K. Huang

Publications

38 publications· 6 under review

View on Google Scholar ↗
5,055
Citations
22
h-index
28
i10-index
PhD, Yale
21 citations
34 citations
41 citations
38 citations
46 citations
55 citations
Postdoc, NCAR
176 citations
478 citations
Asst Prof, NYU Shanghai
650 citations
742 citations
818 citations
979 citations
910 citations
14
15
16
17
18
19
20
21
22
23
24
25
26
Data from Google Scholar, updated automatically

In press

  • Applies a fine-tuned BERT-based language model to 278 million geotagged social media posts to quantify how people perceive and express heat across 347 cities in 50 countries
  • Finds that heat perception extends well beyond summer: 35.4% of cities sustain multi-season perception peaks and 16.1% show counter-seasonal regimes where winter perception surpasses summer
  • Shows spatial concentration of heat perception follows continent-specific modes — intensity-driven in Africa and variability-driven in South America
  • Demonstrates that co-occurring extreme temperature and elevated perception yields the strongest health associations, especially for adults aged ≥70 (mortality RR = 1.479) and respiratory mortality (RR = 1.826)
  • Establishes expressed heat perception as a scalable, season-resolved signal that complements temperature and identifies when heat becomes socially salient
heat perceptionurban heatclimate and healthsocial medialarge language modelBERTclimate adaptationheat mortality
Under review(6)
  • Assesses where a 100 MW computing load could be supported year-round by solar across 121 countries, combining monthly solar yield, cooling demand, land suitability, transmission, and connectivity
  • Identifies 6,280 cells with sufficient energy in every month (8,657 TWh yr⁻¹ across 100 countries under within-month balancing), with middle-income countries holding 55% of that potential
  • Partitions solar-powered AI potential by dual supply-chain access to Chinese energy hardware and US-controlled accelerators into three groups (open to both 40%; hardware barriers with open chips 31%; open hardware with chip-licensing constraints 29%)
  • Finds that under the baseline coding no country faces both supply-chain constraints at once, so solar-powered AI is limited by two different chains rather than a single bottleneck
data centerssolar energyartificial intelligencesupply chainsexport controlsCell Reports Sustainability
  • Applies a floor-area-weighted 3D mobility-distance metric to 6,582 cities and over 650 million buildings across 173 countries, showing that vertical urban structure can either compress or stretch inferred mobility space
  • Finds 3D mobility distances are smaller than 2D estimates in 79% of cities due to vertical intensification, but rapidly urbanizing megacities such as Shanghai and New Delhi exhibit 'volumetric sprawl' where vertical growth exceeds 2D estimates by 9–12%
  • Identifies stage-dependent scaling breakpoints at ~2.7 million residents and ~$12,476 per capita, beyond which the mobility-distance scaling exponent declines from 0.52 to 0.36, marking a transition toward more efficient urban organization
  • Shows that the mitigation potential of vertical urban form is strongest during early-to-mid development, with marginal gains diminishing in high-income large cities where volumetric distribution and spatial organization become the primary determinants of mobility efficiency
3D urban formurban scalingmobility distancetransport energysustainable urbanizationvertical urban growthvolumetric sprawl
  • Moves beyond land-area exposure to a multidimensional assessment of how future coastal inundation affects the built environment
  • Evaluates future impacts on both built-environment exposure and the material stock embedded in buildings and infrastructure
  • Informs coastal adaptation by indicating what is at stake in built assets and materials, not just inundated land
coastal inundationsea-level risematerial stockbuilt environmentcoastal exposureclimate adaptationmultidimensional assessment
  • Assesses how sea-level rise and vertical land motion jointly reshape coastal built-environment exposure patterns, using Tianjin as a case study
  • Integrates land subsidence and relative sea-level change to map shifting exposure of coastal buildings and infrastructure
  • Highlights implications for coastal adaptation planning where vertical land motion amplifies or redistributes flood and inundation risk
sea-level risevertical land motioncoastal exposurebuilt environmentTianjinland subsidenceclimate adaptation
  • Evaluates material-saving strategies — including material efficiency, lightweighting, and longer building lifetimes — as levers for decarbonizing the global building stock
  • Links reductions in construction-material demand to lower embodied carbon emissions across the world's buildings
  • Offers a global-scale perspective on how dematerialization can complement energy and fuel measures in building-sector climate mitigation
building stockmaterial efficiencydecarbonizationembodied carbonconstruction materialsclimate mitigationdematerialization
  • Examines the mismatch between infrastructure reach and capacity pressures driven by rapid urban expansion across sub-Saharan Africa
sub-Saharan Africaurban expansioninfrastructureurban growthdevelopmentNature Communications

