Prediction model of moisture content of dead fine fuel in forest plantations on Maoer Mountain,Northeast China

Preventing and suppressing forest fires is one of the main tasks of forestry agencies to reduce resource loss and requires a thorough understanding of the importance of factors affecting their occurrence.This study was carried out in forest plantations on Maoer Mountain in order to develop models for predicting the moisture content of dead fine fuel using meteorological and soil variables.Models by Nelson(Can J For Res 14:597-600,1984)and Van Wagner and Pickett(Can For Service 33,1985)describing the equilibrium moisture content as a function of relative humidity and temperature were evaluated.A random forest and generalized additive models were built to select the most important meteorological variables affecting fuel moisture content.Nelson’s(Can J For Res 14:597-600,1984)model was accurate for Pinus koraiensis,Pinus sylvestris,Larix gmelinii and mixed Larix gmelinii—Ulmus propinqua fuels.The random forest model showed that temperature and relative humidity were the most important factors affecting fuel moisture content.The generalized additive regression model showed that temperature,relative humidity and rain were the main drivers affecting fuel moisture content.In addition to the combined effects of temperature,rainfall and relative humidity,solar radiation or wind speed were also significant on some sites.In P.koraiensis and P.sylvestris plantations,where soil parameters were measured,rain,soil moisture and temperature were the main factors of fuel moisture content.The accuracies of the random forest model and generalized additive model were similar,however,the random forest model was more accurate but underestimated the effect of rain on fuel moisture.

Diurnal variation models for fine fuel moisture content in boreal forests in China

Studying diurnal variation in the moisture content of fine forest fuel(FFMC)is key to understanding forest fire prevention.This study established models for predicting the diurnal mean,maximum,and minimum FFMC in a boreal forest in China using the relationship between FFMC and meteorological variables.A spline interpolation function is proposed for describing diurnal variations in FFMC.After 1 day with a 1 h field measurement data testing,the results indicate that the accuracy of the sunny slope model was 100%and 84%when the absolute error was<3%and<10%,respectively,whereas the accuracy of the shady slope model was 72%and 76%when the absolute error was<3%and<10%,respectively.The results show that sunny slope and shady slope models can predict and describe diurnal variations in fine fuel moisture content,and provide a basis for forest fire danger prediction in boreal forest ecosystems in China.

Spectral feature characterization and nitrogen content prediction in soils with different particle sizes and moisture contents

The objective of this research is to analyze the influences of light source incidence angle,fiber height,moisture content,and particle size on loamy mixed soil spectra.Nitrogen(N)content calibration and cross-validation models at different moisture contents and particle sizes were obtained using partial least squares(PLS)analysis.Spectral data were collected using a spectrophotometer.Fiber height of 100 mm and light source angle at 45°were chosen to obtain the sharpest spectra without apparent scattering effect.The results show that moisture content and particle size strongly influenced the absorbance of the spectra,and a better N prediction model was obtained when the particle sizes were in the ranges of 0.5-1.0,1.0-2.0 and 2.0-5.0 mm,with the correlation coefficients(r)of 0.819,0.815 and 0.818,and standard errors of prediction(SEP)of 2.29,2.41 and 2.42 mg/kg,respectively.Poor N prediction model was obtained when the soil was kept in its natural moisture content with r of 0.575 and SEP of 3.275 mg/kg,compared to the performance of dried soil samples with r of 0.815 and SEP of 2.425 mg/kg.

基于EMD-GRU的港口堆场煤炭含水率智能预测与实验研究

针对煤炭港口堆场洒水抑尘需求,提出了基于EMD-GRU的煤炭含水率预测模型并进行了实验验证。通过建立煤炭堆场含水率预测模型,利用实时数据驱动预测煤炭堆垛含水率变化,根据气象数据与含水率变化情况判断煤炭堆垛未来起尘情况并制定相应的洒水策略。实验结果表明,EMD-GRU模型的均方根误差、平均绝对误差、平均绝对百分比误差和决定系数分别为0.768、0.566、9.52%、0.944,与SVR、DTR、RNN、LSTM、GRU等预测模型相比,EMD-GRU预测模型的各误差值均最低,决定系数为最高,且预测精度与拟合效果最好。

