基于土壤水分和气象要素的林火预报研究——以广东省为例

基于国内现行的森林火险气象指数和单因子火险贡献度模型,以及逻辑回归模型和随机森林模型,在林火预报中引入微波遥感土壤水分信息,使用MCD14DL火点数据集和地面气象观测资料对广东省不同时间尺度的林火发生概率进行预测。结果表明：逻辑回归模型和随机森林模型构建的林火预测模型显著优于现行的森林火险气象指数和单因子火险贡献度模型,预测精度提升约20%。其中,随机森林模型对林火频数的解释程度最高（两者相关系数为0.476）。此外,加入微波土壤水分信息后,相较原有的基于气象要素的林火预测模型,2种机器学习模型的预测精度均略有提升,体现了表层土壤水分信息在林火预报中的重要性。研究可为高效提取对地观测信息,以改进华南地区不同时间尺度的林火预报工作提供参考。     

气象自动站资料在黔南森林火险等级监测预报中的应用

对大量国家级和区域级气象自动站资料的获取和订正,通过对降水、气温、湿度和风速等气象因子对森林火险贡献度数学模型的计算,充分应用在黔南森林火险气象等级的实时监测和预报中。通过多点、多时段的预报值、实时监测值与实况对比检验统计,得出应用系统的实时监测准确率和预报准确率。     

Spatial Ecology of Pre –Euro‐American Fires in a Southern Rocky Mountain Subalpine Forest Landscape

Natural disturbances such as fires have been widely studied, but less is known about their spatial ecology than about other aspects of them. We reconstructed and mapped pre–Euro‐American fire history in a subalpine forest landscape in southeastern Wyoming, and analyzed the fires using GIS. Mean fire interval varies little with topography (elevation, aspect, slope) and is spatially autocorrelated at distances of at least 2 km. Fires often spread downslope, and spread more than expected from the north and south and less than expected from the west, under the influence of particular synoptic climatic conditions. The landscape of 1868 a.d., at the time of Euro‐American settlement, was strongly influenced by fires. However, it contained large patches of connected forest and few high‐contrast edges, unlike the modern landscape, which is fragmented by industrial forestry and roads. The spatial ecology of the natural fire regime may be a useful guide for management.     

A geo-information system approach for forest fire likelihood based on causative and anti-causative factors

Innumerable forest fire spread models exist for taking a decision, but far less focus is on the real causative factors which initiate/ignite fire in an area. It has been observed that the majority of the forest fires in India are initiated due to anthropogenic factors. In this study, we develop a geo-information system approach for management of forest fire in Mudumalai Wildlife Sanctuary, Tamil Nadu, India, with the objective to develop a forest fire likelihood model, integrating GIS and knowledge-based approach for predicting fire-sensitive initiation areas considering major causative and anti-causative factors. Amongst the various causative factors investigated, it was found that wildlife-dependent factor (antler collection and poaching) contributed significantly to fire occurrence followed by management-dependent factors (uncontrolled tourism and grazing), with much less influence of demographic factors. Similarly, anti-causative factor (stationing of anti-poaching/ fire camps) was considered as quite significant.

The likelihood model so developed, envisaging various factors and flammability, accounted for different scenarios as a result of pair-wise comparison on an ordinal scale in a knowledge matrix. The inferential statistics computed indicated the robustness of the model and its insensitivity to moderate changes. It makes it possible for this forest fire likelihood model to predict and prevent a forest fire in an effective and scientific manner because it can assume forest fire likelihood in real time and present in proper time.     

Fire In The Management Of Vegetation

Abstract

Fire, either natural or man-caused, has influenced the pattern of vegetation in numerous areas over the earth. Factors that contribute to potential damage to either the vegetation and/or environment are examined. Since the extent of damage varies, a fire can be destructive or useful. The use of “prescribed” fire in land management accomplishes several desired objectives, one of which is the perpetuation of certain fire-climax vegetation types, such as pines and some grasses. In those pine-forested areas where fire is regularly used in management, a related benefit is the reduction of hazardous natural fuels, which results in fewer catastrophic wildfires. Air pollution due to forest burning is also considered.     

