Wetland Health and Carbon Stocks Monitoring Using Advanced Machine Learning Algorithms and Remote Sensing Data

在过去二十年中，地球观测数据在许多环境应用中的价值得到了充分的证明，如使用常规遥感技术进行森林和湿地监测。虽然这些技术是非常有用的，数据的可用性，极大地依赖于功能工程和手工编辑，以及不够的准确性，限制了他们的应用范围小规模的研究领域。然而，最近，遥感技术、人工智能（AI）算法和强大的云计算基础设施（如谷歌地球引擎（GEE））的进步解决了传统技术的局限性。尽管最近取得了一些进展，但由于在以下方面缺乏足够的研究和开发，利用这些新的湿地监测方法仍然有限：消除或减少与需要大量地面实况数据有关的障碍;向减少人类参与数据处理水平的新技术迈进;加强数据处理技术，以提高产出的准确性和可解释性;并逐步对大规模调查进行地球观测大数据分析。在这方面，申请人的研究计划（10年以上）的长期目标是利用先进的人工智能工具和遥感数据在全国范围内开发最先进的开放获取湿地监测工具箱，以支持加拿大丰富的生物多样性。为了实现长期研究目标，这个短期研究计划将解决与湿地监测工具箱开发相关的挑战。特别是，所提出的解决方案的专题应用福尔斯三个主要的重要类别，即湿地分类，湿地变化检测，湿地地上生物量（AGB）估计，所有这些都是必不可少的湿地保护和管理。 拟议的研究成果有可能大大推进和彻底改变湿地测绘和监测技术，特别是在大规模。在这项研究计划中开发的湿地监测工具箱将捕获有关湿地状况的全面知识，在省级范围内，有很大的能力在全国范围内和加拿大境外进行升级，从而有助于全球范围的湿地知识和改善国家和全球协议和政策。最后，该计划的成果也可以很容易地转移到环境监测的其他领域，如森林监测。高素质人员将在多学科环境中接受培训，并将能够向加拿大学术界和工业界传授他们的知识和技能，利用最先进的地理空间技术广泛应用于加拿大自然资源的测绘和监测。

Over the past two decades, the value of Earth Observation (EO) data has been well documented for many environmental applications, such as forest and wetland monitoring, using conventional remote sensing techniques. Although these techniques are of great useful, data availability, great dependency on feature engineering and manual editing, as well as insufficient accuracy, limited their applications to small-scale study areas. However, most recently, advances in remote sensing technologies, artificial intelligence (AI) algorithms, and powerful cloud computing infrastructure, such as Google Earth Engine (GEE), have addressed the limitations of traditional techniques. Despite the recent advances, exploiting these new methodologies for wetland monitoring has remained limited due to the lack of sufficient research and development with respect to the following aspects: removing or reducing barriers related to the need for large volume of ground truth data; advancing toward new technologies that decrease the level of human involvement in data processing; enhancing techniques of data processing to raise the level of accuracy and interpretability of the outputs; and moving toward geo big data analysis for large-scale investigations. In this regard, the long-term objective of the applicant's research program (10+ years) is to develop a state-of-the-art open-access wetland monitoring toolbox at a national scale using advanced AI tools and remote sensing data to support Canada's rich biodiversity. To meet the long-term research goal, this short-term research program will address the challenges related to the development of the wetland monitoring toolbox. Particularly, the thematic application of the proposed solutions falls within three major important categories, namely wetland classification, wetland change detection, and wetland above ground biomass (AGB) estimation, all of which are essential for wetland protection and management. The outcomes of the proposed research have the potential to significantly advance and revolutionize the techniques of wetland mapping and monitoring, particularly at a large scale. The wetland monitoring toolbox developed in this research program will capture comprehensive knowledge about the status of wetlands at a provincial scale with a great capacity to upgrade at the national scale and beyond the borders of Canada, thus contributing to global-scale wetland knowledge and the improvement of both national and global agreements and policies. Finally, the achievements of this program can also be easily transferred to other domains of environmental monitoring, such as forest monitoring. The highly qualified personnel (HQP) will be trained in a multidisciplinary environment and will be able to transfer their knowledge and skills to Canadian academia and industries in the wide variety of applications for mapping and monitoring Canada's natural resources using state-of-the-art geospatial technologies.