土地覆被及生态服务价值变化的多时间尺度模拟——以四川省渔子溪流域为例

根据渔子溪流域1986年和1994年的遥感影像土地覆被解译资料为基础,建立了模拟渔子溪流域生态环境变化的Markov Chain模型和Patch-dynamics模型,并以2002年遥感影像资料作为验证数据对模型进行了误差分析。以此为基础,分别以1年和8年的时间尺度对渔子溪流域土地覆被及其生态服务价值的变化进行了模拟,结果表明,时间尺度对模拟结果具有显著影响：在1年和8年的模拟尺度下,以Markov Chain模型得到的流域生态服务价值在1986~2018年间变化的相对误差为20%;不同土地覆被类型在不同步长下模拟结果的相对误差不一样,表明不同土地覆被类型变化的特征时间尺度不一样,以Patch-dynamics模型的模拟结果为例,步长为1年比步长为8年的相对误差大的土地覆被类型为耕地 (-8.2%/-5.6%)、有林地 (-0.5%/-0.4%)、草地 (0.7%/0.4%) 和居民建设用地 (-29.9%/-16.4%),它们的年际变化较大,其变化趋势不稳定,受到人为偶然因素的影响明显;而相对误差较小的土地覆被类型为灌木林 (-1.5%/-1.7%) 和永久冰雪覆盖 (27.3%/41.9%),它们的变化趋势较为稳定,主要受比较稳定的自然因素的影响。研究还表明,采用Markov Chain模型的模拟结果与采用Patch-dynamics模型的模拟结果总体上是一致的,但后者的模拟结果更稳定、更可靠。

Multi-scale Modeling of Land-cover Change and Ecosystem Service Values:A Case of the Yuzixi Catchment in Sichuan

The Yuzixi Catchment is located in the core of the Panda Natural Reserves in western Sichuan Province, with an area of 1742 km2. Due to influences of climate change and human activities, the land cover shows marked changes in the last decade. Using the land-cover data obtained from the remote sensing images of two years of 1986 and 1994, this paper developed two models based on the Markov Chain and Patch-dynamics approach, to simulate land-cover changes in the Yuzixi Catchment. Both models were verified using the land-cover data in 2002. Starting from the year 1994, the land-cover of the year 2002 was simulated in a time-step of 1 year and 8 years with these two models, respectively. Compared with the real land-cover data in 2002, it was found that the time scale had a marked effect on the simulation results, and the simulation error varied among different land use types. The simulation error of Patch-dynamics model of cultivated land (-8.2%/-5.6%), forest land (-0.5% /-0.4%), grassland (0.7%/0.4%) and residential area (-29.9% /-16.4% ) is bigger based on one-year scale than eight-year scale, and their changing trends are unstable due to the significant effect of anthropogenic factors. By contrast, the simulation error of bush land (-1.5%/-1.7%) and ice land (27.3%/41.9%) is smaller based on one-year scale than eight-year scale, and the changing trends are stable because they are mostly affected by natural factors. The research results also show that the simulation results of Markov Chain model are similar to that of Patch-dynamics model. However, the simulation results of Patch-dynamics model are more stable and reliable.