Published

2026

  • At the neighborhood scale, doubling population density only increases built mass by about 68% (δ ≈ 0.75) — economies of scale intensify inside the city versus city-level scaling (β ≈ 0.90)
  • Analyzes over 3,000 cities globally to reveal universal sub-linear scaling of built mass with population, with a city-level scaling exponent of β ≈ 0.90 meaning larger cities use less built mass per capita
  • Proposes a theoretical framework showing that city-level sub-linear scaling emerges from the aggregation of within-city neighborhood-level disparities, establishing a nested multi-scale scaling relationship
  • Redefines 'over-built' and 'under-built' conditions as deviations from expected scaling patterns, linking excess built mass to environmental impact and deficient built mass to inadequate living standards
Animation: nested economies of scale in global city mass
urban scaling lawsZipf's lawcity size distributionbuilt massurban allometryurban economics
  • Integrating building height and protection standards shifts the global flood damage distribution: Southeast Asia accounts for 42% of damages versus only 15% under conventional flood-depth-only assessments
  • Sole reliance on flood depth overestimates risk in regions with resilient vertical urban forms and robust protection (e.g., Europe, East Asia) while underestimating it in areas with low-rise, unprotected development
  • Chinese cities show strong dependence on protection systems that can reduce potential flood damages by over 90%, as illustrated by Wuhan where damages drop to 0.12%–0.17% after accounting for its 100-year protection standard
  • Countries such as Vietnam, Cambodia, Bangladesh, Thailand, and Pakistan show high vulnerability, with more than 30% of building footprints located within floodplains
urban flood risk3D flood assessmentbuilding heightflood protection standardsclimate riskglobal flood modeling
  • Uses 1.5 million mobile phone records from 254 urban parks in Tokyo to classify parks by temporal visitation rhythms across daily, weekly, and seasonal scales, challenging traditional static size-based park classification
  • Reveals that park visitation patterns are shaped by user preference variation and accessibility barriers, with parks rhythmized by seasonal events and daily routines such as commuting and exercise
  • Proposes four modes of action for temporal park planning: year-round activity tracing, temporary and tactical designs, inclusive park systems for spatiotemporal equity, and public engagement cultivating a sense of time and identity
  • Demonstrates how temporal insights inform targeted design interventions, such as fitness infrastructure for morning-popular neighborhood parks and extended operating hours for long-evening central parks
urban parkspark planningtemporal classificationurban recreationpublic space designurban planning
  • Introduces the Urban Homogeneous Scale (UHS) metric — the spatial scale of functionally homogeneous building clusters — as a novel urban morphology measure for understanding urban heat island effects
  • Analyzes 232,874 observations across 101 global cities using machine learning (XGBoost) building classification and Double Machine Learning causal inference to establish UHS-heat relationships
  • Finds that a 100-metre increase in UHS leads to approximately 5% decrease in daytime urban heat island intensity, with daytime cooling more pronounced in inland, humid, and high-income regions
  • Proposes homogeneous scale planning as a land-use strategy for mitigating urban heat, with implications for mixed-use zoning and urban design guidelines
urban heat islandUrban Homogeneous Scaleurban morphologyclimate adaptationurban planningcausal inferencemachine learning

