Improving the predictive performance of CLUE-S by extending demand to land transitions: The trans-CLUE-S model

•The trans-CLUE-S model employs demand at the finer level of land type transitions.•The trans-CLUE-S predictions were twice more accurate than the CLUE-S model's.•The trans-CLUE-S accuracy was less dependent on the amount of environmental data.•Algorithmic addition of a land assignment task enabled and sped up full convergence.•R code is provided for our models and auxiliary functions. The CLUE-S model is a popular choice for modelling land use and land cover change from local to regional scales, but it spatially allocates the demand for only the total cover of each land class in the predicted map. In the present work, we introduce a CLUE-S variant that allocates demand at the more detailed level of land type transitions: the trans-CLUE-S model. We implemented this extension algorithmically in R, without the need of new parameters. By processing each row of the land transition matrix separately, the model allocates the demand of each land category's transitions via the CLUE-S allocation routine for only the cells which were of that category in the map of the previous time step. We found that the trans-CLUE-S model had half the total and configuration disagreement of the CLUE-S predictions in an empirical landscape, and in simulated landscapes of different characteristics. Moreover, the trans-CLUE-S performance was less sensitive to the number of environmental predictors of land type suitability for allocating demand. Although trans-CLUE-S is computationally more demanding due to running a CLUE-S allocation for each land class, we appended the solution of a land-use assignment optimisation problem that facilitates the convergence and acceleration of allocation. We additionally provide R functions for: CLUE-S variants at other levels of demand resolution; random instead of environment-based allocation; and for simulating landscapes of desired characteristics. Our R code for the models and functions can contribute to more reproducible, transparent and accurate modelling, analysis and interpretation of land cover change.