【摘要】：土壤濕度是一類重要的地表環(huán)境變量,獲取高質(zhì)量的區(qū)域土壤濕度數(shù)據(jù)對于科學(xué)研究和生產(chǎn)實踐具有重要意義。傳統(tǒng)的土壤濕度數(shù)據(jù)是基于地面站點監(jiān)測,其優(yōu)勢是精度高、時間連續(xù)性好,但是空間連續(xù)性較差。為了獲取高精度、空間連續(xù)的土壤濕度數(shù)據(jù),本文以黑河中游盈科/大滿灌區(qū)域為實驗區(qū),利用地面觀測得到的多個空間點的土壤濕度時間序列數(shù)據(jù),對該區(qū)域的土壤濕度進行了空間分析,并對土壤濕度的空間插值方法和時間序列數(shù)據(jù)預(yù)測方法進行了研究。主要工作如下:1、對土壤濕度進行空間異質(zhì)性分析土壤濕度的空間異質(zhì)性分析主要包括統(tǒng)計分析、變異分析以及自相關(guān)分析三個方面。為了解本實驗區(qū)域的空間異質(zhì)特性,為接下來的實驗作為對比與參考,本文分別通過數(shù)值統(tǒng)計分析了解數(shù)據(jù)布局以及變異程度;以模型擬合度和殘差大小為檢驗指標(biāo),選取合適的變異模型,利用變異分析剖析區(qū)域化變量的空間結(jié)構(gòu)特征;計算土壤濕度的全局自相關(guān)Moran’s I系數(shù),計算結(jié)果通過Z值檢驗,檢驗具有空間位置的土壤濕度觀測點是否顯著地與其相鄰觀測點的觀測值相關(guān)聯(lián)。2、基于疊加理論,建立了考慮時空特征的土壤濕度插值方法依據(jù)信號疊加基本理論,信號可以分解為互不重合的部分,經(jīng)過處理后仍可以重構(gòu)得到原數(shù)據(jù)。以疊加理論作為理論基礎(chǔ),利用小波分解方法,對土壤濕度時序數(shù)據(jù)進行分解,得到低頻部分和殘差部分。分別利用高精度曲面建模(HASM)和變異函數(shù)對兩部分進行插值。將插值結(jié)果進行擬合,得到了精度較高的插值結(jié)果。3、建立了一種基于改進BP神經(jīng)網(wǎng)絡(luò)的土壤濕度時序預(yù)測方法針對普通BP神經(jīng)網(wǎng)絡(luò)收斂速度慢,易陷入局部最優(yōu)的問題,提出動量因子和自適應(yīng)學(xué)習(xí)率的BP神經(jīng)網(wǎng)絡(luò)改進方法,并且利用粒子群算法優(yōu)化BP神經(jīng)網(wǎng)絡(luò)的初始閾值和權(quán)值。針對標(biāo)準(zhǔn)粒子群算法(PSO)中慣性權(quán)重線性遞減、學(xué)習(xí)因子取常數(shù)而導(dǎo)致的PSO收斂速度慢、易錯過全局最優(yōu)解等問題,將迭代次數(shù)和適應(yīng)度值相結(jié)合改進慣性權(quán)重和學(xué)習(xí)因子,有效提高算法找到全局最優(yōu)解的速度。
[Abstract]:Soil moisture is a kind of important surface environmental variable. It is very important to obtain high quality regional soil moisture data for scientific research and production practice. The traditional soil moisture data is based on ground site monitoring, which has the advantages of high precision, good time continuity, but poor spatial continuity. In order to obtain soil moisture data with high accuracy and spatial continuity, the soil moisture time series data of several spatial points obtained from ground observation are used in this paper. The experimental area is Yingke / Daoman irrigation area in the middle reaches of Heihe River. The spatial analysis of soil moisture in this area was carried out, and the spatial interpolation method and time series data prediction method of soil moisture were studied. The main work is as follows: 1. The spatial heterogeneity analysis of soil moisture includes three aspects: statistical analysis, variation analysis and autocorrelation analysis. In order to understand the spatial heterogeneity of the experimental area, for the following experiments as a comparison and reference, this paper respectively through the numerical statistical analysis to understand the data layout and variability; model fitting and residual size as the test index, Selecting the appropriate variation model, using variation analysis to analyze the spatial structure characteristics of regionalized variables, calculating the global autocorrelation Moran's I coefficient of soil moisture, and calculating the results through Z test. To test whether the soil moisture observation point with spatial position is significantly related to the observed value of its adjacent observation points. Based on the superposition theory, a soil moisture interpolation method considering spatial and temporal characteristics is established according to the basic theory of signal superposition. The signal can be decomposed into non-coincidence parts, and the original data can be reconstructed after processing. Based on the superposition theory and wavelet decomposition method, the soil moisture time series data are decomposed to obtain the low frequency part and the residual part. (HASM) and variable function are used to interpolate the two parts respectively. By fitting the interpolation results, the interpolation results with high precision are obtained. A soil moisture time series prediction method based on improved BP neural network is established to solve the problem that the convergence rate of ordinary BP neural networks is slow and is prone to fall into local optimum. An improved method of BP neural network based on momentum factor and adaptive learning rate is proposed, and the initial threshold and weight of BP neural network are optimized by particle swarm optimization (PSO). Aiming at the problem of linear decline of inertia weight in standard particle swarm optimization (PSO), slow convergence speed of PSO caused by learning factor taking constant, and easy missing global optimal solution, the iterative times and fitness value are combined to improve inertia weight and learning factor. The speed of finding the global optimal solution is improved effectively.
【學(xué)位授予單位】：山東農(nóng)業(yè)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：S152.71

