Calibration of an urban cellular automaton model by using statistical techniques and a genetic algorithm. Application to a small urban settlement of NW Spain

Cellular automata (CA) stand out among the most commonly used urban models for the simulation and analysis of urban growth because of their ability to reproduce complex dynamics, similar to those found in real cities, from simple rules. However, CA models still have to overcome some shortcomings related to their flexibility and difficult calibration. This study combines various techniques to calibrate an urban CA that is based on one of the most widely used urban CA models. First, the number of calibration parameters is reduced by using various statistical techniques, and, second, the calibration procedure is automated through a genetic algorithm. The resulting model has been assessed by simulating the urban growth of Ribadeo, a small village of NW Spain, characterized by low, slow urban growth, which makes the identification of urban dynamics and consequently the calibration of the model more difficult. Simulation results have shown that, by automating the calibration procedure, the model can be more easily applied and adapted to urban areas with different characteristics and dynamics. In addition, the simulations obtained with the proposed model show better values of cell-to-cell correspondence between simulated and real maps, and the values for most spatial metrics are closer to real ones.     

Identifying regions based on flexible user-defined constraints

The identification of regions is both a computational and conceptual challenge. Even with growing computational power, regionalization algorithms must rely on heuristic approaches in order to find solutions. Therefore, the constraints and evaluation criteria that define a region must be translated into an algorithm that can efficiently and effectively navigate the solution space to find the best solution. One limitation of many existing regionalization algorithms is a requirement that the number of regions be selected a priori. The recently introduced max-p algorithm does not have this requirement, and thus the number of regions is an output of, not an input to, the algorithm. In this paper, we extend the max-p algorithm to allow for greater flexibility in the constraints available to define a feasible region, placing the focus squarely on the multidimensional characteristics of the region. We also modify technical aspects of the algorithm to provide greater flexibility in its ability to search the solution space. Using synthetic spatial and attribute data, we are able to show the algorithm’s broad ability to identify regions in maps of varying complexity. We also conduct a large-scale computational experiment to identify parameter settings that result in the greatest solution accuracy under various scenarios. The rules of thumb identified from the experiment produce maps that correctly assign areas to their ‘true’ region with 94% average accuracy, with nearly 50% of the simulations reaching 100% accuracy.     

A hybrid genetic algorithm with local optimiser improves calibration of a vegetation change cellular automata model

Cellular automata (CA) models are commonly used to model vegetation dynamics, with the genetic algorithm (GA) being one method of calibration. This article investigates different GA settings, as well as the combination of a GA with a local optimiser to improve the calibration effort. The case study is a pattern-calibrated CA to model vegetation regrowth in central Victoria, Australia. We tested 16 GA models, varying population size, mutation rate, and level of allowable mutation. We also investigated the effect of applying a local optimiser, the Nelder?Mead Downhill Simplex (NMDS) at GA convergence. We found that using a decreasing mutation rate can reduce computational cost while avoiding premature GA convergence, while increasing population size does not make the GA more efficient. The hybrid GA-NMDS can also reduce computational cost compared to a GA alone, while also improving the calibration metric. We conclude that careful consideration of GA settings, including population size and mutation rate, and in particular the addition of a local optimiser, can positively impact the efficiency and success of the GA algorithm, which can in turn lead to improved simulations using a well-calibrated CA model.     

Calibrating a Land Parcel Cellular Automaton (LP-CA) for urban growth simulation based on ensemble learning

The reliability of raster cellular automaton (CA) models for fine-scale land change simulations has been increasingly questioned, because regular pixels/grids cannot precisely represent irregular geographical entities and their interactions. Vector CA models can address these deficiencies due to the ability of the vector data structure to represent realistic urban entities. This study presents a new land parcel cellular automaton (LP-CA) model for simulating urban land changes. The innovation of this model is the use of ensemble learning method for automatic calibration. The proposed model is applied in Shenzhen, China. The experimental results indicate that bagging-Naïve Bayes yields the highest calibration accuracy among a set of selected classifiers. The assessment of neighborhood sensitivity suggests that the LP-CA model achieves the highest simulation accuracy with neighbor radius r = 2. The calibrated LP-CA is used to project future urban land use changes in Shenzhen, and the results are found to be consistent with those specified in the official city plan.     

