http://om.ciheam.org/om/pdf/c01-2/95605231.pdf

http://www.wgianalytics.net/wgitransport/docs/comparisons/USDA_Review.pdf

http://www.fao.org/3/a-a1080e.pdf

https://www.researchgate.net/publication/227125122_Land-use_change_modeling_Current_practice_and_research_priorities

Article Land-use change modeling: Current practice and research priorities

The third chapter of this book gives a good overview of various land use simulation models: https://books.google.hr/books?id=FHJ9NVr6IMIC&pg=PA35&lpg=PA35&dq=Core+Principles+and+Concepts+in+Land-Use+Modelling:+A+Literature+Review&source=bl&ots=mjUJ8vHl-d&sig=h-OuFoAMvezZ691zPige4-le1VA&hl=hr&sa=X&ved=0ahUKEwiW_u-8z5rTAhWGlCwKHQoPBeoQ6AEIJDAB#v=onepage&q=Core%20Principles%20and%20Concepts%20in%20Land-Use%20Modelling%3A%20A%20Literature%20Review&f=false

Dear Tomasz, 

In the master thesis, Cellular automata modelling and urban simulation (http://hdl.handle.net/2099.1/24250) and in the dissertation of the PHD thesis, Multi-scale integrated cellular modelling for the study of urban change phenomena (http://www-cpsv.upc.es/tesis/presentacio_norte.pdf), both authored by Nuno Norte Pinto, you will find information regarding Cellular Automata models to develop urban simulation technology for assisting planning processes.

And, these references could be usefully;

- BARREDO, J., M. Kasanko, N. McCormick and C. Lavalle (2003). Modelling dynamic spatial processes: simulation of urban future scenarios through cellular automata. In Landscape and Urban Planning, 64: 145-160

- BENENSON, I. and P. M. Torrens (2004). Geosimulation - Automata-based modeling of urban phenomena. Chechester, John Wiley & Sons Ltd

- COUCLELIS, H. (1997). From cellular automata to urban models: new principles for model development and implementation. In Environment and Planning, 24: 165-174

- COUCLELIS, H. (2005). Where has the future gone? Rethinking the role of integrated land-use models in spatial planning. In Environment and Planning, 37: 1353-1371

- LEE, D. (1994). Retrospective on large-scale urban models. In Journal of the American Planning Association, 60(1): 35-40

- WADDELL, P. and G. Ulfarsson (2004). Introduction to urban simulation: design and development of operational models. Handbook of Transport Geography and Spatial Systems, Volume 5 (Handbooks in Transport). D. Hensher, K. Button, K. Haynes and P. Stopher, Elsevier Science

I hope help you.

Best regards.

Rolando.

Hi there. The following is my opinion, references mostly from memory, caveat emptor!

1) agent-based models;

hard to calibrate, especially if there are many agents, not always spatially explicit. Probably better for understanding the process of change rather than creating future simulations, especially if the main goal is to understand the interaction of the agents of change. But recent work is beginning to address these limitations, e.g. see the CRAFTY model (Murray-Rust et al)

2) artificial neural networks;

rules of transition are calculated by the neural network and are invisible to the user - the model is effectively a "black box", these models have a tendency to overfit to the calibration date. On the other hand, they do produce good simulations by standard goodness-of-fit measures. See Pijanowski and collaborators, from 2005 (Grand Traverse Bay Watershed Michigan and numerous other cases up to the present day)

3) cellular automata;

most practical examples rely on trial and error calibration, rather than empirical evidence of association between land use and drivers. But transition rules are transparent (unlike ANN) since the user must explicitly define them. The modeller achieves a fuller understanding of the change processes and they perform well on pattern-based or cell-by-cell statistical measures. Convincingly model the dynamics of urban growth. See work by Keith Clarke and collaborators (1997, 1998 etc), White and Engelen 1993, White et al 1997 (city of Cincinnati). See also work on automatic or semi-automatic calibration of these models - Straatman, B., White, R., & Engelen, G. (2004). Towards an automatic calibration procedure for constrained cellular automata. Computers, Environment and Urban Systems, 28(1), 149-170, and also more recent work (see esp. Barreira-Gonzalez et al 2015, and Norte Pinto and Pais Antunes 2010)

4) economics-based models; and

a range of econometric models exist that claim to reliably associate the size of cities with three key drivers, distance from central business district, agricultural land value, and population. Problem is that this strong association at small spatial scales (e.g. all US cities) breaks down at large spatial scales (e.g within an individual region). Also does not really explain the process or pattern of change. See Brueckner and Fansler 1984, etc etc also critique by Bockstael and Irwin 1999

http://ageconsearch.tind.io//bitstream/197860/2/agecon-maryland-99-04.pdf

We talked about some of these issues in a recent working paper, maybe its of interest.

http://pure.iiasa.ac.at/13939/1/WP-16-016.pdf

5) Markov chains;

Don't know, I've not looked into these in detail.

Hope this helps!

Thank you very much for all answers and sugestions!

Maybe ANN can provide better result as the drivers of land use change will give some strength to perform better.

Logistic regression (logit) models are a common approach to model urban expansion. They predict the outcome of a categorical variables using a set of quantitative and/or qualitative predictors. The model’s ability to include as many factors as necessary allows us to better understand the main drivers behind urbanization processes. Neighborhood interactions can also be captured in Logit models by including them as part of the explanatory variables as in Hu and Lo (2007) and Verburg et al. (2004). However, because Logit models are not temporally explicit, they cannot reveal the path-dependent and self-organizing development which is typical for urban expansion. The most well-known approach to calculating the neighborhood interactions on a dynamic basis is cellular automata (CA) based model, in which the neighborhood state is updated during each simulation step. Cellular models are simple and widely available (Clarke and Gaydos, 1998). However, pure CA models focus on the calculation of land use transitions by explicitly consider the immediate neighbors of each landscape unit, i.e. cell, rather than on the interpretation of drivers of urban expansion. Several studies try to overcome this limitation of pure CA models by integrating CA with other modeling methods to consider several urbanization drivers. In this context, logit and CA are commonly combined to create a so-called ‘CA-logit model’, which considers both the urbanization static drivers and the dynamic neighborhood interactions.

One of the clear drawbacks of CA-logit approach is related to the lack of theoretical links between the spatial transitions rules and agents within the urban environment and their decisions. Agent-Based (AB) models, which are less frequently used in the context of urban expansion modeling, consider agents as goal-oriented entities capable of responding to their environment and interacting with each other. The agents are commonly grouped into homogeneous sets of individuals with comparable characteristics and behaviors. Generally, the decision-making criteria of agents require a large amount of data stemming from surveys that depict people's choices and utilize experts’ knowledge. In a large study area, such an intensive data gathering is limited by the presence of a large number of agents (Valbuena et al., 2008). In order to overcome data limitations, a number of studies used empirical data, such as distance to road network, slope etc., to represent agents decision-making for which we have no behavioral information (Mustafa et al., 2017).

Mustafa, A., Cools, M., Saadi, I., & Teller, J. (2017). “Coupling agent-based, cellular automata and logistic regression into a hybrid urban expansion model (HUEM).” Land Use Policy, 69C, 529–540. https://doi.org/10.1016/j.landusepol.2017.10.009

http://om.ciheam.org/om/pdf/c01-2/95605231.pdf