• 地理空间分析综合应用 •

### CA-Markov与LCM模型的黄河三角洲湿地变化模拟比较

1. 1. 国家海洋环境监测中心,大连 116023
2. 辽宁师范大学城市与环境学院, 大连 1160293. 中南林业科技大学生命科学与技术学院,长沙 410004
• 收稿日期:2018-12-13 修回日期:2019-05-27 出版日期:2019-12-25 发布日期:2019-12-25
• 通讯作者: 曲丽梅 E-mail:lmqu2016@126.com
• 作者简介:陈柯欣（1994-）,女,辽宁营口人,硕士生,主要从事湿地遥感和应用研究。E-mail: 457261689@qq.com
• 基金资助:
国家重点研发计划专项(2017YFC0505901)

### Comparison of the CA-Markov and LCM Models in Simulating Wetland Change in the Yellow River Delta

CHEN Kexin1,2, CONG Pifu1, LU Weizhi3, QU Limei1,*()

1. 1. National Marine Environmental Monitoring Center, Dalian 116023, China
2. School of Urban and Environmental Science, Liaoning Normal University, Dalian 116029, China
• Received:2018-12-13 Revised:2019-05-27 Online:2019-12-25 Published:2019-12-25
• Contact: QU Limei E-mail:lmqu2016@126.com
• Supported by:
National Key Research and Development Program of China(2017YFC0505901)

Abstract:

The aim of this paper is to select the optimization model of the region and understand the future quantity and spatial variation trend of the wetland landscape types in the Yellow River Delta. We used the classified maps of the three periods of 1996, 2006 and 2016, of which the 1996 and 2006 maps were modeled for predicting 2016; we then compared the classified and simulated maps of 2016 to assess the model performances. The best model were used to take the classified 2006 and 2016 maps to simulate the landscape of the Yellow River Delta in 2026. We found that: ① For the simulation of the landscape types of the Yellow River Delta, under the influence of the same driving force factors, the LCM (Land Change Modeler) model performed better than the CA-Markov model in terms of spatial error, while CA-Markov was more suitable for the actual wetland change trend modeling than the LCM model in terms of numerical error. For the areas of larger landscape changes, the advantages of the two models should be combined to best simulate the change trend of wetlands. ② The interference of some human factors and the impact of natural disasters on the landscape types cannot be considered the model, it would cause some interference to simulation accuracy. For the LCM model, the number of transition sub-models had an effect on the simulation results with the same driving force factor, the more transition sub-models were used to generate suitable images, the higher the simulation accuracy. For CA-Markov model, the setting of proportional error coefficient was suitable for improving the accuracy of numerical simulation. ③ Assuming the continuation of the landscape dynamics trend during 2006-2016, and by simulation via combining the two simulation methods up to year 2026, the simulated natural wetlands area was 1252.69 km 2, the human-made wetlands area was 1265.00 km 2, and the non-wetlands area was 924.51 km 2. The simulated results suggest that natural wetlands and non-wetlands area will likely reduce, and human-made wetlands area will increase and expand to even shallow sea areas. Our findings can provide a scientific basis for the rational layout planning of the regional development space and the rational and effective utilization and management of wetland resources.