A Formally and Algorithmically Efficient LULC change Model-Building Environment

TL;DR

This paper introduces a formal, efficient framework for land use and land cover change modeling, improving accuracy and reducing biases through novel calibration and allocation algorithms, validated by a real-world case study.

Contribution

It presents a new modeling environment with a kernel-based calibration method and an unbiased allocation algorithm, enhancing accuracy and efficiency over existing models.

Findings

Significant reduction in estimation error compared to existing software

Methods require few user-defined parameters

Framework is computationally efficient and unbiased

Abstract

The use of spatially explicit land use and land cover (LULC) change models is widespread in environmental sciences and of interest in public decision-help. However, it appears that these models suffer from significant biases and shortcomings, the sources of which can be mathematical, conceptual or algorithmic. We formalize a modeling environment that distinguishes a calibration-estimation module and an allocation module. We propose an accurate calibration-estimation method based on kernel density estimation and detail an unbiased allocation algorithm. Moreover, a method of evaluation of LULC change models is presented and allows us to compare them on various fronts (accuracy, biases, computational efficiency). A case study based on a real land use map but with known (enforced) transition probabilities is used. It appears that the estimation error of the methods we propose is substantially improved over the best existing software. Moreover, these methods require the specification of very few parameters by the user, and are numerically efficient. This article presents an overview of our LULC change modeling framework; its various formal and algorithmic constituents will be detailed in forthcoming papers.