Wildfire Smoke Detection And Fire-affected Vegetation Assessment Based On MODIS Data

With the development of global warming and the increasing frequency of extreme climates,wildfires have occurred frequently in recent years.The wildfire research is of critical importance for the protection of people's lives and property,and the understanding of fire impacts and ecosystem dynamics.Recently,a variety of environmental analysis applications have been advanced by the use of satellite remote sensing,which is an effective technology that can provide valuable information for wildfire monitoring.Numerous methods and analysis were investigated using the satellite imageries to detect wildfires and assess their effects.However,the effectiveness and applicability of wildfire identification need to be further improved,and more investigations are imperative for the spectral characteristics,mapping method,and vegetation dynamics of fire-affected areas.In this thesis,based on the satellite remote sensing data from the Moderate Resolution Imaging Spectroradiometer(MODIS)sensors onboard Aqua and Terra satellites,a couple of models were developed to improve the accuracy of wildfire smoke and burned area mapping,along with the in-depth analysis of the spectral features of post-fire areas and the fire-affected vegetation dynamics.The third chapter presents a new large-scale satellite imagery smoke detection benchmark generated from MODIS images,namely USTC_SmokeRS,consisting of 6225 satellite images from six classes(i.e.,cloud,dust,haze,land,seaside,and smoke)and covering various areas/regions over the world.To build a baseline for smoke detection in satellite imagery,we evaluate several state-of-the-art deep learning-based image classification models.Moreover,we propose a new convolution neural network(CNN)model,SmokeNet,which incorporates spatial and channel-wise attention in CNN to enhance feature representation for scene classification.The experimental results of our method using different proportions(16%,32%,48%,and 64%)of training images reveal that our model outperforms other approaches with higher overall accuracy and Kappa coefficient.Specifically,the proposed SmokeNet model trained with 64%training images achieves the best overall accuracy of 92.75%and Kappa coefficient of 0.9130.Besides,the model trained with 16%training images can also improve the classification overall accuracy and Kappa coefficient by at least 4.99%and 0.06,respectively,over the state-of-the-art models.In the fourth chapter,for the purpose of the accurate assessment of post-fire areas,we investigated the spectral characteristics of burned area and developed a new burned area mapping method that surpasses the detection accuracy of previous methods,while still using a single MODIS image.The key innovation is integrating the optimal spectral indices and a neural network algorithm.We used the traditional empirical formula method,multi-threshold method,and visual interpretation method to extract the sample sets of five typical types(burned area,vegetation,cloud,bare soil,and cloud shadow)from the MODIS data of several wildfires in the USA states of Nevada,Washington,and California in 2016.Afterward,the separability index M was adopted to assess the capacity of seven spectral bands and 13 spectral indices to distinguish the burned area from four unburned land cover types.Based on the separability analysis between the burned and unburned areas,the spectral indices with an M value higher than 1.0 were employed to generate the training sample sets that were assessed to have an overall accuracy of 98.68%and Kappa coefficient of 0.9746.Finally,we utilized a back-propagation neural network(BPNN)to learn the spectral differences of different types from the training sample sets and obtain the output burned area map.The proposed method was applied to three wildfire cases in the USA states of Idaho,Nevada,and Oregon in 2017.A comparison of detection results between the new MODIS-based burned area map and the reference burned area map compiled from Landsat-8 Operational Land Imager(OLI)data indicates that the proposed method can effectively exploit the spectral characteristics of various land cover types.Also,this new method can achieve higher accuracy with the reduction of commission error(CE,>10%)and omission error(OE,>6%)compared to the traditional empirical formula method.To understand the vegetation dynamic processes induced by wildfires,in the fifth chapter,we investigated the time dynamics of MODIS Aqua and Terra Normalized Difference Vegetation Index(NDVI)and Enhanced Vegetation Index(EVI)data acquired during nearly two decades from three study sites(L1,L2,and L3),of which site L1 was affected by two wildfires(Btu Fire,2008 and Camp Fire,2018),site L2 only affected by the Camp Fire,and site L3 not affected by any fire.The statistical analysis methods including the multifractal detrended fluctuation analysis(MFDFA)and the Fisher-Shannon analysis(FSA)were employed to investigate the heterogeneity and the organization/disorder of the vegetation time series,and the relationship with the recovery process of vegetation after fire.Our results indicate that multiple fires increase the heterogeneity of NDVI and EVI time series,whose time dynamics is governed by the large fluctuations.In addition,the NDVI and EVI data for the sites affected by wildfires are more organized and less disordered than those for the fire-unaffected sites.Furthermore,the three sites affected by different number of wildfires can be well discriminated by the multifractal exponents of the hq-range and the multifractal width,as well as the Fisher-Shannon Information plane,with the p-value for the comparisons generally lower than 0.05.The study develops the improved methods for fire smoke detection and burned area mapping,also provides deeper insights on the impacts of wildfires on vegetation dynamics and the changes of its characteristics.Additionally,the recommendations for further research are put forward.