Porting of a wildfire risk and fire spread application into a cloud computing environment

Effective wildfire management is an essential part of forest firefighting strategies to minimize damage to land resources and loss of human lives. Wildfire management tools often require a large number of computing resources at a specific time. Such computing resources are not affordable to local fire agencies because of the extreme upfront costs on hardware and software. The emerging cloud computing technology can be a cost- and result-effective alternative. The purpose of this paper is to present the development and the implementation of a state-of-the-art application running in cloud computing, composed of a wildfire risk and a wildfire spread simulation service. The two above applications are delivered within a web-based interactive platform to the fire management agencies as Software as a Service (SaaS). The wildfire risk service calculates and provides daily to the end-user maps of the hourly forecasted fire risk for the next 112 hours in high spatiotemporal resolution, based on forecasted meteorological data. In addition, actual fire risk is calculated hourly, based on meteorological conditions provided by remote automatic weather stations. Regarding the wildfire behavior simulation service, end users can simulate the fire spread by simply providing the ignition point and the projected duration of the fire, based on the HFire algorithm. The efficiency of the proposed solution is based on the flexibility to scale up or down the number of computing nodes needed for the requested processing. In this context, end users will be charged only for their consumed processing time and only during the actual wildfire confrontation period. The system utilizes both commercial and open source cloud resources. The current prototype is applied in the study area of Lesvos Island, Greece, but its flexibility enables expansion in different geographical areas.     

遥感、地图、地理信息系统—体化应用

遥感、地图和地理信息系统（ＯＩＳ）三者呈“你中有我，我中有你”的相辅相成关系．三者一体化应用使地球科学得以进展，又能在资源开发、环境保护、自然灾害监测评价等方面发挥重要作用．一体化应用的基础是掌握三者的学科一技术特性与相通关系．     

基于GIS和视频监控的森林防火指挥决策系统研建

森林火灾能对森林资源造成巨大的危害,因此发现林区火源并进行科学决策具有重要意义.本系统数据主要包括1∶1万地形图、1∶1万森林资源二类调查数据、扑火队分布数据、瞭望台站数据、扑火物资储备数据、林业局管理机构分布数据、研究区2010年landsat TM遥感影像图等;系统功能主要包括数据的处理和管理、森林火灾监测、最佳扑火路径分析、防火扑火预案制定和决策、森林火险等级评价、林火趋势模拟和预测、灾害损失评估、相关图表和报告制作等.系统采用Microsoft Visual Studio 2005(C#)作为开发语言,基于Arc Engine 9.3组件进行二次开发,三维地图显示采用skyline Terra Explorer开发平台,数据库采用SQL Server 2005数据库管理系统,利用Arc SDE 9.3空间数据库引擎对空间数据进行管理.     

Representing and reducing error in natural‐resource classification using model combination

Artificial Intelligence (AI) models such as Artificial Neural Networks (ANNs), Decision Trees and Dempster–Shafer's Theory of Evidence have long claimed to be more error‐tolerant than conventional statistical models, but the way error is propagated through these models is unclear. Two sources of error have been identified in this study: sampling error and attribute error. The results show that these errors propagate differently through the three AI models. The Decision Tree was the most affected by error, the Artificial Neural Network was less affected by error, and the Theory of Evidence model was not affected by the errors at all. The study indicates that AI models have very different modes of handling errors. In this case, the machine‐learning models, including ANNs and Decision Trees, are more sensitive to input errors. Dempster–Shafer's Theory of Evidence has demonstrated better potential in dealing with input errors when multisource data sets are involved. The study suggests a strategy of combining AI models to improve classification accuracy. Several combination approaches have been applied, based on a ‘majority voting system’, a simple average, Dempster–Shafer's Theory of Evidence, and fuzzy‐set theory. These approaches all increased classification accuracy to some extent. Two of them also demonstrated good performance in handling input errors. Second‐stage combination approaches which use statistical evaluation of the initial combinations are able to further improve classification results. One of these second‐stage combination approaches increased the overall classification accuracy on forest types to 54% from the original 46.5% of the Decision Tree model, and its visual appearance is also much closer to the ground data. By combining models, it becomes possible to calculate quantitative confidence measurements for the classification results, which can then serve as a better error representation. Final classification products include not only the predicted hard classes for individual cells, but also estimates of the probability and the confidence measurements of the prediction.     