赣南马尾松林地表细小死可燃物含水率动态及模型

[目的]建立森林地表细小死可燃物(枯落叶、细枯枝、枯草等)含水率预测模型,预警区域森林火灾引燃的可能性及其潜在火行为。[方法]基于野外长期定位观测的赣南地区典型植被类型马尾松林地表细小死可燃物含水率数据,在不同地形条件和时间段进行气象因子随机森林相对重要性排序和皮尔逊相关性分析,建立地表细小死可燃物含水率随机森林模型和气象要素回归模型,比较不同模型精度指标,筛选适合赣南地区的森林火灾预测模型。[结果]赣南地区马尾松林地表细小死可燃物含水率具有明显变异性,阴坡含水率显著高于阳坡,在防火期初期最明显。地表细小死可燃物含水率与各气象要素(温度、相对湿度、风速、光照强度)具有极显著相关性(P<0.001);随机森林模型预测精度高于气象要素回归模型,阴坡2种模型精度均高于阳坡;具有滞后效应的光照强度因子对地表细小死可燃物含水率影响最大,影响地表细小死可燃物含水率的关键因素在阳坡是相对湿度、阴坡是风速。[结论]具有滞后效应的气象因子对赣南地区马尾松林地表细小死可燃物含水率有显著影响,考虑增加这些因素能更好预测地表细小死可燃物含水率变化,为火险预警提供可靠依据。

密度与含水率双因素作用下重组竹顺纹弹性模量的预测与分析

本研究选取3组不同密度(1.10、1.20、1.30 g/cm^(3))和5组不同含水率(4%,8%,12%,16%,20%)的重组竹试件进行顺纹抗拉、抗压试验,分析不同密度和含水率作用下重组竹顺纹抗拉、抗压弹性模量的变化规律。结果发现,重组竹顺纹抗拉、抗压弹性模量随密度的增大而增大,随含水率的升高而减小并趋于平衡;通过试验研究和数据拟合提出重组竹顺纹抗拉、抗压弹性模量的双参数耦合模型,线性组合模型和ASTM组合模型均能较好地预测重组竹顺纹抗拉、抗压弹性模量随密度和含水率变化后的理论值,顺纹抗拉弹性模量的实测值与线性组合模型和ASTM组合模型预测值之间最大误差为5.28%和8.20%;顺纹抗压弹性模量的实测值与线性组合模型和ASTM组合模型预测值之间最大误差为8.91%和8.67%,因线性组合模型计算更为简便,建议在实际应用中优先选用。

基于PCA-BP神经网络的烟叶含水率预测研究

为了实现对复烤下机烟叶含水率的准确预测,提出了基于主成分分析法和BP神经网络的烟叶含水率预测模型。首先,采用主成分分析法提取最具表征意义的复烤烟叶含水率特征因子,获得特征矩阵。然后将特征矩阵输入BP神经网络,构建包括特征矩阵与复烤下机烟叶含水率的预测模型。仿真结果表明,提出的模型在复烤烟叶含水率预测方面呈现出显著的预测能力,决定系数达0.92。文中方法可辅助优化烟叶复烤控制参数,提升复烤烟叶品质。

人工神经网络在果蔬干燥领域应用进展

果蔬干燥是农产品加工中的重要环节,构建精确的干燥动力学模型成为干燥领域的重点方向。综述人工神经网络在果蔬干燥过程中的应用现状、分析存在的问题和做出展望。针对神经网络在干燥过程中的各种场景分类为四个部分:含水率预测、品质检测、工艺优化和控制系统方面,总结各部分的应用类型及发展创新;再对比传统干燥模型和人工神经网络模型;最后介绍混合神经网络的应用场景。发现人工神经网络比传统干燥模型更精确,且混合神经网络结合专家系统、模糊逻辑等理论能够提供准确的预测,作为一种新颖高效的建模技术,可以广泛应用于果蔬加工的优化、控制、自动化等领域。其中应用最广泛的就是与遗传算法结合的GA-BP神经网络,BP负责预测、GA负责寻优,在这样的算法中不仅可以精确预测结果还可以优化工艺。这样的模型更适合果蔬干燥且在未来有更广阔的发展空间,以期这些探讨和分析对果蔬干燥领域具有参考意义。