【參考文獻】

相關(guān)期刊論文 前10條

1 李雨鴻;孫倩倩;李輯;李晶;劉維;李琳琳;陶蘇林;;東北地區(qū)春播期土壤水分預(yù)測方法研究[J];干旱地區(qū)農(nóng)業(yè)研究;2015年06期

2 徐英;夏冰;;綜合BME和BNN法的農(nóng)田土壤水分與養(yǎng)分分布空間插值[J];農(nóng)業(yè)工程學(xué)報;2015年16期

3 萬曙靜;張承明;馬靖;;微波遙感反演地表土壤含水量的方法研究[J];山東農(nóng)業(yè)大學(xué)學(xué)報(自然科學(xué)版);2015年02期

4 楊勇;梅楊;張楚天;張若兮;廖祥森;;基于時空克里格的土壤重金屬時空建模與預(yù)測[J];農(nóng)業(yè)工程學(xué)報;2014年21期

5 孫倩倩;劉晶淼;梁宏;;東北地區(qū)土壤濕度的區(qū)域性預(yù)報模型研究[J];自然資源學(xué)報;2014年06期

6 趙明偉;岳天祥;趙娜;;高精度曲面建模優(yōu)化方案[J];中國圖象圖形學(xué)報;2014年02期

7 李小濤;黃詩峰;宋小寧;馬建威;;衛(wèi)星遙感結(jié)合地面觀測數(shù)據(jù)的土壤墑情監(jiān)測分析系統(tǒng)[J];水利學(xué)報;2013年S1期

8 王紅梅;王仲良;王X;陳歡;劉安樂;;華北農(nóng)牧交錯帶農(nóng)田-草地景觀鑲嵌體土壤水分空間異質(zhì)性[J];生態(tài)學(xué)報;2013年19期

9 肖緒培;宋乃平;謝騰騰;王興;楊明秀;;荒漠草原區(qū)不同土地利用類型土壤水分時空特征[J];生態(tài)與農(nóng)村環(huán)境學(xué)報;2013年04期

10 韓焱紅;矯梅燕;陳靜;陳法敬;;基于貝葉斯理論的集合降水概率預(yù)報方法研究[J];氣象;2013年01期

相關(guān)博士學(xué)位論文 前1條

1 胡紅英;局域波分解方法、特征剖析及應(yīng)用研究[D];大連理工大學(xué);2006年

相關(guān)碩士學(xué)位論文 前3條

1 周禹瑩;大慶市大同區(qū)土壤養(yǎng)分空間異質(zhì)性分析[D];哈爾濱師范大學(xué);2015年

2 張宇;不同載畜率下荒漠草原土壤水分空間異質(zhì)性的研究[D];內(nèi)蒙古農(nóng)業(yè)大學(xué);2014年

3 吳維臻;坡面尺度土壤水分空間異質(zhì)性特征及其與地形因子的關(guān)系[D];蘭州大學(xué);2014年

，

本文編號：2227511