湿地生态系统是地球上最具有价值的生态系统之一，具有水源涵养、气候调节和物种多样性保护等重要生态系统服务，对环境的保护和人类福祉延续至关重要。近年来随着全球气候变化和社会经济的发展，湿地生态系统退化问题日益显著。为更加有效地实现湿地资源保护修复、可持续管理及利用，亟需精细可靠的湿地资源信息和生态系统服务的评估，以支持相关政府部门的政策规划。生态系统服务是指人类能够从生态系统得到的各种不同的服务，如支持服务、调节服务、文化服务和供给服务。生态系统为人类的生存和发展提供了各种各样的资源，包括提供人类生态必须的环境，可调节的生命支持系统，食物和其它生活所需材料，休闲娱乐和美学文化等。因此及时获取生态系统服务的变化，对于生态的健康发展和调节维护具有重要意义。国际湿地城市的提出是为了城市发展和湿地保护深度融合，是对城市湿地的生态保护认可的荣誉称号。
本研究以国际湿地城市武汉市、南昌市、合肥市为研究区，基于Landsat和Sentinel-2的2015年、2018年、2022年融合后的年平均遥感影像数据，使用随机森林算法结合知识规则的湿地空间遥感提取方法，开展了三个年份的武汉市、南昌市、合肥市湿地分类，实现了6种精细湿地类型和5种非湿地类型信息遥感提取，最后分析了湿地生态系统服务变化过程。本文主要研究内容和结论如下：
（1）基于随机森林算法和知识规则实现湿地精细类别空间信息遥感提取。本研究结合多源地理空间数据、遥感专题数据和目视解译生产土地利用样本点；其次在Google earth engine平台上，基于融合遥感影像和样本点数据，利用随机森林算法得到逐像元土地利用粗分类结果。再针对水体，根据知识规则进行二次分类，使用eCognition软件实现精细类型湿地空间信息遥感提取，最终得到10 m空间分辨率的6种湿地类型和5种非湿地类型土地覆盖信息空间分布数据。数据分类结果的OA和Kappa均在0.80以上，可保障后续分析结果的准确性。在2015年、2018年、2022年的，武汉市湿地面积分别为1644.01km2、1396.72km2、1896.19km2、南昌市湿地面积分别为1549.40km2、1133.01km2、1911.31km2合肥市湿地面积分别为1397.22km2、1381.82km2、1588.01km2。三个城市的湿地面积呈现先降低后增加的趋势。
（2）基于InVEST模型开展国际湿地城市生态系统服务评估和时空变化。使用InVEST模型对研究区的三个城市开展生态系统服务的水源涵养、碳储量、生境质量、土壤保持的功能评估。从评估的结果可知，三个城市的水源涵养功能主要受降雨量的影响，但是湿地范围内湿地的碳储量的变化受湿地类别变化的影响。三个城市湿地碳储量的功能在城市整体碳储量的占比较少，原因是三个城市主要的湿地类别是水体，水体的含碳量较低。三个城市的生境质量总体是增加的，生境质量的变化与湿地的增加相关，三个城市在整体湿地面积是增加的，但是湿地范围的单位面积的生境质量的值在降低。在土壤保持功能方面，三个城市的侵蚀量是减少的，土壤侵蚀值与降雨量相关，南昌市的变化最大，湿地范围内的土壤侵蚀增加，这与当地湿地类型的变化较大有关。
（3）湿地变化和城市政策背景在国际湿地城市中的作用。国际湿地城市需要满足湿地率和湿地保护率的指标，三个城市的湿地保护量化指标发生波动，但是在2022年均满足湿地率10%和湿地保护率50%。对三个城市的湿地变化典型区域对比变化分析，合肥市的湿地增加位于巢湖的河流汇入口周边，南昌的湿地增加位于鄱阳湖南部边界，武汉市的湿地增加位于官渡湖和黄家湖的周边。武汉和合肥典型区域的湿地增加和减少以人工湿地为主，南昌的湿地典型区域的湿地变化以自然湿地为主。三个城市均出台湿地保护政策，武汉的政策最为完备，合肥的政策结合自身湿地特点，南昌的湿地保护和动物保护结合，均取得较好成果。

Wetland ecosystems are one of the most valuable ecosystems on Earth, with important ecological values such as water conservation, climate regulation, and species diversity protection. They are crucial for environmental protection and human well-being. In recent years, with the intensification of rapid economic development and global climate change, the problem of wetland ecosystem degradation has become increasingly prominent. In order to achieve more effective protection, restoration, sustainable management, and utilization of wetland resources, there is an urgent need for precise and reliable wetland resource information to support policy planning and implementation by relevant government departments. Ecosystems provide a variety of resources for human survival and development, including providing the necessary environment for human ecology, adjustable life support systems, food and other necessary materials for life, leisure and entertainment, and aesthetic culture. Therefore, obtaining changes in ecosystem services is of great significance for the healthy development and regulation and maintenance of the ecosystem. The proposal of International Wetland City is aimed at the deep integration of urban development and wetland protection, and is an honorary title recognized for the ecological protection of urban wetlands.