2025

  • Across the 1,000 largest global cities (2003–2020), declining urban population density attenuates the rise in population-weighted surface urban heat island extremes by approximately 0.5 K per decade per 1,000 persons/km²/year decrease in density
  • Cities undergoing densification face disproportionately greater increases in population-weighted exposure to surface heat extremes compared to area-average trends
  • Vegetation changes show a moderate association with population-weighted heat exposure (R² = 0.56), while variations in surface albedo exhibit a negligible relationship (R² = 0.03)
  • Highlights the need for strategic urban planning that balances compact development co-benefits against amplified heat exposure risks in rapidly densifying cities
urban densityheat exposuresurface urban heat islandurban sprawlclimate equitysatellite land-surface temperature
  • Implements a probabilistic inundation mapping method on Google Earth Engine using Sentinel-1 SAR data and land-cover-specific local thresholds derived from fitted Gamma distributions
  • Achieves strong accuracy with Brier Scores of 0.05–0.07, Logarithmic Loss of 0.1–0.2, and Expected Calibration Error of 0.03–0.04 across flood events on five continents
  • Produces pixel-level inundation probability estimates rather than binary classification, enabling uncertainty quantification in flood mapping
  • Enables rapid, scalable flood mapping through Google Earth Engine with user-refinable local thresholds for operational disaster response across diverse geographic settings
flood mappingSentinel-1 SARGoogle Earth Engineinundation detectionremote sensingprobabilistic mapping
  • Among 3,763 Shanghai residents surveyed during the 2022 COVID-19 lockdown, lower socioeconomic status was significantly associated with higher depressive symptoms and anxiety (B = 0.173 and B = 0.147, p < 0.001)
  • Neighborhood social capital, community management quality, and frequency of online social connections via WeChat partially mediated the association between socioeconomic status and mental health outcomes
  • Neighborhood social capital and online social connections also moderated SES-driven mental health inequalities, mitigating the disparity in depressive symptoms and anxiety among lower-SES residents
COVID-19 lockdownmental healthShanghaineighborhood environmentsocial connectionshealth equity
  • Short-term exposure to all six ambient air pollutants (PM2.5, PM10, NO2, CO, SO2, O3) was significantly associated with increased daily hospital visits for anxiety disorders in Qingyuan, Guangdong Province, even at concentrations below China's and WHO's standards
  • NO2 showed the strongest association (OR = 1.19, 95% CI: 1.05–1.34 per IQR increase at lag 0–7 days), followed by CO (OR = 1.14 at lag 0–2 days), with robust results across sensitivity analyses
  • Subgroup analyses revealed that females and middle-aged individuals were more susceptible to air-pollution-related anxiety disorder hospital visits than males and other age groups
air pollutionanxiety disordersPM2.5hospital visitsenvironmental healthsouthern China
  • For every 10% increase in urban tree cover, surface temperatures are reduced by 0.25 °C during the day and 0.04 °C at night, with trees in humid regions showing highest daytime cooling and arid zones the greatest nighttime cooling
  • While climatic factors contribute nearly twice as much to tree cooling efficiency as nonclimatic factors, the nonlinear effects of nonclimatic variables such as impervious surface coverage offer actionable optimization thresholds
  • Optimal impervious surface coverage for maximizing tree cooling is approximately 60% in arid regions and 40% in humid areas, providing practical guidance for climate-resilient nature-based urban cooling
urban tree coolinggreen infrastructureurban canopy managementnature-based coolingurban forestry
  • The urban heat island effect reduces global cold-related mortality by more than fourfold compared to the increase in heat-related mortality it causes, revealing a previously overlooked beneficial health effect during cold seasons
  • Widely implemented urban cooling strategies such as green infrastructure and reflective roofs can have adverse net effects in high-latitude cities while benefiting some tropical cities, highlighting the need for place-specific approaches
  • Proposes seasonal adjustments to roof albedo as an actionable strategy to simultaneously reduce both heat-related and cold-related mortality across diverse climate zones
urban heat mortalityurban overheatingheat-related deathsnighttime heatNature Climate Changepublic health
  • Urban moist heat in more than 2,300 Global South cities increased by 0.41 ± 0.01 °C per decade between 2003 and 2020, with urban browning contributing an additional 0.05 °C per decade in affected cities
  • The loss of urban vegetation in Global South cities compounds warming trends driven by climate change and urbanization, disproportionately affecting vulnerable populations in regions such as Nigeria
  • Introduces a new dataset for measuring urban moist heating rates, revealing that urban browning is an underrecognized driver of heat stress exacerbation in the developing world
urban browningheat stressGlobal Southvegetation lossurban greennessNature Cities