内蒙古大兴安岭森林火险等级预报模型研究

为提高森林火险等级预报的准确率和及时率,减少森林资源损失。以内蒙古大兴安岭地区为研究区域,利用遥感MODIS-NDVI数据反演可燃物的湿度指数,以气象指数、植被指数、闪电指数、湿度指数共同计算得出森林火险指数,以火险指数作为内蒙古大兴安岭地区森林火险等级预报模型的量化指标,并对火险等级进行分级,获得内蒙古大兴安岭森林火险等级预报方法。该方法结合当地的实际情况,将闪电指数引入到预报模型中,并实现了定量化估测。同时,以近几年发生在该地区重特大森林火灾为例,对该预报模型进行了验证。该方法可较好地对内蒙古大兴安岭地区森林火险等级进行定量化预报。

Establishment of soil moisture model based on hyperspectral data and growth parameters of winter wheat

Large area of soil moisture status diagnosis based on plant canopy spectral data remains one of the hot spots of agricultural irrigation.However,the existing soil water prediction model constructed by the spectral parameters without considering the plant growth process will inevitably increase the prediction errors.This study carried out research on the correlations among spectral parameters of the canopy of winter wheat,crop growth process,and soil water content,and finally constructed the soil water content prediction model with the growth days parameter.The results showed that the plant water content of winter wheat tended to decrease during the whole growth period.The plant water content had the best correlations with the soil water content of the 0-50 cm soil layer.At different growth stages,even if the soil water content was the same,the plant water content and characteristic spectral reflectance were also different.Therefore,the crop growing days parameter was added to the model established by the relationships between characteristic spectral parameters and soil water content to increase the prediction accuracy.It is found that the determination coefficient(R^(2))of the models built during the whole growth period was greatly increased,ranging from 0.54 to 0.60.Then,the model built by OSAVI(Optimized Soil Adjusted Vegetation Index)and Rg/Rr,two of the highest precision characteristic spectral parameters,were selected for model validation.The correlation between OSAVI and soil water content,Rg/Rr,and soil water content were still significant(p<0.05).The R^(2),MAE,and RMSE validation models were 0.53 and 0.58,3.19 and 2.97,4.76 and 4.41,respectively,which was accurate enough to be applied in a large-area field.Furthermore,the upper and lower irrigation limit of OSAVI and Rg/Rr were put forward.The research results could guide the agricultural production of winter wheat in northern China.

Predicting sandy soil moisture content with hyperspectral imaging

In this study,a rapid and non-invasive technology for predicting soil moisture content(SMC)was presented based on hyperspectral imaging(HSI).Firstly,a set of HSI system was developed to collect both spectral(400-1000 nm)and spatial(1620×841 pixels)information from sandy soil samples with variable SMC levels in the laboratory.Principal component analysis(PCA)transformation,K-means clustering,and several other image processing methods were performed to obtain a region of interest(ROI)of soil sample from the original HSI data.Then,256 optimal spectral wavelengths were selected from the average reflectance of the ROI,and 28 textural features were extracted using a gray-level co-occurrence matrix(GLCM).Data dimensionality reduction was conducted on both the spectral information and textural information by using a partial least square algorithm.Six latent variables(LVs)extracted from the spectral information,four LVs extracted from the textural information and fused data were used to build regression models with a three-layer BPNN,respectively.The results showed that all of the three calibration models achieved high prediction accuracy,particularly when using spectral information with R^(2)_(C)=0.9532 and RMSEC=0.0086.However,validation models demonstrate that predicting SMC using fused data is more effective than using spectral reflectance and textural features separately,with a R^(2)_(P)=0.9350 and RMSEP=0.0141,thus proving that the HSI technique is capable of detecting SMC.

Resistivity is used as a tool to evaluate the variability of soil water content

Resistivity is used to evaluate soil water content(SWC),which has the advantages of not causing soil disturbance and in low price.It is an effective way to assess the SWC variability.This paper aims to evaluate the variability of loess slope SWC through the change of resistivity.It provides a simple way for long term SWC monitoring to solve the expensive cost of deploying moisture sensors.In this context,geoelectric and environmental factors such as soil temperature and SWC were monitored for three years.The prediction model of apparent resistivity and SWC was calibrated.The post processing of geoelectric data was introduced.In addition,the SWC collected by Time-Domain Reflectometry(TDR)was used to verify the feasibility of electrical resistivity tomography(ERT)data.The SWC variability in the process of rainfall,the evolution of four seasons,and the alternation of drying and wetting were evaluated.The research results show that:i)the SWC monitored by ERT and TDR can reflect the response and hysteretic effect of water content at 0.5-3.0 m depth.ii)The moisture content monitored by ERT reflects that the soil is relatively wet in summer and autumn and dry in winter and spring.iii)From 2017 to 2020,the SWC increased in August,and the soil became dry in January.iv)Two areas with high SWC and three areas with low SWC on loess slope are reflected by resistivity.The outcome can provide the change information of SWC to a great extent without excavating boreholes.