This study focuses on the international wetland cities of Wuhan, Nanchang, and Hefei as research areas. Based on Landsat and Sentinel-2 remote sensing images from 2015, 2018, and 2022, a wetland spatial remote sensing extraction method framework using random forest algorithm and knowledge rule decision model was adopted. Wetland classification in Wuhan, Nanchang, and Hefei from 2000 to 2022 was carried out year by year, achieving remote sensing extraction of information for 6 fine wetland types and 5 non wetland types, Finally, the process of changes in wetland ecosystem service functions was analyzed. The main research content and conclusions of this article are as follows:
(1) Implementing remote sensing extraction of wetland fine category spatial information based on forest algorithm and knowledge rule decision model. This study combines multisource geospatial and remote sensing thematic data with visual interpretation to produce wetland samples; Secondly, based on the fusion of remote sensing images and sample data, the random forest algorithm is used to obtain pixel by pixel land use coarse classification results. Then, a knowledge rule decision model is used to achieve fine type wetland spatial information remote sensing extraction for water objects. Finally, the spatial distribution of land cover information for 6 wetland types and 5 non wetland types with a spatial resolution of 10 meters is obtained. The OA and Kappa of the classification results are both above 0.85, which can ensure the accuracy of subsequent analysis results.
(2) Conduct an assessment and spatiotemporal evolution of ecosystem service functions in international wetland cities. Use the InVEST model to conduct functional evaluations of ecosystem services for carbon storage, water conservation, soil conservation and habitat quality in three cities in the study area. From the evaluation results, it can be seen that the water conservation function of the three cities is mainly affected by rainfall, but the changes in water conservation of wetlands within the wetland area are mainly influenced by changes in the wetland's own category. The function of wetland carbon storage in the three cities is relatively small in the overall carbon storage of the cities, because the main wetland category in the three cities is water. The changes in habitat quality in the three cities are different, and the changes in habitat quality reflect the areas where wetlands are added in the three cities. As the overall wetland area increases, there is an increase and a decrease in habitat quality in the three cities. This indicates that the areas where wetland area increases are more in areas with strong human activities such as construction land and farmland. In terms of soil conservation function, the water systems of the three cities are different, and the distribution of DEM has its own characteristics. In the case of reduced rainfall, the soil erosion value of the three cities decreases, and the wetland area is affected by the changes in the river water system of the wetland itself, especially in Nanchang, where the changes are significant, resulting in an increase in soil erosion within the wetland area.
(3) The role of wetland changes and urban policy background in international wetland cities. The international wetland cities in the research area need to meet the indicators of wetland rate and wetland protection rate. The indicators of the three cities fluctuate, but in 2022, they will meet the wetland rate of 10% and wetland protection rate of 50%. Comparative analysis of typical regional changes in wetland changes in three cities shows that the increase in wetlands in Hefei is located around the river confluence of Chaohu Lake, the increase in wetlands in Nanchang is located at the southern boundary of Poyang Lake, and the increase in wetlands in Wuhan is located around Guandu Lake and Huangjia Lake. The increase and decrease of wetlands in typical areas of Wuhan and Hefei are mainly artificial wetlands, while the wetland changes in typical areas of wetlands in Nanchang are mainly natural wetlands. The carbon storage of three cities is positively correlated with wetland area. All three cities have introduced wetland protection policies, with Wuhan having the most comprehensive policies. Hefei's policies combine its own wetland characteristics, while Nanchang's wetland protection and animal protection have achieved good results.