Weiqi Zhou

An object-oriented approach for analysing and characterizing urban landscape at the parcel level

This paper presents an object‐oriented approach for analysing and characterizing the urban landscape structure at the parcel level using high‐resolution digital aerial imagery and LIght Detection and Ranging (LIDAR) data. Additional spatial datasets including property parcel boundaries and building footprints were used to both facilitate object segmentation and obtain greater classification accuracy. The study area is the Gwynns Falls watershed, which includes portions of Baltimore City and Baltimore County, MD. A three‐level hierarchical network of image objects was generated, and objects were classified. At the two lower levels, objects were classified into five classes, building, pavement, bare soil, fine textured vegetation and coarse textured vegetation, respectively. The object‐oriented classification approach proved to be effective for urban land cover classification. The overall accuracy of the classification was 92.3%, and the overall Kappa statistic was 0.899. Land cover proportions as well as vegetation characteristics were then summarized by property parcel. This exercise resulted in a knowledge base of rules for urban land cover classification, which could potentially be applied to other urban areas.

Trees grow on money: Urban tree canopy cover and environmental justice

This study examines the distributional equity of urban tree canopy (UTC) cover for Baltimore, MD, Los Angeles, CA, New York, NY, Philadelphia, PA, Raleigh, NC, Sacramento, CA, and Washington, D.C. using high spatial resolution land cover data and census data. Data are analyzed at the Census Block Group levels using Spearman’s correlation, ordinary least squares regression (OLS), and a spatial autoregressive model (SAR). Across all cities there is a strong positive correlation between UTC cover and median household income. Negative correlations between race and UTC cover exist in bivariate models for some cities, but they are generally not observed using multivariate regressions that include additional variables on income, education, and housing age. SAR models result in higher r-square values compared to the OLS models across all cities, suggesting that spatial autocorrelation is an important feature of our data. Similarities among cities can be found based on shared characteristics of climate, race/ethnicity, and size. Our findings suggest that a suite of variables, including income, contribute to the distribution of UTC cover. These findings can help target simultaneous strategies for UTC goals and environmental justice concerns.

Can Money Buy Green? Demographic and Socioeconomic Predictors of Lawn-Care Expenditures and Lawn Greenness in Urban Residential Areas

It is increasingly important to understand how household characteristics influence lawn characteristics, as lawns play an important ecological role in human-dominated landscapes. This article investigates household and neighborhood socioeconomic characteristics as predictors of residential lawn-care expenditures and lawn greenness. The study area is the Gwynns Falls watershed, which includes portions of Baltimore City and Baltimore County, MD. We examined indicators of population, social stratification (income, education and race), lifestyle behavior, and housing age as predictors of lawn-care expenditures and lawn greenness. We also tested the potential of PRIZM market cluster data as predictors for these two dependent variables. Lawn greenness was found to be significantly associated with lawn-care expenditures, but with a relatively weak positive correlation. We also found lifestyle behavior indicators to be the best predictors for both dependent variables. PRIZM data, especially the lifestyle segmentation, also proved to be useful predictors for both.

Comparing Machine Learning Classifiers for Object-Based Land Cover Classification Using Very High Resolution Imagery

This study evaluates and compares the performance of four machine learning classifiers—support vector machine (SVM), normal Bayes (NB), classification and regression tree (CART) and K nearest neighbor (KNN)—to classify very high resolution images, using an object-based classification procedure. In particular, we investigated how tuning parameters affect the classification accuracy with different training sample sizes. We found that: (1) SVM and NB were superior to CART and KNN, and both could achieve high classification accuracy (>90%); (2) the setting of tuning parameters greatly affected classification accuracy, particularly for the most commonly-used SVM classifier; the optimal values of tuning parameters might vary slightly with the size of training samples; (3) the size of training sample also greatly affected the classification accuracy, when the size of training sample was less than 125. Increasing the size of training samples generally led to the increase of classification accuracies for all four classifiers. In addition, NB and KNN were more sensitive to the sample sizes. This research provides insights into the selection of classifiers and the size of training samples. It also highlights the importance of the appropriate setting of tuning parameters for different machine learning classifiers and provides useful information for optimizing these parameters.