密度和含水率对竹基纤维复合材料抗弯性能的影响

通过试验探究不同密度和含水率对竹基纤维复合材料的静曲强度(modulus of rupture,MOR)以及弹性模量(modulus of elasticity,MOE)的影响,并建立密度和含水率对MOR以及MOE影响的预测模型。结果表明,在测试区间内试板的密度与抗弯性能呈正相关;随着含水率升高,试板抗弯性能先增大后减小。根据预测模型,压制密度1.17 g/cm^(3)、含水率11%左右的板材可以获得较高的MOR,大约为151.32 MPa;压制密度1.20 g/cm^(3)、含水率10.65%左右的板材可以获得较高的MOE,大约为19.68 GPa。该研究为提高竹基纤维复合材料的抗弯性能以及生产实践提供一定理论指导。

哈尔滨典型林分地表凋落物含水率预测模型

为构建高精度的森林地表凋落物含水率预测模型,以黑龙江省哈尔滨市3种典型林分白桦林(Betula platyphylla)、蒙古栎林(Quercus mongolica)和兴安落叶松林(Larix gmelini)为研究对象,采用自制的便携式气象因子检测仪进行实时检测,通过随机森林(Random Forest,RF)、反向传播(Back Propagation,BP)和长短期记忆网络(Long Short Term Memory,LSTM)3种代表性的机器学习方式对3种林分分别构建凋落物含水率预测模型。研究表明,不同的林分在不同的机器学习方式下构建的预测模型的预测精度均不同;在不同的机器学习构建的模型中3种林分的预测精度均为LSTM>BP>RF;LSTM在地表凋落物含水率预测模型的研究中效果最优。此研究结果说明LSTM在地表凋落物含水率的预测中表现突出,对提高基于气象要素法预测地表凋落物含水率模型精度具有重要的理论意义和实践指导价值。

基于BP神经网络的5HNH-15干燥机出粮水分研究

精确的工业化粮食干燥过程数学模型是实现其过程动态跟踪、闭环控制的前提。为此,基于5HNH-15连续式粮食干燥机,构建了8-11-1的BP神经网络预测模型,模型的输入为5HNH-15连续式干燥机的8个干燥影响因素,输出为出口粮食含水率。利用MmatLab软件进行BP神经网络模型的建立及验证,结果表明:模型在67次迭代后,均方误差MSE达到2.8361e-6,绝对误差小于±0.1,平均绝对误差MAE=0.0288,相对误差小于1.2%,回归系数R=0.99996,决定系数R 2=0.9998。新增1组验证试验,结果显示:模型预测值与实际值的绝对误差小于±0.1,平均绝对误差MAE=0.0121,相对误差小于1.1%,证明了所构建模型的精确性与普适性,可为实现工业化粮食干燥的智能控制提供理论依据和技术支撑。

基于3S的小开河灌区农业灌溉需水量预测研究

土壤含水率的空间分布是影响农业灌溉决策的关键因素。采用3S技术对灌区的土壤含水率空间分布进行探索,以土壤含水率为预报对象,结合作物生长情况和气象预报,以田间水量平衡方程为理论基础,利用土壤含水率预测灌区农业灌溉需水量。结果表明:小开河灌区5月下旬的预测需水量为621.77万m^(3),与灌区的统计数据一致,基于3S的灌区农业灌溉需水量预测技术可实现灌区的精准灌溉。

