Social Structure-Land Use-Water Flows: Modelling Relationships using Discrete Bayesian Networks

Bayesian networks are one of the most powerful tools for the design ofexpert systems located in an uncertainty framework (probabilistic expert system).We have studied the relationships between social structure, land use and waterflows in two Spanish catchments using discrete Bayesian networks. The aim wasto predict how social changes influence both land use and green and blue waterflows. Land use in the Nacimiento catchment is related to woodland and traditionalagriculture, while land use in the Adra catchment comprises traditional andgreenhouse agriculture. The socioeconomic variables selected were emigrationrate, immigration rate, and the proportion of people older than 65 (p65). Green andblue water flows were calculated using the BalanceMED hydrological model. Datawere discretized into three intervals using the Equal Frequency method and aBayesian network was trained for each catchment. We studied two scenarios ofsocial evolution: emigration and ageing; and immigration. The results indicated thatin the Nacimiento catchment, social changes have little influence on changes inland use and water flows. The network for the Adra catchment showed the stronginfluence that social change plays on land use and water flow. The increase inemigration rates and p65, implies a decrease in immigration rate, an increase inwoodland uses and a decrease in agricultural uses. In this context, non-productivegreen water flow decreased. If there were an increase in immigration rate,emigration rate and p65 would decrease; woodland uses would decrease andgreenhouse and irrigated land would increase. Therefore, non-productive greenwater flows and consumptive blue water would increase. These results highlightthat the Nacimiento catchment has more resilience than the Adra catchment.