Mapping the concentrations of total suspended matter in Lake Taihu, China, using Landsat‐5 TM data

Remote sensing techniques can be used to estimate and map the concentrations of suspended matter in inland water, providing both spatial and temporal information. Although an empirical approach to remote sensing of inland waters has been carried out frequently, satellite imagery has not been incorporated into routine lake monitoring programmes due in part to the lack of a standard prediction equation with multi‐temporal capacity for suspended matter. Empirical and physical models must be developed for each lake and its corresponding turbidity composition if they are to be compared over time, or with other bodies of water. This study aimed to develop and apply multi‐temporal models to estimate and map the concentrations of total suspended matter (TSM) in Lake Taihu, China. Two Landsat‐5 Thematic Mapper (TM) images and nearly contemporaneous in situ measurements of TSM were used. A modified Dark‐Object Subtraction (DOS) method was used, and appeared to be adequate for atmospheric correction. The relationships were examined between TSM concentrations and atmospherically corrected TM band and band ratios. Results of this study show that the ratio TM4/TM1 has a strong relationship with TSM concentrations for lake waters with relatively low concentrations of phytoplankton algae. However, TM3 provided a strong predictive relationship with TSM concentrations despite varied water quality conditions. Different prediction models were developed and compared using multiple regression analysis. The Akaike Information Criteria (AIC) approach was used to choose the best models. The validation of the multi‐temporal capability of the best models indicated that it is feasible to apply the linear regression model using TM3 to estimate TSM concentrations across time in Lake Taihu, even if no in situ data were available.

Increasing impact of urban fine particles (PM2.5) on areas surrounding Chinese cities

The negative impacts of rapid urbanization in developing countries have led to a deterioration in urban air quality, which brings increasing negative impact to its surrounding areas (e.g. in China). However, to date there has been rare quantitative estimation of the urban air pollution to its surrounding areas in China.We thus evaluated the impact of air pollution on the surrounding environment under rapid urbanization in Chinese prefectures during 1999 – 2011. We found that: (1) the urban environment generated increasing negative impact on the surrounding areas, and the PM2.5 concentration difference between urban and rural areas was particularly high in large cities. (2) Nearly half of the Chinese prefectures (156 out of 350) showed increased impact of urban PM2.5 pollution on its surrounding areas. Those prefectures were mainly located along two belts: one from northeast China to Sichuan province, the other from Shanghai to Guangxi province. Our study demonstrates the deterioration in urban air quality and its potential impacts on its surrounding areas in China. We hope that the results presented here will encourage different approaches to urbanization to mitigate the negative impact caused by urban air pollution, both in China and other rapidly developing countries.

An Object-Based Approach for Urban Land Cover Classification: Integrating LiDAR Height and Intensity Data

Digital surface models (DSMs) derived from light detection and ranging (LiDAR) data have been increasingly integrated with high-resolution multispectral satellite/aerial imagery for urban land cover classification. Fewer studies, however, have investigated the usefulness of LiDAR intensity in aid of urban land cover classification, particularly in highly developed urban settings. In this letter, we use an object-based classification approach to investigate whether a combination of LiDAR height and intensity data can accurately map urban land cover. We further compare the approach to a method that uses multispectral imagery as the primary data source, but LiDAR DSM as ancillary data to aid in classification. The study site is a suburban area in Baltimore County, MD. The LiDAR data were acquired in March 2005, from which DSM and two intensity layers (first and last returns), with 1-m spatial resolution were generated, respectively. Four classes were included: 1) buildings; 2) pavement; 3) trees and shrubs; and 4) grass. Our results indicated that the object-based approach provided flexible and effective means to integrate LiDAR height and intensity data for urban land cover classification. A combination of the LiDAR height and intensity data proved to be effective for urban land cover classification. The overall accuracy of the classification was 90.7%, and the overall Kappa statistics equaled 0.872, with the users and producers accuracies ranging from 86.8% to 93.6%. The accuracy of the results were far better than those using multispectral imagery alone, and comparable to using DSM data in combination with high-resolution multispectral satellite/aerial imagery.