基于原位光谱的滨海滩涂土壤含水量预测模型

【目的】滩涂土壤含水量对滨海区域生态保护和植物生长都至关重要,及时了解土壤含水量对生态恢复、土壤资源管理和土壤可持续利用具有重要意义。针对土壤含水量获取复杂的问题,探求更加快速精准地获取滨海滩涂土壤含水量的方法。【方法】沿沧州滨海区域采集了共计14个表层(0~20 cm)滩涂土壤样品,在实验室利用烘干法测得滩涂土壤含水量。在分析野外实测滨海滩涂土壤原位高光谱反射率(R)和土壤含水量特性的基础上,采用反射率倒数(1/R)、反射率倒数对数(log(1/R))、反射率一阶微分(R')和反射率去包络线(CR)等变换形式。充分挖掘光谱信息,并探求与滨海滩涂土壤高相关性的可见近红外(VIS-NIR)光谱波段,构建基于多元逐步线性回归(MSR)和支持向量机回归(SVR)方法的滨海滩涂土壤含水量预测模型。结合适用于小样本的留一交叉验证(LOO-CV)法验证模型精度,并对比分析两种预测模型的性能以及模型的稳定性。【结果】研究发现:1)土壤光谱反射率与对应土壤含水量呈显著负相关,且二者在1400~1600 nm和1900~2400 nm内密切相关。2)除R外,同一种光谱变换形式下,采用SVR方法构建的滨海滩涂土壤含水量预测模型精度和稳定性明显高于MSR方法,对比得出基于R´的滨海滩涂土壤含水量SVR预测模型精度最高,Adjusted-R2、RPD和RMSE分别为0.81、2.08和2.56。【结论】在R'变换形式下利用SVR方法建立的模型能够较准确地预测滨海滩涂土壤含水量,为滨海湿地土壤管理、植物生长和环境保护提供必要的数据支持,并为基于高光谱影像的区域尺度土壤含水量预测提供方法借鉴。

植被含水率对导线触树单相接地故障特性的影响分析

输配电线路走廊的导线触树单相接地故障可能引燃植被、诱发火灾,威胁线路的正常运行。植被含水率是影响导线触树单相接地故障的重要因素,但含水率对故障发展过程、故障特性的影响尚不明确。以典型植被桉树为对象,开展了导线触树单相接地故障试验研究。基于试验现象及泄漏电流波形,从炭化通道发展及含水率变化的角度分析了植被高阻接地故障发展过程,研究了含水率对故障过程中泄漏电流、炭化通道、等效电阻等特性的影响规律。随后,揭示了泄漏电流第一峰值与含水率关系,建立了基于植被长度、直径、含水率的泄漏电流第一峰值预测模型,通过试验验证了预测模型的准确性。该研究成果对于导线触树单相接地故障的检测、诊断及预测具有重要意义。

基于LSTM网络的粮食干燥机水分预测与优化

粮食干燥机的出机粮食水分预测有助于实现干燥机的智能化控制,从而可以减少干燥过程中的粮食损耗,对于粮食产后干燥环节有着重大意义。通过机器学习的方式进行预测,可以规避传统数学模型所存在的一系列缺陷。研究根据连续式谷物干燥机所提取的数据特征,提出了一种基于优化长短期记忆神经网络(LSTM)的稻谷出机水分预测模型。实验结果表明,出机水分与M_(in)、T_(o2)、T_(o3)、T_(d1)、T_(d2)、T_(d3)具有十分明显的相关性,通过设定不同的网络参数,确立了批尺寸50,学习率0.001,迭代次数50,时间步长50,神经元数100×100时效果最佳,此外还发现增加训练数据量,可以有效提高LSTM网络预测性能。将研究建立的LSTM模型与BP、ELMAN、NARX等算法以及普通LSTM网络(无dropout,单隐藏层)进行比较。结果发现,相较于其他网络模型,研究所采用的LSTM模型可以更好地预测稻谷出机水分,其平均绝对误差(MAE)、均方根误差(RMSE)和决定系数(R^(2))分别为0.12%、0.20%和0.94。研究所采用的优化LSTM模型具有较高预测精度,稳定性以及泛化性,可以为粮食干燥机的水分预测控制提供参考。

我国北方森林地下火燃烧温度预测

森林地下火燃烧是一个隐蔽、漫长的阴燃过程,整个燃烧过程都靠自身释放的热量维持,所以开展地下火燃烧温度预测研究,对于地下火监测预警具有重要意义.以大兴安岭地区呼中国家级自然保护区典型植被兴安落叶松林为研究对象,通过室内模拟点烧试验,基于非线性混合效应模型预测不同含水率下地下火燃烧温度.结果表明:基于Logistic和Richards回归的非线性混合效应模型适合地下火燃烧温度预测;3参数的混合效应模型拟合效果最好;不同含水率下不同深度地下火燃烧温度预测模型的预测值和真实值间都不存在显著差异(P>0.05),模型预测精度较高,尤其是在3 cm深度处,平均相对误差(MRE)都小于15%,具有较高的应用价值.