Projects

Purposes: Investigate how social inequality relates to intra-urban traffic CO2 emissions by incorporating the fine-grained mobility flows using smart data

Methods: The research design includes four major: 

1) Extract mobility flows from individual GPS tracks linked to home locations based on mobile phone app data (Huq), summarize the O-D (Origin-Destination) flows, incorporate the O-D flows into SWM, and compute the 2D and 3D urban form factors as control variables.

2) Estimate the traffic CO2 produced in available DataZones based on traffic count, vehicle types, and emission factors.

3) Test the spatial autocorrelation and construct the spatial panel models to predict traffic CO2 emissions based on contributing factors in each focal and nearby DataZones based on a customized SWM. 

4) Interpret the environmental justice through hot spot analysis, mobility flows visualization, and Gini coefficients.

Results: The results demonstrate that models based on customized SWM with real mobility better predict traffic CO2 than traditional distance-based models. According to model results, income and car ownership rates are dominant factors associated with traffic CO2. Noticeably, traffic CO2 emissions are closely related to incoming mobility flows from neighborhoods with high income and car ownership rates. Moreover, the 20% top-ranked urban areas in terms of income and car ownership are responsible for 37.21% and 49.52% of total traffic CO2, respectively, indicating that disadvantaged groups bear the costs of emissions disproportionately generated by residents of wealthier areas. Finally, urban planners should not only consider reducing traffic emissions but also ensure that disadvantaged residents benefit from improved transport systems to mitigate emission inequality. 

This work, as the first author, has been under 2nd review of the journal "Cities".

Purposes: Investigate the combined effects of local meteorological, morphological, and socioeconomic factors on air quality in 244 prefectural-level Chinese cities. 

Methods: We leverage an improved Apriori algorithm to mine association rules between these factors and their combined effects on air pollution concentration. Unlike the traditional Apriori algorithm, which requires repetitive database scans to generate candidates, consuming substantial time and memory, our study introduces lift as an evaluation metric. This significantly reduces the algorithm's runtime and allows for more effective rule filtering, yielding more persuasive rules, and enhance the applicability and contribution to environmental protection strategies.

Results: Secondary industry in GDP and saturation vapor pressure strongly relate to air quality. Severe air pollution occurs when urban development is coupled with reduced green areas and high temperatures, confirming that a single factor cannot predict air quality well. For example, we find that combining low population, low regional GDP, high maximum temperatures, and longer roads worsens air quality in small urban built-up areas. Additionally, temperature and altitude differences associate with highway passenger volume, regional GDP, and population differently. 

This work, as the corresponding author finished by my student, has been published at "Transactions in GIS" (https://onlinelibrary.wiley.com/doi/10.1111/tgis.13156).

Purposes:  Prediction of intra-urban PM (Particulate Matter) through mobile monitoring

Methods: Cooperated with the “Senseable City Lab” at MIT and participated in the “City Scanner” project (http://senseable.mit.edu/cityscanner), we collected PM2.5 and NC10 (number concentration for particles with diameters of 1 μm-10 μm) through mobile monitoring in the Bronx, New York City. To quantify the urban context comprehensively, I incorporated the volunteered geographic information, Open Street Map, and crowdsourced data, Microsoft Building Footprint Data, to calculate the 2-D and 3-D urban morphological characteristics, respectively.  I also developed the "wind-sensitive" model to consider the local wind effects on the urban form-air quality relationship. Particularly, as a crucial component of 3D urban form, I improved the traditional FAI (Frontal Areal Index) by considering the remaining windward areas for the rest of the buildings in the urban plot.

Results: This research revealed the environmental justice issues by finding that residents in the southern Bronx had a higher exposure risk than in other regions. In addition, 3-D urban morphological factors showed higher feature importance than the 2-D counterparts for PM2.5 prediction according to the Random Forest. Besides, using mobile measurements from low-cost sensors had the potential to develop fine-scale granular maps of PM concentrations in cities around the world.  

This work, as the first author, has been published in the journal "Environmental Science & Technology" (https://pubs.acs.org/doi/full/10.1021/acs.est.1c04854).

Purposes:  Prediction of intra-urban NO2 through mobile monitoring by incorporating tree diversity

Methods: We collected the NO2 in the Bronx in the "City Scanner" project at "Senseable City Lab" at MIT. To increase the model robustness, I also obtained the mobile monitoring NO2 in Oakland, CA from the volunteered geographic information platform, OpenAQ. To better characterize the local context, I incorporated street trees from the crowdsourced database, OpenTrees, and incorporated Simpson and Shannon index into air quality models.  In addition, I quantified the maximum value of the spatial heterogeneity for each land use type through the “Lacunarity” analysis and used it as the upper bounds of the buffer radius.  To better characterize the local context, I incorporated street trees from the crowdsourced database, OpenTrees, and incorporated Simpson and Shannon index into air quality models. 

Results:  Using "Lacunarity" analysis greatly saved computational resources for extracting information in large areas and proved to keep a similar level of accuracy as the conventional Land Use Regression, simultaneously.  Besides, it was necessary to strike a balance between the richness and dominance of tree species to achieve overall better air quality on the city scale. 