2024

  • Under SSP3-7.0, labor workability across 217 BRI projects in 70 countries could decrease by 31% by end of century, three times the 10% decline under SSP1-2.6, due to rising wet bulb globe temperatures
  • Under SSP2-4.5, floods with a historical 10-year return period are projected to occur every 5 years at BRI project sites, effectively doubling flood frequency
  • Regional flood risk varies substantially, with 100-year floods projected to recur every 15 years in South Asia and every 24 years in Sub-Saharan Africa
  • These largely overlooked climate risks threaten the commercial viability and loan repayment capacity of BRI infrastructure projects collectively valued at an estimated $1 trillion
Belt and Road Initiativeclimate riskheat riskflood riskinfrastructureinternational development
  • Global urban rivers emit an estimated 78.1 ± 3.5 Tg CO₂-equivalent annually, with GHG emissions per unit area nearly double those from non-urban rivers (~815 vs ~414 mmol CO₂-eq/m²/day)
  • Elevated CH₄ and N₂O emissions are driven by widespread eutrophication and altered carbon and nutrient cycling, with the highest emissions in lower-middle-income countries where pollution control is deficient
  • Highlights that pollution control in urban rivers is an underappreciated but important pathway for reducing urban greenhouse gas emissions, particularly in developing nations
urban riversgreenhouse gas emissionswater pollutionurban waterwaysclimate mitigation
  • Produces a building-scale height map across the contiguous United States circa 2020, revealing downtown areas average 12.4 m in height — more than double the 5.4 m average in suburban areas
  • Three-dimensional building volume analysis reveals a significantly larger urban-rural gap in living spaces than two-dimensional footprint analysis alone, highlighting the importance of vertical structure
  • Urban regions offer more limited per-capita living spaces compared to rural areas, but the higher density reduces inequality in living space distribution within urban settings
building height mappingurban-rural gap3D urban structurebuilt environmentUnited Statesremote sensing
  • Air conditioning systems can reduce indoor temperatures by up to 18 °C but cause substantial nighttime outdoor warming of 2–2.5 °C in high-density urban neighborhoods, creating a thermal equity problem
  • Increasing the target indoor AC temperature by just 2 °C can reduce outdoor temperatures by up to 1 °C in high-density areas, while reducing the spatial cooled fraction can cut energy consumption by up to 50%
  • Uses coupled WRF and building energy modeling in Chongqing to demonstrate that moderate AC usage adjustments and temporal scheduling can balance indoor comfort, outdoor heat mitigation, and energy resilience
air conditioningthermal disparitywaste heaturban heat islandindoor-outdoor temperatureenergy efficiency

2023

  • Convection-permitting WRF simulations coupled with an Urban Canopy Model show that both mega-city development and near-future warming independently enhance the frequency and intensity of hourly extreme rainfall over the Greater Bay Area
  • Urban expansion has a comparable effect on local extreme rainfall as near-future (2001–2050) climate warming, nearly doubling the frequency of very heavy rainfall events exceeding 90 mm/hr over urban areas
  • Observational records from 1971 to 2016 confirm that both intensity and frequency of extreme rainfall increased by more than 5% per decade across most Greater Bay Area mega-cities, concurrent with rapid urban expansion
extreme rainfallmega-city developmentGreater Bay AreaWRF modelingurban precipitationclimate warming

2022

  • News & Views commentary in Nature Climate Change highlighting that more than two-thirds of urban tree species worldwide are projected to face climate risk by 2050
  • Discusses a global assessment of 3,129 species across 164 cities showing cities at low latitudes face the greatest vulnerability, with nearly all urban tree species exceeding climatic safety margins
  • Argues that urban forests are central to climate adaptation and nature-based solutions, yet species selection for urban greening rarely accounts for future climatic conditions
  • Calls for proactive strategies including planting climate-resilient species and adaptive management to sustain the ecological, economic, and public health benefits of urban trees
urban forestsurban trees climate changeurban canopynature-based solutionsclimate adaptationurban cooling

