Effectively Classifying Wetland-Dominated Landscapes

Wetland ecosystems straddle both terrestrial and aquatic habitats, performing many ecological functions directly and indirectly benefitting humans. However, global wetland losses are substantial. Satellite remote sensing and classification informs wise wetland management and monitoring. Both pixel- and object-based classification approaches using parametric and non-parametric algorithms may be effectively used in describing wetland structure and habitat, but which approach should one select? We conducted both pixel- and object-based image analyses (OBIA) using parametric (Iterative Self-Organizing Data Analysis Technique, ISODATA, and maximum likelihood, ML) and non-parametric (random forest, RF) approaches in the Barguzin Valley, a large wetland (~500 km2) in the Lake Baikal, Russia, drainage basin. Four Quickbird multispectral bands plus various spatial and spectral metrics (e.g., texture, Non-Differentiated Vegetation Index, slope, aspect, etc.) were analyzed using field-based regions of interest sampled to characterize an initial 18 ISODATA-based classes. Parsimoniously using a three-layer stack (Quickbird band 3, water ratio index (WRI), and mean texture) in the analyses resulted in the highest accuracy, 87.9% with pixel-based RF, followed by OBIA RF (segmentation scale 5, 84.6% overall accuracy), followed by pixel-based ML (83.9% overall accuracy). Increasing the predictors from three to five by adding Quickbird bands 2 and 4 decreased the pixel-based overall accuracy while increasing the OBIA RF accuracy to 90.4%. However, McNemar’s chi-square test confirmed no statistically significant difference in overall accuracy among the classifiers (pixel-based ML, RF, or object-based RF) for either the three- or five-layer analyses. Although potentially useful in some circumstances, the OBIA approach requires substantial resources and user input (such as segmentation scale selection—which was found to substantially affect overall accuracy). Hence, we conclude that pixel-based RF approaches are likely satisfactory for classifying wetland-dominated landscapes.

Decision-Tree, Rule-Based, and Random Forest Classification of High-Resolution Multispectral Imagery for Wetland Mapping and Inventory

Efforts are increasingly being made to classify the world’s wetland resources, an important ecosystem and habitat that is diminishing in abundance. There are multiple remote sensing classification methods, including a suite of nonparametric classifiers such as decision-tree (DT), rule-based (RB), and random forest (RF). High-resolution satellite imagery can provide more specificity to the classified end product, and ancillary data layers such as the Normalized Difference Vegetation Index, and hydrogeomorphic layers such as distance-to-a-stream can be coupled to improve overall accuracy (OA) in wetland studies. In this paper, we contrast three nonparametric machine-learning algorithms (DT, RB, and RF) using a large field-based dataset (n = 228) from the Selenga River Delta of Lake Baikal, Russia. We also explore the use of ancillary data layers selected to improve OA, with a goal of providing end users with a recommended classifier to use and the most parsimonious suite of input parameters for classifying wetland-dominated landscapes. Though all classifiers appeared suitable, the RF classification outperformed both the DT and RB methods, achieving OA >81%. Including a texture metric (homogeneity) substantially improved the classification OA. However, including vegetation/soil/water metrics (based on WorldView-2 band combinations), hydrogeomorphic data layers, and elevation data layers to increase the descriptive content of the input parameters surprisingly did not markedly improve the OA. We conclude that, in most cases, RF should be the classifier of choice. The potential exception to this recommendation is under the circumstance where the end user requires narrative rules to best manage his or her resource. Though not useful in this study, continuously increasing satellite imagery resolution and band availability suggests the inclusion of ancillary contextual data layers such as soil metrics or elevation data, the granularity of which may define its utility in subsequent wetland classifications.

Downscaling land surface temperature data by fusing Suomi NPP-VIIRS and landsat-8 TIR data

ABSTRACT Land surface temperature (LST) is a key parameter of great interest in many remote sensing applications. However, no single satellite system can produce thermal infrared (TIR) images at both high spatial and temporal resolution to retrieve LST. Various algorithms have been developed to enhance the spatial or temporal resolution of TIR data in the past decades. Among them, the Spatio-temporal Adaptive Data Fusion Algorithm for Temperature mapping (SADFAT) model is one of the most widely used algorithms for fusing Landsat and Moderate Resolution Imaging Spectroradiometer (MODIS) imagery. To our knowledge, Visible Infrared Imaging Radiometer Suite (VIIRS) TIR data have not yet been used in thermal downscaling with Landsat-8 TIR data. This study aims to generate daily LST images at Landsat-8 resolution (100 m) by fusing VIIRS and Landsat-8 TIR data for the first time with the SADFAT algorithm. The results indicate that the prediction accuracy for the study area ranged from 1.1 K to 1.4 K, which suggests that VIIRS data can be used as a good alternative for MODIS data for generating daily LST images by fusing Landsat TIR data.