IOD is the only criterion used. The study mainly focuses on the geometric importance of the building, and the building types are ignored. The existing buildings in green spaces should also be taken into account in the calculation, but these constraints and practical conditions are not taken into account.
Remote sensing image (Landsat images) to land-use maps; distance to roads map; distance to business center map; distance to population center map; environmental-sensitive area map; slope map; and elevation map.
Conversion label; distance to urban center; distance to town center; distance to commercial housing area; distance to main roads; distance to agricultural land; and stochastic.
SLEUTH parameterization with supervised classification (land use and urban) and the use of statistical measures: sng (cumulative number of urbanized pixels by spontaneous neighborhood growth), og (cumulative number of urbanized pixels by organic growth), rt (cumulative number of urbanized pixels by road influenced growth), area (total number of urban pixels), edges (number of urban to nonurban pixel edges), clusters (number of urban pixel clusters), rad (radius of cluster, which encloses the urban area), slope coefficient, spread coefficient, breed coefficient, road gravity coefficient, percent urban (percent of urbanized pixels divided by the number of pixels available for urbanization), urban growth rate, and number of growth pixels each year.
Two different urban growth types: organic growth (based on segment) and spontaneous growth (based on pixels), which where identified and separately simulated introducing a landscape expanse index (LEI) that built on neighborhood density analysis. CA components: suitability surface, neighborhood, stochastic perturbation, and development probability.
Socioeconomic and natural climatic factors: climate change, socioeconomic changes, historical land use, and interactions between variables to obtain land-use demand in each decade. Later, to ANN for land use in 2010: neighborhood influence, weight factor, self-adaptively land inertia, and converting cost. These combined probabilities with probability-of-occurrence surfaces and with roulette wheel selection detect the land use in time.
Constrained relations among factors were applied in DE to generate different sets of CA parameters for the prediction of future scenarios. Variables that affect land-use changes: distance to urban center; distance to district center; distance to main roads; distance to the roads along Dianchi Lake; distance to protected areas; and DEM.
Prior to classifying the images using a supervised classification algorithm, unsupervised classification and normalized difference vegetation index (NDVI) were calculated to help select suitable polygons as training sites and to improve the overall classification process. The classification scheme was established based on auxiliary information from the field survey, local knowledge of the study area, and visual interpretation of the images. Image classification was performed using the maximum likelihood classification (MLC) algorithm, which is a supervised classification, and one of the most widely applied parametric classification algorithms.
A series of grayscale probabilistic maps with different parameters were produced to be used as the basis for the evolutionary model. The parameters were captured in the following gradient maps: land ownership, regulatory master plans, vertical urban consolidation, accessibility, and allocation. Land ownership: public vs. private; land redevelopment master plans; vertical density; accessibility; allocation parameters; and economic and real estate parameters.
They used the entropy method to determine weights for the selected factors. Potential factors for multicriteria evaluation: population change; change in employment; population density; median housing income; and highway accessibility, transit accessibility, slope, distance to each of the existing land use types, administrative constraints, and natural constraints.
Criteria for comparing CA metaheuristic models: best objective function value; iteration or generation; computational time; initial urban area; hit; correct rejection; failure; false alarm; assignment; and quantity. Spatial input variables: distance to city center; distance to country center; distance to main road; distance to railroad; distance to coast; DEM; and restricted areas.
This study uses remote sensing and geographic information. From 1992 and 2003 Landsat 5 TM images, and 2014 Landsat 8 OLI images and DEM, a land-use classification map was obtained for each year. The cell automaton model is mainly composed of cell, cell space, neighbor, ruler, and time. The closer the distance between the nuclear cell and the neighbor, the higher the weight factor. The weight factor is combined with transition probabilities to predict the state of adjacent grid cells so that land-use change is not a completely random decision. The Markov chain model component controls the temporal dynamics between LULC classes based on the transition probabilities, while the spatial dynamics are controlled by local rules determined by the CA spatial filter or transition potential maps.
They use spatial variables in UMCNN. For RFA-CA, the factors used are neighborhood effects, constraint factors, development suitability, and stochastic factors.
Inputs for the training phase of the CA-Markov model: time-1 land-use map; time-2 land-use map; simulated n-time transition area matrix; and simulated n-time Markov conditional probability image. Inputs for the simulation phase of the CA-Markov model: simulated n-time transition area matrix and simulated n-time Markov conditional probability image.
Preprocessing tasks, such as radiation calibration, FLAASH atmospheric corrections, image mosaicking, and image cropping, were applied before classifying the images with the ENVI tool.
Agents of urbanization: existing built-up; hot spots; commutation, high-preference roads; medium-preference roads; least-preference roads; railways; red category industries; orange category industries; green category industries; white category industries; high land prices; low land prices; medium land prices; places of public interests; public utility centers; and population.
Image preprocessing techniques include the following: layer stacking; mosaicking; and subsetting or clipping to study area boundaries. The LULC classification scheme comprised seven LULC classes, identified by codes, to prepare different LULCs to simulate future land use.
The research described the continuing historical increase in built-up space through the consumption of other ecologically valuable LULC classes. Driving factors: elevation, slope, road distance, highway distance, rail distance, and urban centers.
The analysis was based on three equal interval LULC maps derived from satellite images: Landsat TM for 1998, 2008, and 2018, together with topographic spatial layers (elevation aspects and terrain slopes) derived from the ASTER digital elevation model. Other spatial parameters (population density, proximity to urban centers, and proximity to major roads) were also incorporated into the simulation process.