2021

  • Urban expansion through 2050 slightly reduces daytime heat stress (~0.2 °C wet-bulb globe temperature) but intensifies nighttime heat stress by ~1 °C on average and up to 2–3 °C in mega-urban regions of China, India, and Nigeria
  • Approximately half of the nighttime outdoor urban heat stress increase from urban expansion persists even after large-scale deployment of cool roofs as a heat island mitigation strategy
  • Uses WRF simulations with spatially explicit urban land expansion projections and wet-bulb globe temperature (WBGT) as a combined temperature-humidity metric to assess physiological heat stress
  • Demonstrates that infrastructure-based mitigation measures such as cool and green roofs are insufficient alone, highlighting the need for complementary adaptation strategies to address persistent nighttime heat stress
nighttime heat stressurban heat island mitigationurban expansionWRF simulationnocturnal heat exposureclimate and health
  • Proposes a local climate zone (LCZ) land-use change simulation framework that groups ten urban land types into three categories (high-rise, mid-rise, low-rise) for more reliable urbanization simulation
  • Shows that the spatial pattern of occurrence probabilities of grouped urban land types aligns with the economic land price bid curve, linking urban morphology transitions to land market dynamics
  • Validated in the Pearl River Delta metropolitan area, demonstrating applicability to rapidly urbanizing coastal regions where frequent transitions among urban land types alter climate-relevant surface properties
  • Addresses a critical gap in urban climate forecasting by coupling LCZ-based land-use change modeling with meteorological simulation for scenario-based warming projections
urban climate forecastland-use change modelingPearl River DeltaGreater Bay Arearapid urbanizationSLEUTH model
  • Monitors the detailed spatiotemporal evolution of land development in the Qiantang New District reclamation area in China, revealing dramatic land-use changes under rapid urbanization pressures
  • Identifies key historical drivers of land-use change in reclaimed coastal areas where understanding of long-term dynamics has been limited
  • Provides a framework and insights for sustainable land management planning in rapidly urbanizing coastal reclamation regions
land reclamationurbanizationremote sensing monitoringland developmentland-use change

2020

  • Couples urban growth projections with WRF regional climate modeling to quantify future urbanization impacts on local climate in the Pearl River Delta mega-urban region
  • Shows that urban expansion substantially modifies local temperature and precipitation patterns beyond background climate change under multiple RCP scenarios
  • Finds that future land-use change contributes an additional warming signal comparable in magnitude to the urban heat island effect under current conditions
Pearl River Deltaurbanization impactclimate scenariosregional climate modelingland-use changemega-urban region
  • Projects global urban land expansion through 2100 under all five SSPs, with SSP5 yielding the greatest total urban land area and SSP1/SSP4 showing turning points in the 2080s/2070s as socioeconomic growth slows
  • Approximately 50–63% of newly expanded urban land worldwide is projected to replace existing croplands, making agricultural land the dominant land-cover type lost to urbanization
  • The resulting cropland loss is estimated to reduce global crop production by 1–4%, equivalent to the annual food requirements of 122 to 1,389 million people depending on the SSP scenario
  • China and the United States face the largest urban expansion under SSP5, whereas Latin American regions experience greater expansion under SSP3, reflecting regionally distinct urbanization pressures
global urban expansionSSP scenariosurban land projectionsNature CommunicationsIPCCurban growth 2100
  • Produces annual maps of global urban land at 30-meter resolution from 1985 to 2015, revealing an 80% increase in the world's urban footprint from approximately 362,000 km² to 653,000 km²
  • China and the United States together accounted for nearly half of the global net increase in urban land, with China's urban area expanding more than fourfold during the study period
  • Global urban expansion disproportionately consumed cropland, with approximately 300,000 km² of agricultural land converted to urban uses, raising food security concerns
  • The rate of global urbanization accelerated markedly after 2000, with developing countries in Asia and Africa driving the fastest urban growth
global urban changeLandsatremote sensingurban land coverGAIA datasetimpervious surface mapping30-meter resolution
  • Projects that urban expansion will convert approximately 290,000 km² of natural habitat to urban land uses between 2000 and 2030, with the most severe losses in tropical coastal regions of China, Brazil, and Nigeria
  • A systematic review of 922 studies found that 72% of research on direct urban impacts on biodiversity was conducted in high-income countries, despite the greatest projected habitat loss occurring in lower-income countries
  • Indirect urban impacts on biodiversity, such as food consumption and resource demand, affect a far greater land area than direct habitat conversion, yet only 34% of reviewed studies quantified these indirect effects
  • By 2030, urban areas will accommodate over 1.2 billion additional people globally, yet critical research gaps persist in tropical and Southern Hemisphere regions where urbanization-driven biodiversity loss is expected to be most severe
urban biodiversityhabitat lossurban ecologyconservation planningbiodiversity hotspotsurban growth impact
  • Produces 100-meter resolution gridded population projections for China under the SSPs using machine learning, revealing that 30–43% of population in existing coarse-resolution SSP grids is misallocated to non-urban land
  • Projects that exposure to extreme heat in China will increase by 121–136% under RCP4.5 and 164–191% under RCP8.5 from 2015 to 2050, concentrated in high-density urbanized areas like the Yangtze and Pearl River Deltas
  • All scenarios project accelerated population decline in China from 2020 to 2050, though provincial-level trends diverge significantly from the national trajectory
population projectionsChinaSSP scenariosgridded populationurbanizationclimate impact assessment

