Multiple land use change simulation with Monte Carlo approach and CA-ANN model, a case study in Shenzhen, China

Environmental Systems Research

Table 1 Class kappa coefficients and overall accuracy for each scenarios

Scenario*	Woodland	Cropland	Wetland	Water body	Build-up	*Overall Kappa*	Overall accuracy (%)
I	0.72	0.49	0.87	0.79	0.68	0.71	81.91
II	0.76	0.33	0.55	0.71	0.77	0.73	83.13
III	0.79	0.46	0.76	0.78	0.74	0.76	85.33

*Scenario I: training samples from 1996 and 2000 land use maps are input and output of the model respectively, the map of 1996 was input to the model and the map of 2000 was predicted and evaluated with the actual map of 2000 using testing samples.
Scenario II: training samples from 1996 and 2000 land use maps are input and output of the model respectively, the map of 2000 was input to the model and the map of 2004 was predicted and evaluated with the actual map of 2004 using testing samples.
Scenario III: training samples from 2000 and 2004 land use maps are input and output of the model respectively, the map of 2000 was input to the model and the map of 2004 was predicted and evaluated with the actual map of 2004 using testing samples.