基于CLUE-S的香溪河土地利用变化预测与总磷污染负荷分析

本研究旨在比较不同土地利用数量预测方法的适用性,并结合空间分布预测模型对香溪河流域未来的土地利用进行预测,以期为该区域的土地利用规划提供决策支持,有效控制总磷污染负荷的产生。结合土地利用数量预测的3种算法(线性外推法、马尔科夫链模型以及遗传算法)与空间预测模型CLUE-S(Conversion of land use and its effect at small regional extent),对流域内的土地利用变化进行预测,通过输出系数法对流域内的总磷污染负荷进行估算。结果表明,基于线性外推与马尔科夫链两种无约束预测模型,2020年流域内林地面积将比2010年减少约1%,绝大多数林地转变为水田与旱地,且大多发生在流域中部坡度较平缓区域。而基于存在自然社会经济约束条件的遗传算法优化情景下,水田面积减少1060 hm2、旱地面积减少3370 hm2,其面积主要转化为林地,且大多发生在流域高海拔、较陡峭的北部区域。基于输出系数法分析得到的流域内总磷污染负荷在线性外推与马尔科夫链预测情景下相比于2010年均有所增加,分别增加11 000 kg和8000 kg,而在遗传算法情景下,总磷负荷相比于2010年减少约24 000kg。空间分布上,在线性外推与马尔科夫链情景下增加的负荷主要位于流域中部区域,而在遗传算法情景下流域北部区域总磷负荷减少量最为明显。研究结果表明,遗传算法在土地利用优化预测方面表现优异,结合CLUE-S模型,可以对未来土地利用规划起到一定的支持作用,有效控制非点源污染负荷的产生。

Land use change predictions based on the CLUE-S model and total phosphorus load analysis

In this study, we analyzed the feasibility of three quantitative land use prediction methods. These three models were the linear extrapolation method, the Markov chain model, and the genetic algorithm. They were combined with the CLUE-S model to predict land use changes in 2020 in the Xiangxi Watershed. Our research provided supporting information for land use plans in the study area and gave indications to reduce the discharge of non-point source pollution. When using either the linear extrapolation method or the Markov chain model, the forest area decreased by more than 1% from 2010 to 2020. This area mainly changed to paddy fields and dry land, which was located in the central area of the watershed on a relatively gentle slope. However, when using the genetic algorithm and appropriate environmental, social, and economic constraints, the areas of paddy field and dryland decreased by 1060 hm² and 3370 hm², respectively. This area was mainly converted to forest located in the north of the watershed at high altitude on steep slopes. Based on the export coefficient method, the total phosphorus loads from the whole watershed were calculated to be 11 000 kg and 8000 kg higher in 2020 than in 2010, using the linear extrapolation method and the Markov chain method, respectively. The phosphorus load in 2020 predicted by the genetic algorithm was 24 000 kg lower than in 2010. Spatially, the increased phosphorus load and decreased phosphorus load mainly occurred in the central area and the northern area of the watershed, respectively. This research thus compared three land use area prediction methods and identified the best one. The land use structure in 2020 was simulated and the total phosphorus load and distribution were predicted by integrating the quantity prediction method and the spatial land use distribution model. The result would provide a good reference for future land use planning in the study area.