2019

  • Projects global urban land areas to expand by 0.6–1.3 million km² (78%–171%) between 2015 and 2050 under shared socioeconomic pathways (SSPs)
  • Estimates urban heat island intensification of 0.5–0.7 °C on average, with localized warming reaching up to 3 °C due to urban land-cover change
  • Finds that urban expansion-driven warming can be about half, and in some locations up to two times, as strong as greenhouse gas-induced warming under RCP 4.5
  • Identifies tropical regions in the Southern Hemisphere as facing disproportionately high extreme heat risk from compounding urbanization and climate warming
  • Combines spatially explicit urban expansion modeling at 5-km resolution with mathematical urban heat island models across roughly half of the world's future urban population
urban expansionurban heat islandSSP scenariosglobal urbanization projectionsland-use change modelingWRF climate modelurban land cover 2050
  • Fugitive road dust PM2.5 emissions in urban Lanzhou were approximately 1,141 kg per day, accounting for 24.6% of total PM2.5 emissions, concentrated in areas of intense traffic and construction
  • Constructs a high-resolution (500 × 500 m) road dust PM2.5 emissions inventory, finding emissions exceeding 30,000 µg/m²/day in areas with smaller particle sizes and higher traffic intensity
  • Estimates 234.5 premature deaths attributable to road dust PM2.5 exposure in Lanzhou in 2017, underscoring that reducing road dust emissions is critical for public health in developing urban regions
PM2.5road dust emissionsair pollutionurban air qualityhealth impactsparticulate matter
  • Natural wetlands in Hangzhou Bay decreased at ~10 km² per year from 1984 to 2016, with the degradation rate quadrupling from ~4 to ~18 km²/year after establishment of an economic development zone in 2001
  • Under a continued high-development scenario, the coastline is projected to move approximately 5.89 km inland, undermining coastal protections against sea level rise
  • Wetland protection scenarios could mitigate projected degradation by approximately 20%, demonstrating that land-use simulation models can identify effective conservation strategies under rapid urbanization
wetland degradationremote sensingland-use modelinghabitat lossconservationland cover change

2018

  • Natural dust sources contribute 81.0% of global dust emissions while anthropogenic sources contribute 19.0%, with natural emissions concentrated in hyper-arid regions and anthropogenic emissions in semi-arid and sub-humid regions
  • Semi-arid regions exhibit the most complex interplay, accounting for 42.99% of global anthropogenic dust emissions due to diverse land cover types including grasslands, croplands, and urban areas
  • The complex mix of natural and anthropogenic dust in semi-arid regions may be a crucial factor enhancing the observed warming trend over these areas under global climate change
dust emissionsanthropogenic dustnatural dustclimate regionsair qualityatmospheric environment
  • Integrates local self-similarity (LSS) descriptors, which capture common internal layout features across sensor types, with mutual information to handle images with both rigid and non-rigid radiometric distortions
  • Uses a Bayesian probabilistic model for matching LSS feature ensembles, followed by particle swarm optimization to maximize combined self-similarity and mutual information for optimal correspondences
  • Overcomes limitations of conventional registration algorithms that fail when common underlying visual features between different sensor images are not distinct
image registrationself-similaritymulti-sensorremote sensingmutual informationfeature matching

