森林是一种主要的自然资源，在维持环境平衡中发挥着重要作用，而频发的森林火灾会直接导致森林退化，严重威胁地表植被和野生动物的生命健康，及时、准确地检测森林火点具有重要意义。MODIS时空分辨率、光谱分辨率、仪器参数等方面的设计使其在火灾监测方面具有显著优势，但是MODIS火点产品空间分辨率低，且仅代表像元含火，不包含亚像元火点特性。本文针对MODIS数据的时间分辨率优势，提出时间序列分析和上下文空间分析相结合的MODIS火点检测方法，对MODIS V4火点检测算法进行改进；基于能量平衡方程，建立火点发射辐射端元库、背景辐射端元库以及阴影光谱库，利用MESMA对火点像元进行混合像元分解，得到亚像元火点面积和温度。本文主要研究工作及成果总结如下：1）改进MODIS V4算法，提出基于时间序列分析和空间上下文分析的火点检测方法。针对MODIS V4火点检测算法设定固定阈值判别潜在火点而没有考虑区域性差异的不足，研究基于时间序列断点识别算法的潜在火点判别和基于空间上下文分析的相对火点判别。结果表明：相对于MODIS V4火点检测算法，本文的火点检测算法提高了火点检测的正确率，对于面积很小且造成的NDVI和亮温时间序列数据趋势几乎没有变化的火点可能出现漏判情况。2）研究基于多端元光谱混合分析（MESMA）的亚像元火点特性。针对MESMA亚像元火点研究中，不包含烟羽背景，限制其使用范围的情况，将背景为烟羽的光谱信息加入反射辐射端元光谱库，基于MESMA反演得到亚像元火点信息。结果表明：MESMA反演的火点面积与MOD14的火点辐射功率FRP不具有统计相关性；MESMA反演的火点面积与AMS火点像元数相关性（r=0.743，p<0.01）高于FRP与AMS火点像元数相关性（r=0.293），烟羽端元的加入对二者极其显著的相关关系（仅有植被p<0.01）没有改变；烟羽端元的加入使MESMA反演的火点面积与AMS火点像元数极其显著的相关性（r=-0.707）降低，但二者相关性仍是显著的（r=-0.339）。总体而言，烟羽信息的加入扩大了MESMA使用范围。本文创新点有：1）引入时间序列断点识别算法BFAST提取潜在火点，克服固定阈值区域性限制；2）MESMA反射辐射端元光谱库加入了烟羽信息，提高了MESMA的适用范围，丰富了亚像元火点分解结果信息。

As major natural resource, forest plays an important role in maintaining environmental balance. However, frequent forest fire could directly lead to forest degradation; seriously threaten the vegetation and wildlife, which makes timely and accurate fire detection very important. The spatial, temporal, spectral resolution and the instrument parameters of MODIS enable it with apparent advantage in monitoring fire, but MODIS fire product has low spatial resolution, which only indicates the fire pixel, not sub-pixel fire characteristics.This paper proposes the improved algorithm of detecting forest fire based on times series analysis and contextual analysis. Based on energy balance equation, generate fire emission radiance library, background reflection radiance library and shade radiance library, through multimember spectral mixture analysis, fire temperature and fire size is derived. The main work and conclusions from this paper are as follows:1) Propose the improved MODIS V4 algorithm to detect forest fire based on times series analysis and contextual analysis. The results showed that: compared with MODIS V4 fire detection algorithm, the method put forward by this paper improves the accuracy of fire detection and the fire may be omitted when its size is small. 2) Study on sub-pixel fire characteristics based on multimember spectral mixture analysis (MESMA). Given the insufficiency of MESMA in studying sub-pixel fire characteristics, add plume radiance spectral to background reflection radiance endmember library and derive the fire area and temperature based on MESMA. The results showed that: fire area retrieved by MESMA has no correlation with FRP; The correlation between fire area retrieved by MESMA and AMS fire pixels （r=0.743，p<0.01）is higher than the correlation between FRP and AMS fire pixels（r=0.293）.Adding smoke endmember does not change the significant correlation of the two; The adding of smoke endmember lowers the significant correlation between fire area retrieved by MESMA and AMS fire pixels(r=-0.707), with the two still having significant correlation (r=-0.339). Overall ,the add of plume spectral enlarge the scope of MESMA.Main innovations of this paper are as follows:1) Introduce change detection algorithm of phenology BFAST to extract potential fire pixel.2) Add plume radiance spectral to reflected radiation endmember library, which enhanced MESMA scope and enriched the sub-pixel fire characteristics result.