Computational improvements to multi-scale geographically weighted regression

ABSTRACT

Geographically Weighted Regression (GWR) has been broadly used in various fields to model spatially non-stationary relationships. Multi-scale Geographically Weighted Regression (MGWR) is a recent advancement to the classic GWR model. MGWR is superior in capturing multi-scale processes over the traditional single-scale GWR model by using different bandwidths for each covariate. However, the multiscale property of MGWR brings additional computation costs. The calibration process of MGWR involves iterative back-fitting under the additive model (AM) framework. Currently, MGWR can only be applied on small datasets within a tolerable time and is prohibitively time-consuming to run with moderately large datasets (greater than 5,000 observations). In this paper, we propose a parallel implementation that has crucial computational improvements to the MGWR calibration. This improved computational method reduces both memory footprint and runtime to allow MGWR modelling to be applied to moderate-to-large datasets (up to 100,000 observations). These improvements are integrated into the mgwr python package and the MGWR 2.0 software, both of which are freely available to download.     

A partitioned and asynchronous cellular automata model for urban growth simulation

Cellular automata (CA) models are used to analyze and simulate the global phenomenon of urban growth. However, these models are characterized by ignoring spatially heterogeneous transition rules and asynchronous evolving rates, which make it difficult to improve urban growth simulations. In this paper, a partitioned and asynchronous cellular automata (PACA) model was developed by implementing the spatial heterogeneity of both transition rules and evolving rates in urban growth simulations. After dividing the study area into several subregions by k-means and knn-cluster algorithms, a C5.0 decision tree algorithm was employed to identify the transition rules in each subregion. The evolving rates for cells in each regularly divided grid were calculated by the rate of changed cells. The proposed PACA model was implemented to simulate urban growth in Wuhan, a large city in central China. The results showed that PACA performed better than traditional CA models in both a cell-to-cell accuracy assessment and a shape dimension accuracy assessment. Figure of merit of PACA is 0.368 in this research, which is significantly higher than that of partitioned CA (0.327) and traditional CA (0.247). As for the shape dimension accuracy, PACA has a fractal dimension of 1.542, which is the closest to that of the actual land use (1.535). However, fractal dimension of traditional CA (1.548) is closer to that of the actual land use than that of partitioned CA (1.285). It indicates that partitioned transition rules play an important role in the cell-to-cell accuracy of CA models, whereas the combination of partitioned transition rules and asynchronous evolving rates results in improved cell-to-cell accuracy and shape dimension accuracy. Thus, implementing partitioned transition rules and asynchronous evolving rates yields better CA model performance in urban growth simulations due to its accordance with actual urban growth processes.     

A high-performance cellular automata model for urban simulation based on vectorization and parallel computing technology

Cellular automata (CA) models can simulate complex urban systems through simple rules and have become important tools for studying the spatio-temporal evolution of urban land use. However, the multiple and large-volume data layers, massive geospatial processing and complicated algorithms for automatic calibration in the urban CA models require a high level of computational capability. Unfortunately, the limited performance of sequential computation on a single computing unit (i.e. a central processing unit (CPU) or a graphics processing unit (GPU)) and the high cost of parallel design and programming make it difficult to establish a high-performance urban CA model. As a result of its powerful computational ability and scalability, the vectorization paradigm is becoming increasingly important and has received wide attention with regard to this kind of computational problem. This paper presents a high-performance CA model using vectorization and parallel computing technology for the computation-intensive and data-intensive geospatial processing in urban simulation. To transfer the original algorithm to a vectorized algorithm, we define the neighborhood set of the cell space and improve the operation paradigm of neighborhood computation, transition probability calculation, and cell state transition. The experiments undertaken in this study demonstrate that the vectorized algorithm can greatly reduce the computation time, especially in the environment of a vector programming language, and it is possible to parallelize the algorithm as the data volume increases. The execution time for the simulation of 5-m resolution and 3 × 3 neighborhood decreased from 38,220.43 s to 803.36 s with the vectorized algorithm and was further shortened to 476.54 s by dividing the domain into four computing units. The experiments also indicated that the computational efficiency of the vectorized algorithm is closely related to the neighborhood size and configuration, as well as the shape of the research domain. We can conclude that the combination of vectorization and parallel computing technology can provide scalable solutions to significantly improve the applicability of urban CA.     