This work, as the first author, has been published in the journal "Applied Geography" (https://doi.org/10.1016/j.apgeog.2023.102943).

Purposes: Quantify individual exposure to air pollutants through field measurements

Methods: In this project, I used the mobile sensor "AirBeam2" to collect the PM1, PM2.5, and PM10 in the Atlanta downtown area with high spatial (<1m) and temporal (900ms) resolution and analyzed their synergic effects on local meteorology on air pollution distribution. I also compared the model performances between Random Forest, Artificial Neural Network, and Multiple Linear Regression models with both fixed station data from EPA (Environmental Protection Agency) and mobile monitoring. To consider the wind direction effects, I divided the area into the windward and leeward areas around the buffer zones for each data point. The urban form was evaluated in both windward and leeward areas, respectively. 

Results: The results showed that southern Atlanta was exposed to a higher level of air pollutants than their northern counterparts, which reconfirmed the spatial and social inequalities (e.g., most low-income people resided in southern Atlanta).  Besides, meteorological factors and urban form had synthetic effects on air pollution distribution because the impacting factors of air pollution varied in different wind subsections.

This work, as the first author, has been published in the journal "Landscape Ecology" (https://link.springer.com/article/10.1007/s10980-020-01094-6).

Purposes:  Estimation of air pollution distribution through urban morphological analysis 

Methods:  To improve the characterization of the urban context, I collected the POI (Point of Interests) data from Gaode Maps, a popular app that the Chinese use to navigate. Besides, I calculated the MBI (Morphological Building Index), an index to detect buildings on high-resolution imagery, from ZY-3 (Chinese No.3 Resources Satellite) in Beijing, China. The NDVI (Normalized Differential Vegetation Index) and LST (Land Surface Temperature) were calculated from Landsat-8. All variables were added to the proposed “Spatiotemporal Autocorrelation” model.

Results: The results showed that residents in southern Beijing were exposed to a higher level of PM2.5, PM10, and SO2, which could attribute to the compact urban development and higher rates of traffic congestion and industrial emissions, and lower proportions of greenspace in southern areas. 

This work, as the first author, has been published in the journal "Computers, Environment and Urban Systems" (https://www.sciencedirect.com/science/article/pii/S0198971518301960) and got the Best Paper Award (2nd place) at the American Association of Geography in 2019.

Purposes:  Applying crowdsourced data to estimate the mobility patterns of outdoor exercisers

Methods: I collected the individual routes of outdoor exercisers from crowdsourced data, Strava, a popular app with an active community, in Atlanta downtown. The Strava data can further divide the exercisers by gender, trip purposes, and age groups.o quantify the urban context around exercisers accurately, I extracted the most popular routes (corridors) through Strava and calculated both 2-D and 3-D urban form metrics within the buffer areas around each corridor. Particularly, I proposed a dynamic scheme of “Wind-Related Buffer” (the radius was computed by the product of the wind speed and the trip duration on each corridor) and assigned spatial measurements of urban form at a variety of scales to mitigate the “Modifiable Areal Unit Problem”. 

Results:  Urban form could impact outdoor exercisers by affecting location preferences and such impact varied among different subgroups. Besides, meteorological factors and urban form factors work together to influence outdoor exercise decisions. The results also revealed higher rates of Asthma-Related Emergency Room Visits and lower levels of outdoor exercise activities in southern Atlanta compared to their northern counterparts. In addition, places became attractive for exercisers only when multiple requirements were met (e.g., accessibility to public parks and proximity to residential communities).

This work, as the first author, has been published in the "Journal of Geographical Systems" (https://link.springer.com/article/10.1007/s10109-023-00424-x).

Purposes:  Understanding the interrelations between urban context, air quality, and traffic volume 

Methods: I collected traffic volume and air quality data from the fixed monitoring stations based on CCS (Continuous Count Stations) and EPA (Environmental Protection Agency), respectively, in Atlanta city. In addition, I compared the model performances among Random Forest, Random Tree, and M5 Model Tree within different spatiotemporal stratum by developing the method of “Spatiotemporal Stratification”. In this way, each variable and model would be summarized and constructed in their stratum, respectively. 

Results: This work revealed that the relationships between urban form, traffic volume, and air quality varied under different spatiotemporal conditions (e.g., workday or holiday, polluted or unpolluted zones). 

This work, as the first author, has been published in the journal "Environment and Planning B: Urban Analytics and City Science" (https://journals.sagepub.com/doi/abs/10.1177/2399808321995822). 

Purposes:  Estimation of fine-scale population distribution based on 3D urban buildings models

Methods: The building height was calculated through the relationship between solar angle and the building shadow length in high-resolution remote sensed imagery, ZY-3, and the population was redistributed by the average height per floor and living area per person. 

Results: Fine-scale population estimation was achieved by 3D reconstruction of urban residential buildings, and a deterministic model in a relatively large scope through a more feasible approach was proposed. This method does not need the classification of land use types for the model input and the final results show great potential in determining urban citizen distributions at finer resolutions in the future.

This work, as the corresponding author, has been published in the journal "Sensors" (https://www.mdpi.com/1424-8220/16/10/1755).