Vegetative succession and controlled fire in a glades ecosystem A geographical information system approach

Abstract

A geographical information system (GIS) approach was used successfully on a federal wilderness area in southwestern Missouri to examine vegetative succession relative to fire management in a glades ecosystem. Maps of vegetation were obtained by interpreting aerial photographs taken in 1938, 1958, 1966, 1975 and 1986. Maps of topography, streams, soils and the location of fires which burned during the period 1938 to 1986 were also procured from a variety of public agencies. All maps were digitized and incorporated into a raster-based GIS with 25 m pixels. It was concluded that (1) both glades and oak-hickory forest have an affinity for distinct physiographic areas and (2) fire can help decrease the rate of processes of natural succession which cause glades to convert to forest. The probable effects of controlled fire on three areas proposed by the Forest Service were evaluated and summarized.     

Impacts of land use and fire on the loss and degradation of lowland forest in 1983–2000 in East Kutai District, East Kalimantan, Indonesia

Deforestation and forest degradation are proceeding rapidly in the lowland forests of Indonesian Borneo. Time series analysis of satellite imagery provides an ideal means of quantifying landscape change and identifying the pathways which lead to the changes. This study investigates the forest and land cover changes by classifying Landsat MSS (Multispectral Scanner), TM (Thematic Mapper) and ETM+ (Enhanced Thematic Mapper Plus) images over three time periods (1983–90, 1990–98, and 1998–2000), creating land cover maps for each year and change trajectories for each year-pair. The study area chosen covers an area of 2160 km² of undulating topography and alluvial plains in the East Kutai District of East Kalimantan Province, which in the 1980s was covered mostly with lowland dipterocarp forest; today the landscape is a patchwork dominated by oil palm and timber plantations and degraded forest. We relate land cover change data to land use allocation and to fire impacts based on fire hotspot distribution and fire damage information. The multidate land cover change trajectories provide an insight into the forest loss and degradation pathways over the 17-year period spanning the first entry of commercial logging concessionaires, followed by a government-sponsored transmigration scheme, government-licensed timber and oil palm plantations and, finally, the devastating fires of 1998. The results show a mean deforestation rate of 42 km² or 6 per cent per year for 1983–2000, rising to 10 per cent per year for 1990–98; by 2000, 70 per cent of forest initially damaged by fire and drought during the 1982–83 El Niño event was classified as non-forest. Although our study area is perhaps a worst-case scenario in terms of land use planning outcomes, the lessons from this research are directly applicable to scenario prediction for informed forest and land use planning and monitoring.     

Research Article. Modelling inside GIS: Part 1. Model structures,exploratory spatial data analysis and aggregation

Abstract

This is the first of two papers elaborating a framework for embedding urban models within GIS. This framework is based upon using the display capabilities of GIS as the user interface to the conventional modelling process, beginning with data selection and analysis, moving to model specification and calibration, and thence to prediction. In this paper, we outline how various stages in this process based on purpose-built software outside the system, are accessed and operated through the GIS. We first deal with display based on thematic maps, surfaces, graphs and linked windows, standard to any data from whatever source, be it observations, model estimates or predictions. We then describe how various datasets are selected, how the spatial system can be partitioned or aggregated, and how rudimentary exploratory spatial data analysis enables scatterplots to be associated with thematic maps. We illustrate all these functions and operations using the proprietary GIS ARC-INFO applied to population data at the tract level in the Buffalo region. In the second part of the paper, various residential location models are outlined and the full modelling framework is assembled and demonstrated.     

Modelling soil map-unit inclusions by Monte Carlo simulation

Abstract

Error and uncertainty in spatial databases have gained considerable attention in recent years. The concern is that, as in other computer applications and, indeed, all analyses, poor quality input data will yield even worse output. Various methods for analysis of uncertainty have been developed, but none has been shown to be directly applicable to an actual geographical information system application in the area of natural resources. In spatial data on natural resources in general, and in soils data in particular, a major cause of error is the inclusion of unmapped units within areas delineated on the map as uniform. In this paper, two alternative algorithms for simulating inclusions in categorical natural resource maps are detailed. Their usefulness is shown by a simplified Monte Carlo testing to evaluate the accuracy of agricultural land valuation using land use and the soil information. Using two test areas it is possible to show that errors of as much as 6 per cent may result in the process of land valuation, with simulated valuations both above and below the actual values. Thus, although an actual monetary cost of the error term is estimated here, it is not found to be large.     