2017

  • Constructs both indirect and direct anthropogenic dust emission schemes validated against CALIPSO satellite observations, identifying high emission centers in India, eastern China, North America, and Africa
  • Pasturelands contribute approximately 60% of indirect anthropogenic dust emissions, substantially more than croplands, while urban areas emit more (13.5 µg/m²/s) than rural areas (7.9 µg/m²/s)
  • Unlike natural dust emissions with strong seasonal variation, indirect anthropogenic dust emissions show minimal seasonal fluctuation, highlighting that human-driven dust is a persistent year-round concern
anthropogenic dust emissionsglobal dustatmospheric modelingair qualityhuman-caused dust

2015

  • Develops a snow depth retrieval algorithm based on support vector regression that combines passive microwave brightness temperature with visible/infrared reflectance to improve accuracy over microwave-only approaches
  • Visible and near-infrared measurements contain supplementary information about surface snowpack properties that, when integrated with 19 and 37 GHz data from SSM/I and SSMI/S sensors, significantly reduce retrieval uncertainties
  • Validated in Xinjiang, China, demonstrating that combining data modalities captures snow cover characteristics more comprehensively than single-sensor retrieval methods
snow depth retrievalmicrowave remote sensingdata fusionvisible/infrared reflectancecryosphereremote sensing

2013

  • Proposes AIS-MOLA, an improved artificial immune system for multi-objective land-use allocation that seeks the Pareto front to reveal trade-offs among spatial suitability and compactness objectives
  • Introduces heuristic hypermutation based on compromise programming, non-dominated neighbour-based proportional cloning, and a crossover operator that preserves connected land-use patches
  • Outperforms Pareto Simulated Annealing (PSA) by generating solutions closer to the true Pareto front while requiring only 5.1% of PSA's computation time
  • Validated on a large-area case study in Panyu, Guangdong, China (389 × 337 cells) demonstrating optimal trade-offs between spatial suitability and compactness
land-use optimizationPareto frontartificial immune systemmulti-objective optimizationspatial planningGIScience
  • Introduces a novel SMI metric integrating spatial information from detected features with mutual information between reference and sensed images, achieving more robust registration than either metric alone
  • Employs a two-phase searching strategy using ant colony optimization for continuous domains, first identifying approximate solutions via spatial information then refining with the combined SMI metric
  • Demonstrates highly accurate automatic registration on multi-sensor remote sensing imagery (TM, SPOT, SAR) where conventional single-metric methods fail
image registrationmulti-sensormutual informationremote sensingSARIEEE TGRS

2012

  • Proposes the MACO method, assigning different ant types to different land uses, outperforming simulated annealing by ~4.5% and genetic algorithms by ~1.3% in overall land-use allocation utility
  • Generates more spatially compact land-use patterns than competing methods, improving compactness over SA and GA by 8.9% and 1.7% respectively with significantly less computation time
  • Addresses the exponentially growing solution space of large-area land-use allocation by incorporating spatial exchange mechanisms and optimizing across suitability, compactness, and conversion cost objectives
ant colony optimizationland-use allocationspatial optimizationmulti-objective planningGIScience

2010

  • Integrates three major computational components — cellular automata (CA), multi-agent systems (MAS), and swarm intelligence (SI) — into the unified GeoSOS platform for process-based geographical modeling
  • Includes multiple CA algorithms (MCE-CA, Logistic-CA, PCA-CA, ANN-CA, Decision-tree-CA) that automatically derive transition rules through data mining for simulating urban growth and land-use change
  • Couples simulation models with optimization models, demonstrating more satisfactory modeling results under complex changing environments than either approach alone
GeoSOSgeographical simulationcellular automataland-use simulationspatial optimizationurban growth modeling