Automatic extraction of road intersection points from USGS historical map series using deep convolutional neural networks

ABSTRACT

Road intersection data have been used across a range of geospatial analyses. However, many datasets dating from before the advent of GIS are only available as historical printed maps. To be analyzed by GIS software, they need to be scanned and transformed into a usable (vector-based) format. Because the number of scanned historical maps is voluminous, automated methods of digitization and transformation are needed. Frequently, these processes are based on computer vision algorithms. However, the key challenges to this are (1) the low conversion accuracy for low quality and visually complex maps, and (2) the selection of optimal parameters. In this paper, we used a region-based deep convolutional neural network-based framework (RCNN) for object detection, in order to automatically identify road intersections in historical maps of several cities in the United States of America. We found that the RCNN approach is more accurate than traditional computer vision algorithms for double-line cartographic representation of the roads, though its accuracy does not surpass all traditional methods used for single-line symbols. The results suggest that the number of errors in the outputs is sensitive to complexity and blurriness of the maps, and to the number of distinct red-green-blue (RGB) combinations within them.     

Knowledge transfer and adaptation for land-use simulation with a logistic cellular automaton

Few studies have been conducted into the use of knowledge transfer for tackling geo-simulation problems. Cellular automata (CA) have proven to be an effective and convenient means of simulating urban dynamics and land-use changes. Gathering the knowledge required to build the CA may be difficult when these models are applied to large areas or long periods. In this paper, we will explore the possibility that the knowledge from previously collected data can be transferred spatially (a different region) and/or temporally (a different period) for implementing urban CA. The domain adaptation of CA is demonstrated by integrating logistic-CA with a knowledge-transfer technique, the TrAdaBoost algorithm. A modification has been made to the TrAdaBoost algorithm by incorporating a dynamicweight-trimming technique. This proposed model, CA_trans, is tested by choosing different periods and study areas in the Pearl River Delta. The ‘Figure of Merit’ measurements in the experiments indicate that CA_trans can yield better simulation results. The variance of traditional logistic-CA is about 2–5 times the variance of CA_trans until the number of new data reaches 30. The experiments have demonstrated that the proposed method can alleviate the sparse data problem using knowledge transfer.     

Coupling fuzzy clustering and cellular automata based on local maxima of development potential to model urban emergence and expansion in economic development zones

ABSTRACT

Modeling urban growth in Economic development zones (EDZs) can help planners determine appropriate land policies for these regions. However, sometimes EDZs are established in remote areas outside of central cities that have no historical urban areas. Existing models are unable to simulate the emergence of urban areas without historical urban land in EDZs. In this study, a cellular automaton (CA) model based on fuzzy clustering is developed to address this issue. This model is implemented by coupling an unsupervised classification method and a modified CA model with an urban emergence mechanism based on local maxima. Through an analysis of the planning policies and existing infrastructure, the proposed model can detect the potential start zones and simulate the trajectory of urban growth independent of the historical urban land use. The method is validated in the urban emergence simulation of the Taiping Bay development zone in Dalian, China from 2013 to 2019. The proposed model is applied to future simulation in 2019–2030. The results demonstrate that the proposed model can be used to predict urban emergence and generate the possible future urban form, which will assist planners in determining the urban layout and controlling urban growth in EDZs.     

