Use of Synthetic Aperture Radar (SAR) images for fuel moisture modelling and land cover mapping in the case of boreal forests and natural grasslands

北方森林和天然草原是世界上两个重要的生态系统，但它们也容易发生火灾。加拿大火灾危险评级系统的火灾天气指数系统在世界许多地方使用，主要根据天气数据对火灾危险进行评级，这些数据是点源测量。一个有希望的替代方法是使用卫星数据。我的研究计划的长期目标是利用遥感技术绘制北方森林和天然草原的火灾天气指数代码。在短期内，我们将利用加拿大/南非天然草原和阿拉斯加北方森林持续合作中获得的图像和数据，㈠比较同时使用极化合成孔径雷达、光学和其他地理空间数据进行土地覆盖制图的两种分类器; ㈡开发一个经验模型，估计天然草原的FWI干旱代码;以及（iii）使用polSAR图像测试基于物理的DC模型。这项研究将使用雷达卫星2号C波段和ALOS-PALSAR L波段polSAR图像。我们的研究团队在使用SAR图像绘制FWI代码和指数方面拥有独特的专业知识。拟议的研究在许多方面延续了这一创新趋势。从分类器的角度来看，本文提出的遗传算法-神经网络分类器首次用于北方森林和天然草原生态系统的分类。从直流建模的角度来看，所有基于极化SAR数据的土壤水分模型主要是针对农业领域开发的。这是第一次，DC的天然草地将使用polSAR数据建模，并提出了一个基于物理的模型来估计DC从polSAR图像采集的北方森林和天然草地。在加拿大，拟议的研究涉及加拿大森林部长理事会的火灾研究优先事项。在全世界范围内，利用遥感更好地预测FWI代码和指数将从经济和人类安全的角度来看具有显着的好处。拟议中的研究有可能为加拿大带来重大的经济进步，因为它将为MDA公司带来新的市场机会，R-2的拥有者

Boreal forests and natural grasslands are two important ecosystems in the world, but they are also prone to fires. The Fire Weather Index (FWI) system of the Canadian Fire Danger rating system is used in many places in the world to rate fire dangers based primarily on weather data, which are point-source measurements. A promising alternative is using satellite data. The long term goal of my research program is to map Fire Weather Index codes for boreal forests and natural grasslands using remote sensing. In the short term, using images and data available through on-going collaborations on Canadian/South African natural grasslands and Alaska boreal forests, we will (i) compare two classifiers that simultaneously use polarimetric synthetic aperture radar (polSAR), optical and other geospatial data for land cover mapping; (ii) develop an empirical model to estimate the FWI drought code (DC) for natural grasslands; and (iii) test a physics-based DC model using polSAR images. The study will use both RADARSAT-2 C-band and ALOS-PALSAR L-band polSAR images. Our research team has built unique expertise on the use of SAR images for mapping FWI codes and indices. The proposed research continues this innovative trend in many ways. From the classifier point of view, it is the first time that the proposed genetic algorithm-neural network classifier will be use to classify boreal forest and natural grassland ecosystems. From the DC modeling point of view, all the soil moisture models using polSAR data were mainly developed for agricultural fields. It is the first time that DC of natural grasslands will be modeled using polSAR data and that a physics-based model will be proposed to estimate DC from polSAR images acquired over both boreal forest and natural grasslands. In Canada, the proposed research addresses fire research priorities of the Canadian Council of Forest Ministers. Across the world, better prediction of FWI codes and indices using remote sensing will have significant benefits both from the economical and the human safety points of views. The proposed research has a potential for significant economic advancement for Canada, as it will produce new market opportunities for MDA Inc., the R-2 owner.