Assessment Of 2D And 3D Methods For Property Valuation Using Remote Sensing Data At The Neighbourhood Scale In Xi'an,China

The Chinese property market has kept flourishing in the past decades due to fast urbanisation.Xi'an,the capital city of Shaanxi Province,also has become one of the cities with a fast-growing property market.The high-rise apartments dominate the property market to meet the demands of the growing urban population.The continuous construction of the neighbourhoods has led to significant spatial changes in the vertical dimension.However,the change of the geographical information in the vertical dimension cannot be represented by the current 2D-based property valuation methods.Therefore,it is important to introduce 3D indicators and 3D modelling into the property valuation.This research aims to narrow the cognitive and technical gap by assessing the performances of 2D and 3D methods for property valuation.In this research,a mixed qualitative-quantitative methodology was issued to assess 2D and 3D methods for property valuation using remote sensing data at the neighbourhood scale.The author employed semi-structured expert interview,focus group and questionnaire in understanding the pricing policy,identify the current situation of 3D modelling in China,and buyers'preferences for high-rise apartments.The 2D methods applied both ordinary least squares(OLS)and geographically weighted regression(GWR)to run the model of 2D indicators.The 3D model was built in CityEngine to build the high-rise apartment models for visualization and quantification.Finally,the performances of the 2D and 3D models were assessed.This research contributes to the existing literature related to property valuation and 3D modelling by showing the importance of 3D indicators in influencing the property prices of the high-rise apartments and the possibility of applying 3D modelling in property valuation.Future research opportunities were identified,such as machine learning on predicting the property price and developing suitable 3D software for buyers.The main findings were as follows:(1)The pricing strategy had no specific formula,and the current 3D modelling status in Xi'an needs to be developedAccording to the qualitative analysis of semi-structured expert interview and focus group,the results revealed that the government adopts the market comparison method and the real estate developers adopt the cost method for determining the property price at present.3D modelling needs to be further developed in Xi'an.Now the Level of Details(LoD)is still low and limited in application.(2)Different age groups had different preferences for high-rise apartments.The different aspects of buyers'preferences were analysed and taken as selection criteria of 2D and 3D indicators for property valuation.Buyers'preferences were collected by questionnaire.It was revealed that nearly half of the respondents chose to live in the middle storey level,which had a link with the Doctrine of the Mean,a Chinese tradition.The majority chose the“south-north”orientation.Concerning surrounding environment and property physical attributes,“public security”,“property orientation”,“less noise”were the most favourite three indicators,while the bottom three were“sports facility”,“entertainment facility”,and“cultural facility”.People preferred to see green land and water outside of the window.They disliked the view of building and street.(3)The hedonic price model containing the 2D indicators could not generalize the model and explain the property price variation in Xi'an.The regression results showed that GWR performed better than OLS,so GWR was chosen for detailed analysis.It also revealed that density of factory,normalised difference vegetation index(NDVI),distance to Central Business District(CBD),distance to food and distance to the subway were the five significant indicators influencing property price.However,unlike other studies,it could not generalise the model for they both had low R~2.The fixed-price and purchase-restriction policy established by the Xi'an municipal government may influence the results.(4)The 3D model could better explain the property price variation than the 2D model.The 3D method included four indicators,view quality,sky view factor(SVF),sunlight and property orientation and executed model of 3D indicators.SVF,sunlight and property orientation were the three significant indicators.Leave-one-out cross-validation(LOOCV)was carried out based on local knowledge.The error percentage between the predicted price and the real price was 9.76%.In conclusion,it was proved by the comparison results that the3D method could better explain the property price variation at the neighbourhood scale than2D methods.The R~2 in 3D method was 0.451 while R~2 in GWR of 2D methods was 0.217.3D indicators were successfully analysed and quantified in CityEngine.They were also effectively visualised via graphs and videos to show the geographical information change in the vertical dimension.