A meta-modeling approach for spatio-temporal uncertainty and sensitivity analysis: an application for a cellular automata-based Urban growth and land-use change model

The paper presents a computationally efficient meta-modeling approach to spatially explicit uncertainty and sensitivity analysis in a cellular automata (CA) urban growth and land-use simulation model. The uncertainty and sensitivity of the model parameters are approximated using a meta-modeling method called polynomial chaos expansion (PCE). The parameter uncertainty and sensitivity measures obtained with PCE are compared with traditional Monte Carlo simulation results. The meta-modeling approach was found to reduce the number of model simulations necessary to arrive at stable sensitivity estimates. The quality of the results is comparable to the full-order modeling approach, which is computationally costly. The study shows that the meta-modeling approach can significantly reduce the computational effort of carrying out spatially explicit uncertainty and sensitivity analysis in the application of spatio-temporal models.     

Estimation of Blast-Induced Air Overpressure in Quarry Mines Using Cubist-Based Genetic Algorithm

In the present work, blast-induced air overpressure is estimated by an innovative intelligence system based on the cubist algorithm (CA) and genetic algorithm (GA) with high accuracy, called GA–CA model. Herein, CA initialization model was developed first and the hyper-parameters of the CA model were selected randomly. Subsequently, the GA procedure was applied to perform a global search for the optimized values of the hyper-factors of the CA model. Root-mean-square error (RMSE) is utilized as a compatibility function to determine the optimal CA model with the lowest RMSE. Gaussian process (GP), conditional inference tree (CIT), principal component analysis (PCA), hybrid neural fuzzy inference system (HYFIS) and k-nearest neighbor (k-NN) models are also developed as the benchmark models in order to compare and analyze the quality of the proposed GA–CA algorithm; 164 blasting works were investigated at a quarry mine of Vietnam for this aim. The results revealed that GA significantly improved the performance of the CA model. Based on the statistical indices used for model assessment, the proposed GA–CA model was confirmed as the most superior model as compared to the other models (i.e., GP, CIT, HYFIS, PCA, k-NN). It can be applied as a robust soft computing tool for estimating blast-induced air overpressure.

     

Examining the physical meaning of the bank erosion coefficient used in meander migration modeling

Widely used models of meander evolution relate migration rate to vertically averaged near-bank velocity through the use of a coefficient of bank erosion (E). In applications to floodplain management problems, E is typically determined through calibration to historical planform changes, and thus its physical meaning remains unclear. This study attempts to clarify the extent to which E depends on measurable physical characteristics of the channel boundary materials using data from the Sacramento River, California, USA. Bend-average values of E were calculated from measured long-term migration rates and computed near-bank velocities. In the field, unvegetated bank material resistance to fluvial shear (k) was measured for four cohesive and noncohesive bank types using a jet-test device. At a small set of bends for which both E and k were obtained, we discovered that variability in k explains much of the variability in E. The form of this relationship suggests that when modeling long-term meander migration of large rivers, E depends largely on bank material properties. This finding opens up the possibility that E may be estimated directly from field data, enabling prediction of meander migration rates for systems where historical data are unavailable or controlling conditions have changed. Another implication is that vegetation plays a limited role in affecting long-term meander migration rates of large rivers like the Sacramento River. These hypotheses require further testing with data sets from other large rivers.     