土壤数据空间分辨率对水文过程模拟的影响

分布式水文模型的应用,其准确性有赖于输入数据对流域特征的描述,尤其在大尺度流域,输入数据分辨率的增加是否必然改善模型的模拟效果是值得深入研究的问题。本文以鄱阳湖信江流域为研究区,运用SWAT模型为模拟工具,分析了土壤数据空间分辨率对径流、蒸发及土壤含水量等水文要素模拟的影响以及高精度土壤数据在大流域尺度的适应性。结果表明：不同分辨率的土壤数据对SWAT模型中水文响应单元的划分结果差异显著,但在径流模拟和蒸发计算结果中并没有表现出显著的差别;模型率定前后,低分辨率土壤数据的径流模拟结果略好于高分辨率土壤数据,但两者之间的差别不明显;模型模拟的土壤含水量差异显著,高分辨率土壤模拟的月平均土壤含水量整体大于低分辨率土壤模拟结果;研究还发现,模型的蒸发计算对土壤分辨率信息不敏感。本文研究意味着,大尺度SWAT模型的应用中,土壤数据分辨率的提高不一定会改善模型的模拟效果。在具体应用中,应考虑流域本身的尺度以及模拟精度的要求,选择合适分辨率的土壤数据,同时应结合模型原理和关键参数的物理含义来解释模拟结果。     

Identifying strengths and limitations of pan-European forest cover maps through spatial comparison

Detailed and harmonized information on spatial forest distribution is an essential input for forest-related environmental assessments, in particular, for biomass and growing stock modeling. In the last years, several mapping approaches have been developed in order to provide such information for Europe in a harmonized way. Each of these maps exhibits particular properties and varies in accuracy. Yet, they are often used in parallel for different modeling purposes. A detailed spatial comparison seemed necessary in order to provide information on the advantages and limitations of each of these forest cover maps in order to facilitate their selection for modeling purposes.

This article confronts the high-resolution forest cover map recently developed by the Joint Research Centre for the year 2000 (FMAP2000) with previously existing maps for the same time period: the CORINE Land Cover 2000 (CLC2000) and the Calibrated European Forest Map 1996 (CEFM1996). The spatial comparison of these three maps was carried out based on forest proportion maps of 1 km derived from the original maps. To characterize differences according to biogeographic regions, two criteria were used: detail of thematic content within each map and local spatial agreement.

Concerning thematic content, CLC2000 displayed a surfeit of non-forested areas at the cost of low forest proportions, while FMAP2000 showed a more balanced distribution likely to preserve more detail in forest spatial pattern. Good spatial agreement was found for CLC2000 and FMAP2000 within about 70% of the study area, while only 50% agreement was found when compared with CEFM1996. The largest spatial differences between all maps were found in the Alpine and Mediterranean regions. Reasons for these might be different input data and classification techniques and, in particular, the calibration of CEFM1996 to reported national statistics.     

Building Environmental Literacy through Participation in GIS and Multimedia Assisted Field Research

Abstract

A Learning Support System (LSS) that emphasizes experiential research in natural environments using the cutting-edge technologies of GIS and multimedia has been developed for teaching environmental literacy to undergraduate students at the University of Georgia. Computers are used as cognitive tools to create a context in which students become interns in an ecological research center. Students are instructed to conduct research in the form of two field laboratories (the stream and forest laboratories). They accomplish their tasks by collecting data in the field (the State Botanical Garden of Georgia near the campus). They enter the data in the Learning Support System (LSS), and are guided to formulate hypotheses relating to stream water quality and human impact on forest succession for testing. Students also interact with the Environmental Research Support Site (ERSS) within the LSS for explanations to their findings. A specially customized Arc View GIS program within the LSS provides a tool to students for spatial analysis in the case of the forest laboratory. Students and faculty evaluations as well as final examination results confirmed the receptiveness of students to the LSS approach and its effectiveness in the learning of environmental literacy.     