Modeling Residential Lawn Fertilization Practices: Integrating High Resolution Remote Sensing with Socioeconomic Data

This article investigates how remotely sensed lawn characteristics, such as parcel lawn area and parcel lawn greenness, combined with household characteristics, can be used to predict household lawn fertilization practices on private residential lands. This study involves two watersheds, Glyndon and Baisman’s Run, in Baltimore County, Maryland, USA. Parcel lawn area and lawn greenness were derived from high-resolution aerial imagery using an object-oriented classification approach. Four indicators of household characteristics, including lot size, square footage of the house, housing value, and housing age were obtained from a property database. Residential lawn care survey data combined with remotely sensed parcel lawn area and greenness data were used to estimate two measures of household lawn fertilization practices, household annual fertilizer nitrogen application amount (N_yr) and household annual fertilizer nitrogen application rate (N_ha_yr). Using multiple regression with multi-model inferential procedures, we found that a combination of parcel lawn area and parcel lawn greenness best predicts N_yr, whereas a combination of parcel lawn greenness and lot size best predicts variation in N_ha_yr. Our analyses show that household fertilization practices can be effectively predicted by remotely sensed lawn indices and household characteristics. This has significant implications for urban watershed managers and modelers.

90 years of forest cover change in an urbanizing watershed: spatial and temporal dynamics

Landscape structure in the Eastern US experienced great changes in the last century with the expansion of forest cover into abandoned agricultural land and the clearing of secondary forest cover for urban development. In this paper, the spatial and temporal patterns of forest cover from 1914 to 2004 in the Gwynns Falls watershed in Baltimore, Maryland were quantified from historic maps and aerial photographs. Using a database of forest patches from six times—1914, 1938, 1957, 1971, 1999, and 2004—we found that forest cover changed, both temporally and spatially. While total forest area remained essentially constant, turnover in forest cover was very substantial. Less than 20% of initial forest cover remained unchanged. Forest cover became increasingly fragmented as the number, size, shape, and spatial distribution of forest patches within the watershed changed greatly. Forest patch change was also analyzed within 3-km distance bands extending from the urban core to the more suburban end of the watershed. This analysis showed that, over time, the location of high rates of forest cover change shifted from urban to suburban bands which coincides with the spatial shift of urbanization. Forest cover tended to be more stable in and near the urban center, whereas forest cover changed more in areas where urbanization was still in process. These results may have critical implications for the ecological functioning of forest patches and underscore the need to integrate multi-temporal data layers to investigate the spatial pattern of forest cover and the temporal variations of that spatial pattern.

Fine particulate (PM2.5) dynamics during rapid urbanization in Beijing, 1973-2013.

PM2.5 has been given special concern in recent years when the air quality monitoring station started recording. However, long-term PM2.5 concentration dynamic analysis cannot be taken with the limited observations. We therefore estimated the PM2.5 concentration using meteorological visibility data in Beijing. We found that 71 ± 17% of PM10 were PM2.5, which contributed to visibility impairment (y = 332.26e−0.232x; R2 = 0.75, P < 0.05). We then reconstructed a time series of annual PM2.5 from 1973 to 2013, and examined its relationship with urbanization by indicators of population, gross domestic production (GDP), energy consumption, and number of vehicles. Concluded that 1) Meteorological conditions were not the major cause of PM2.5 increase from 1973 to 2013; 2) With population and GDP growth, PM2.5 increased significantly (R2 = 0.5917, P < 0.05; R2 = 0.5426, P < 0.05); 3) Intensive human activity could change air quality in a short period, as observed changes in the correlations of PM2.5 concentration with energy consumption and number of vehicles before and after 2004, respectively. The success of this research provides an easy way in reconstructing long-term PM2.5 concentration with limited PM2.5 observation and meteorological visibility, and insight the impact of urbanization on air quality.