Interest in intermediate soft-classified maps in land change model validation: suitability versus transition potential

This study compares two types of intermediate soft-classified maps. The first type uses land use/cover suitability maps based on a multi-criteria evaluation (MCE). The second type focuses on the transition potential between land use/cover categories based on a multi-layer perceptron (MLP). The concepts and methodological approaches are illustrated in a comparable manner using a Corine data set from the Murcia region (2300 km², Spain) in combination with maps of drivers that were created with two stochastic, discretely operating, commonly used tools (MCE in CA_MARKOV and MLP in Land Change Modeler). The importance of the different approaches and techniques for the obtained results is illustrated by comparing the specific characteristics of both approaches by validating the suitability versus transition potential maps to each other using a Spearman correlation matrix and, between the Corine maps, using classical ROC (receiver operating characteristic) statistics. Then, we propose a new use of ROC statistics to compare these intermediate soft-classified maps with their respective hard-classified maps of the models for each category. The validation of these results can be beneficial in choosing a suitable model and provide a better understanding of the implications of the different modeling steps and the advantages and limitations of the modeling tools.     

Preliminary Cellular-Automata Forecast of Permit Activity from 1998 to 2010, Idaho and Western Montana

For public land management in Idaho and western Montana, the U.S. Forest Service (USFS) has requested that the U.S. Geological Survey (USGS) predict where mineral-related activity will occur in the next decade. Cellular automata provide an approach to simulation of this human activity. Cellular automata (CA) are defined by an array of cells, which evolve by a simple transition rule, the automaton. Based on exploration trends, we assume that future exploration will focus in areas of past exploration. Spatial-temporal information about mineral-related activity, that is permits issued by USFS and Bureau of Land Management (BLM) in the last decade, and spatial information about undiscovered resources, provide a basis to calibrate a CA. The CA implemented is a modified annealed voting rule that simulates mineral-related activity with spatial and temporal resolution of 1 mi² and 1 year based on activity from 1989 to 1998. For this CA, the state of the economy and exploration technology is assumed constant for the next decade. The calibrated CA reproduces the 1989–1998-permit activity with an agreement of 94%, which increases to 98% within one year. Analysis of the confusion matrix and kappa correlation statistics indicates that the CA underestimates high activity and overestimates low activity. Spatially, the major differences between the actual and calculated activity are that the calculated activity occurs in a slightly larger number of small patches and is slightly more uneven than the actual activity. Using the calibrated CA in a Monte Carlo simulation projecting from 1998 to 2010, an estimate of the probability of mineral activity shows high levels of activity in Boise, Caribou, Elmore, Lincoln, and western Valley counties in Idaho and Beaverhead, Madison, and Stillwater counties in Montana, and generally low activity elsewhere.     

Uncertainty analysis of runoff and sedimentation in a forested watershed using sequential uncertainty fitting method

The Soil and Water Assessment Tool(SWAT) was implemented in a small forested watershed of the Soan River Basin in northern Pakistan through application of the sequential uncertainty fitting(SUFI-2) method to investigate the associated uncertainty in runoff and sediment load estimation. The model was calibrated for a 10-year period(1991–2000) with an initial 4-year warm-up period(1987–1990), and was validated for the subsequent 10-year period(2001–2010). The model evaluation indices R~2(the coefficient of determination), NS(the Nash-Sutcliffe efficiency), and PBIAS(percent bias) for stream flows simulation indicated that there was a good agreement between the measured and simulated flows. To assess the uncertainty in the model outputs, p-factor(a 95% prediction uncertainty, 95PPU) and r-factors(average wideness width of the 95 PPU band divided by the standard deviation of the observed values) were taken into account. The 95 PPU band bracketed 72% of the observed data during the calibration and 67% during the validation. The r-factor was 0.81 during the calibration and 0.68 during the validation. For monthly sediment yield, the model evaluation coefficients(R~2 and NS) for the calibration were computed as 0.81 and 0.79, respectively; for validation, they were 0.78 and 0.74, respectively. Meanwhile, the 95 PPU covered more than 60% of the observed sediment data during calibration and validation. Moreover, improved model prediction and parameter estimation were observed with the increased number of iterations. However, the model performance became worse after the fourth iterations due to an unreasonable parameter estimation. Overall results indicated the applicability of the SWAT model with moderate levels of uncertainty during the calibration and high levels during the validation. Thus, this calibrated SWAT model can be used for assessment of water balance components, climate change studies, and land use management practices.     