A hybrid integrated deep learning model for the prediction of citywide spatio-temporal flow volumes

ABSTRACT

The spatio-temporal residual network (ST-ResNet) leverages the power of deep learning (DL) for predicting the volume of citywide spatio-temporal flows. However, this model, neglects the dynamic dependency of the input flows in the temporal dimension, which affects what spatio-temporal features may be captured in the result. This study introduces a long short-term memory (LSTM) neural network into the ST-ResNet to form a hybrid integrated-DL model to predict the volumes of citywide spatio-temporal flows (called HIDLST). The new model can dynamically learn the temporal dependency among flows via the feedback connection in the LSTM to improve accurate captures of spatio-temporal features in the flows. We test the HIDLST model by predicting the volumes of citywide taxi flows in Beijing, China. We tune the hyperparameters of the HIDLST model to optimize the prediction accuracy. A comparative study shows that the proposed model consistently outperforms ST-ResNet and several other typical DL-based models on prediction accuracy. Furthermore, we discuss the distribution of prediction errors and the contributions of the different spatio-temporal patterns.     

THUNDERSTORM CLIMATOLOGY BASED ON LIGHTNING DETECTOR DATA,MANITOBA, CANADA

Comparisons are made between thunderstorm data collected from a lightning detector network and from conventional climatic stations for the province of Manitoba, Canada. The greater resolution in time and space of lightning detector (direction finder) data makes it a valuable source of thunderstorm information and lends itself to some important applications. Data were collected for the forest fire season of 1985 using a network of 7 lightning direction finders distributed throughout the province. Some 67,912 cloud-to-ground lightning strikes were recorded by time and location during 122 thunderstorm days. July was the most active month with 27,260 strikes over 28 days. Two regions of the province had the greatest concentration of lightning strikes, indicating some influence by topography and position of large lakes. Case studies are presented of the most active lightning storms of 1985 and 1986. These storms are exclusively frontal storms, with most having similar synoptic weather patterns to those of large hailstorms and tornadoes in Manitoba. Relationships between meteorological parameters and lightning strike distribution are presented. These relationships may prove useful in the suppression of lightning-caused forest fires, especially in remote areas of the province. [Key words: lightning, thunderstorm, synoptic climatology, natural hazards, fire prevention.]     

汉江流域安康站日径流预测的LSTM模型初步研究

论文基于2003—2014年水文资料,采用长短期记忆神经网络(Long-Short Term Memory,LSTM),构建了汉江上游安康站日径流预测模型,评价了不同输入条件下日径流预测的精度。结果表明：当预见期为1 d时,在仅以安康站前期日径流量作为输入的条件下,LSTM模型在训练期和检验期的效率系数分别达到0.68和0.74;如再将流域前期面雨量和上游石泉站前期日径流量加入LSTM网络作为输入变量,安康站日径流量预测效果将更好,训练期和检验期的效率系数最高可达到0.83和0.84,均方根误差也有显著削减,且对主要洪峰流量的预测能力也有一定提高。此外,LSTM可以有效避免过拟合等问题,具有较好的泛化性能。但当预见期从1 d延长至2、3 d时,LSTM的预测精度显著降低。     

不同树种叶片养分含量提取的高光谱方法及精度评价

以深圳市坝光银叶园和大鹏半岛自然保护区19种湿地森林树种叶片可见光近红外光谱与全氮（Total Nitrogen, TN）、全磷（Total Phosphorus, TP）、全钾（Total Potassium, TK）含量关系为基础,分析了11种光谱预处理方式、3种光谱数据降维方式和2种建模方法对模型精度的影响。结果表明,标准正态变换（Standard Normal Variate, SNV）结合一阶导数（first derivative, 1 ^st）预处理方式下模型精度最高;主成分分析（Principal Component Analysis, PCA）降维处理对模型的降维效果最好;支持向量回归（Support Vector Regression, SVR）的建模效果精度最高。对于TN、TP、TK最佳模型的预测确定系数均在0.80以上,模型RPD值也在2.0以上,SVR模型可以用于树种叶片TN、TP、TK的快速检测。