Bryan C. Pijanowski

Using neural networks and GIS to forecast land use changes: a Land Transformation Model

The Land Transformation Model (LTM), which couples geographic information systems (GIS) with artificial neural networks (ANNs) to forecast land use changes, is presented here. A variety of social, political, and environmental factors contribute to the model’s predictor variables of land use change. This paper presents a version of the LTMparameterized for Michigan’s Grand Traverse Bay Watershed and explores how factors such as roads, highways, residential streets, rivers, Great Lakes coastlines, recreational facilities, inland lakes, agricultural density, and quality of views can influence urbanization patterns in this coastal watershed. ANNs are used to learn the patterns of development in the region and test the predictive capacity of the model, while GIS is used to develop the spatial, predictor drivers and perform spatial analysis on the results. The predictive ability of the model improved at larger scales when assessed using a moving scalable window metric. Finally, the individual contribution of each predictor variable was examined and shown to vary across spatial scales. At the smallest scales, quality views were the strongest predictor variable. We interpreted the multi-scale influences of land use change, illustrating the relative influences of site (e.g. quality of views, residential streets) and situation (e.g. highways and county roads) variables at different scales. # 2002 Elsevier Science Ltd. All rights reserved.

Soundscape Ecology: The Science of Sound in the Landscape

This article presents a unifying theory of soundscape ecology, which brings the idea of the soundscape—the collection of sounds that emanate from landscapes—into a research and application focus. Our conceptual framework of soundscape ecology is based on the causes and consequences of biological (biophony), geophysical (geophony), and human-produced (anthrophony) sounds. We argue that soundscape ecology shares many parallels with landscape ecology, and it should therefore be considered a branch of this maturing field. We propose a research agenda for soundscape ecology that includes six areas: (1) measurement and analytical challenges, (2) spatial-temporal dynamics, (3) soundscape linkage to environmental covariates, (4) human impacts on the soundscape, (5) soundscape impacts on humans, and (6) soundscape impacts on ecosystems. We present case studies that illustrate different approaches to understanding soundscape dynamics. Because soundscapes are our auditory link to nature, we also argue for their protection, using the knowledge of how sounds are produced by the environment and humans.

What is soundscape ecology? An introduction and overview of an emerging new science

We summarize the foundational elements of a new area of research we call soundscape ecology. The study of sound in landscapes is based on an understanding of how sound, from various sources—biological, geophysical and anthropogenic—can be used to understand coupled natural-human dynamics across different spatial and temporal scales. Useful terms, such as soundscapes, biophony, geophony and anthrophony, are introduced and defined. The intellectual foundations of soundscape ecology are described—those of spatial ecology, bioacoustics, urban environmental acoustics and acoustic ecology. We argue that soundscape ecology differs from the humanities driven focus of acoustic ecology although soundscape ecology will likely need its rich vocabulary and conservation ethic. An integrative framework is presented that describes how climate, land transformations, biodiversity patterns, timing of life history events and human activities create the dynamic soundscape. We also summarize what is currently known about factors that control temporal soundscape dynamics and variability across spatial gradients. Several different phonic interactions (e.g., how anthrophony affects biophony) are also described. Soundscape ecology tools that will be needed are also discussed along with the several ways in which soundscapes need to be managed. This summary article helps frame the other more application-oriented papers that appear in this special issue.

Calibrating a neural network‐based urban change model for two metropolitan areas of the Upper Midwest of the United States

We parameterized neural net‐based models for the Detroit and Twin Cities metropolitan areas in the US and attempted to test whether they were transferable across both metropolitan areas. Three different types of models were developed. First, we trained and tested the neural nets within each region and compared them against observed change. Second, we used the training weights from one area and applied them to the other. Third, we selected a small subset (∼1%) of the Twin Cities area where a lot of urban change occurred. Four model performance metrics are reported: (1) Kappa; (2) the scale which correct and paired omission/commission errors exceed 50%; (3) landscape pattern metrics; and (4) percentage of cells in agreement between model simulations. We found that the neural net model in most cases performed well on pattern but not location using Kappa. The model performed well only in one case where the neural net weights from one area were used to simulate the other. We suggest that landscape metrics are good to judge model performance of land use change models but that Kappa might not be reliable for situations where a small percentage of urban areas change.

A big data urban growth simulation at a national scale: Configuring the GIS and neural network based Land Transformation Model to run in a High Performance Computing (HPC) environment

The Land Transformation Model (LTM) is a Land Use Land Cover Change (LUCC) model which was originally developed to simulate local scale LUCC patterns. The model uses a commercial windows-based GIS program to process and manage spatial data and an artificial neural network (ANN) program within a series of batch routines to learn about spatial patterns in data. In this paper, we provide an overview of a redesigned LTM capable of running at continental scales and at a fine (30m) resolution using a new architecture that employs a windows-based High Performance Computing (HPC) cluster. This paper provides an overview of the new architecture which we discuss within the context of modeling LUCC that requires: (1) using an HPC to run a modified version of our LTM; (2) managing large datasets in terms of size and quantity of files; (3) integration of tools that are executed using different scripting languages; and (4) a large number of steps necessitating several aspects of job management. Display Omitted Reconfigure the Land Transformation Model (LTM) using high performance computing.We present the design of software for managing big data simulations.We integrate software environments, such as Python, XML, ArcGIS, SNNS, and the .NET framework.We executed 285 instances of ArcGIS on an HPC.