Simulating urban growth in a metropolitan area based on weighted urban flows by using web search engine

As a consequence of rapid and immoderate urbanization, simulating urban growth in metropolitan areas effectively becomes a crucial and yet difficult task. Cellular automata (CA) model is an attractive tool for understanding complex geographical phenomena. Although intercity urban flows, the key factors in metropolitan development, have already been taken into consideration in CA models, there is still room for improvement because the influences of urban flows may not necessarily follow the distance decay relationship and may change over time. A feasible solution is to define the weights of intercity urban flows. Therefore, this study presents a novel method based on weighted urban flows (CA_WeightedFlow) with the support of web search engine. The relatedness measured by the co-occurrences of the cities’ names (toponyms) on massive web pages can be deemed as the weights of intercity urban flows. After applying the weights, the gravitational field model is integrated with Logistic-CA to fulfill the modeling task. This method is employed to the urban growth simulation in the Pearl River Delta, one of the most urbanized metropolitan areas in China, from 2005 to 2008. The results indicate that our method outperforms traditional methods with respect to two measures of calibration goodness-of-fit. For example, CA_WeightedFlow can yield the best value of ‘figure of merit’. Moreover, the proposed method can be further used to explore various development possibilities by simply changing the weights.     

The environmental history of a mountain lake (Lago Paione Superiore, Central Alps, Italy) for the last c. 100 years: a multidisciplinary, palaeolimnological study

A palaeoecological study of an oligotrophic alpine lake, Paione Superiore (Italy), provided a record of historical changes in water quality. Historical trends in lake acidification were reconstructed by means of calibration and regression equations from diatoms, chrysophycean scales and pigment ratios. The historical pH was inferred by using two different diatom calibration data sets, one specific to the alpine region. These pH trends, together with the record of sedimentary carbonaceous particles and chironomid remains, indicate a recent acidification of this low alkalinity lake.Concentration of total organic matter, organic carbon, nitrogen, biogenic silica (BSiO₂), chlorophyll derivatives (CD), fucoxanthin, diatom cell concentration and number of chironomid head capsules increased during the last 2–3 decades. When expressed as accumulation rates, most of these parameters tended to decrease from the past century to c. 1950, then all except P increased to the present day. A marked increase in sedimentary nitrogen may be related to atmospheric pollution and to the general increases in output of N in Europe. High C/N ratios indicate a prevailing allochthonous source of organic matter.Finally, the increase in measured air temperature from the mid-1800's appeared to be related to lake water pH before industrialization: cold periods generally led to lower pH and vice-versa. The more recent phenomenon of anthropogenic acidification has apparently decoupled this climatic-water chemistry relationship.     

Automatic alignment of contemporary vector data and georeferenced historical maps using reinforcement learning

ABSTRACT

With large amounts of digital map archives becoming available, automatically extracting information from scanned historical maps is needed for many domains that require long-term historical geographic data. Convolutional Neural Networks (CNN) are powerful techniques that can be used for extracting locations of geographic features from scanned maps if sufficient representative training data are available. Existing spatial data can provide the approximate locations of corresponding geographic features in historical maps and thus be useful to annotate training data automatically. However, the feature representations, publication date, production scales, and spatial reference systems of contemporary vector data are typically very different from those of historical maps. Hence, such auxiliary data cannot be directly used for annotation of the precise locations of the features of interest in the scanned historical maps. This research introduces an automatic vector-to-raster alignment algorithm based on reinforcement learning to annotate precise locations of geographic features on scanned maps. This paper models the alignment problem using the reinforcement learning framework, which enables informed, efficient searches for matching features without pre-processing steps, such as extracting specific feature signatures (e.g. road intersections). The experimental results show that our algorithm can be applied to various features (roads, water lines, and railroads) and achieve high accuracy.