Modeling multiple land use changes using ANN, CART and MARS: Comparing tradeoffs in goodness of fit and explanatory power of data mining tools

Abstract Over half of the earths terrestrial surface has been modified by humans. This modification is called land use change and its pattern is known to occur in a non-linear way. The land use change modeling community can advance its models using data mining tools. Here, we present three data mining land use change models, one based on Artificial Neural Network (ANN), another on Classification And Regression Trees (CART) and another Multivariate Adaptive Regression Splines (MARS). We reconfigured the three data mining models to concurrently simulate multiple land use classes (e.g. agriculture, forest and urban) in South-Eastern Wisconsin (SEWI), USA (time interval 1990–2000) and in Muskegon River Watershed (MRW), Michigan, USA (time interval 1978–1998). We compared the results of the three data mining tools using relative operating characteristic (ROC) and percent correct match (PCM). We found that ANN provided the best accuracy in both areas for three land use classes (e.g. urban, agriculture and forest). In addition, in both regions, CART and MARS both showed that forest gain occurred in areas close to current forests, agriculture patches and away from roads. Urban increased in areas of high urban density, close to roads and in areas with few forests and wetlands. We also found that agriculture gain is more likely for the areas closer to the agriculture and forest patches. Elevation strongly influenced urbanization and forest gain in MRW while it has no effect in SEWI.

A REVISION OF LACK'S BROOD REDUCTION HYPOTHESIS

A recent review by T. Amundsen and J. N. Stokland showed that, when food was plentiful, last-hatched nestlings of asynchronously hatched broods died more frequently than their nestmates or nestlings in synchronously hatched broods. They argued that if last-hatched nestlings die as a consequence of hatching asynchrony, then brood reduction must be maladaptive. I construct a model similar to a model by D. M. Temme and E. Charnov. In my model, asynchronously hatched broods experience reduced survival rates of the last-hatched nestling during good food years, and synchronously hatched broods experience entire brood starvation during bad food years. I compare the reproductive performance of parents raising an asynchronously hatched brood to parents raising a synchronously hatched brood. I show that hatching asynchrony is favored over synchronous hatching when good food years are not very frequent, when the survival rate of last-hatched nestlings during good food years is high, when the survival rate of nestlings raised in synchronously hatched broods during bad food years is low, or when bad food years are not severe. I discuss the models assumptions and predictions, and I compare the results of this model to Temme and Charnovs model.

Comparing three global parametric and local non-parametric models to simulate land use change in diverse areas of the world

This paper compares one global parametric land use change model, the artificial neural network - based Land Transformation Model, with two local non-parametric models: a classification and regression tree and multivariate adaptive regression spline model. We parameterized these three models with identical data from different regions of the world; one region undergoing extensive agricultural expansion (East Africa), another region where forests are re-growing (Muskegon River Watershed in the United States), and a third region where urbanization is prominent (South-Eastern Wisconsin in the United States). Independent training data and testing data were used to train and calibrate each model, respectively. Comparisons of simulated maps from the three kinds of land use change patterns were made using conventional goodness-of-fit metrics in land use change science. The results of temporal and spatial comparison of the data mining models show that the artificial neural network outperformed all other models in a short-time interval (East Africa; 5 years) and for coarse resolution data (East Africa; 1 km); however, the three data mining models obtained similar accuracies in a long-time interval (Muskegon River Watershed; 20 years) and for fine resolution data with large numbers of cells (Muskegon River Watershed; 30 m). Furthermore, the results showed that the probability of agriculture gain was more likely in locations closer to towns and large cities in East Africa, urbanization was more likely in locations closer to roads and urban areas in South-Eastern Wisconsin and the probability of forest gain was more likely in locations closer to the forest and shrub land cover and farther away from roads in Muskegon River Watershed.

Hydroclimatic Response of Watersheds to Urban Intensity: An Observational and Modeling-Based Analysis for the White River Basin, Indiana

Impervious surface area (ISA) has different surface characteristics from the natural land cover and has great influence on watershed hydrology. To assess the urbanization effects on streamflow regimes, the authors analyzed the U.S. Geological Survey (USGS) streamflow data of 16 small watersheds in the White River [Indiana (IN)] basin. Correlation between hydrologic metrics (flow distribution, daily variation in streamflow, and frequency of high-flow events) and ISA was investigated by employing the nonparametric Mann‐Kendall method. Results derived from the 16 watersheds show that urban intensity has a significant effect on all three hydrologic metrics. The Variable Infiltration Capacity (VIC) model was modified to represent ISA in urbanized basins using a bulk parameterization approach. The model was then applied to the White River basin to investigate the potential ability to simulate the water and energy cycle response to urbanization. Correlation analysis for individual VIC grid cells indicates that the VIC urban model was able to reproduce the slope magnitude and mean value of the USGS streamflow metrics. The urban model also reproduced the urban heat island (UHI) seen in the Moderate Resolution Imaging Spectroradiometer (MODIS) land surface temperature products, especially for the grids encompassing the city of Indianapolis, IN. The difference of the hydrologic metrics obtained from the VIC model with and without urban representation indicates that the streamflow regime in the White River has been modified because of urban development. The observed data, together with model analysis, suggested that 3%‐5% ISA in a watershed is the detectable threshold, beyond which urbanization effects start to have a statistically significant influence on streamflow regime.

Modelling urbanization patterns in two diverse regions of the world

We present work applying a similarly parameterized urbanization model to two diverse regions of the world, one in the USA and another in Albania. Eight calibration metrics are used to estimate model goodness of fit: four location-based measures (e.g. kappa), and four patch metrics based on patch size, shape and configuration. We conclude that if we use location goodness of fit estimates, the model fits observed data very well for most simulations. The model fit to data better in Albania than in the USA probably owing to top-down land ownership policies occurring in Albania and owing to the fact that commonly used land use change model drivers, such as distance to road, are not likely to capture individual behaviours that are important in the USA. Patch metrics provided additional information on model fit to observed data, and we suggest that, in some circumstances, patch metrics may be more useful than location metrics to calibrate a land use change model.