本文選題：土壤養(yǎng)分 + 模擬退火��； 參考：《西南大學(xué)》2017年碩士論文

【摘要】：土壤養(yǎng)分的空間分布信息是對土壤資源進行研究的基礎(chǔ)信息,把握研究區(qū)內(nèi)土壤養(yǎng)分在地形上的變異特征對農(nóng)業(yè)生產(chǎn)和環(huán)境模擬具有至關(guān)重要的作用。在成土母質(zhì)同源,氣候條件、耕作方式、管理措施相同的條件下,地形是影響土壤養(yǎng)分空間分異的主要因素,對于地形復(fù)雜的丘陵區(qū)域,土壤樣本的獲取需要耗費大量成本,合理布設(shè)采樣點以確保土壤養(yǎng)分空間信息完整表達(dá)顯得尤為重要。有效合理的土壤采樣點布局不僅能夠充分反映土壤養(yǎng)分的空間信息,還能大大降低成本。本文以重慶市江津區(qū)永興鎮(zhèn)典型丘陵區(qū)(2km2)為研究區(qū)域,運用空間分析理論方法研究探討土壤pH值、有機質(zhì)、堿解氮、有效磷和速效鉀含量的空間變異規(guī)律及其與地形的關(guān)系,利用模擬退火方法確定最優(yōu)采樣密度和最佳取樣單元,并在GIS技術(shù)的支撐下,采用神經(jīng)網(wǎng)絡(luò)方法結(jié)合地形因子進行預(yù)測性土壤養(yǎng)分制圖。主要研究結(jié)果為:(1)土壤養(yǎng)分之間關(guān)系密切。土壤pH與有機質(zhì)、堿解氮、有效磷之間呈顯著負(fù)相關(guān),土壤有機質(zhì)與堿解氮、速效鉀之間呈顯著正相關(guān),土壤有效磷與速效鉀之間呈顯著正相關(guān)。土壤養(yǎng)分存在空間自相關(guān)性。土壤堿解氮和有機質(zhì)含量具有強烈的空間自相關(guān)性;土壤pH值、土壤有效磷和速效鉀含量具有中等程度的自相關(guān)性。土壤養(yǎng)分與地形因子存在顯著相關(guān)性。土壤pH值與地形濕度指數(shù)(TWI)呈顯著負(fù)相關(guān),與水平曲率(HORIZC)和坡度(Slope)呈顯著正相關(guān),即隨著土壤中水分含量的增加,土壤酸堿度降低;土壤有機質(zhì)、堿解氮含量與高程(Elevation)、水平曲率(HORIZC)、坡度(Slope)和相對位置指數(shù)(RPI)呈顯著負(fù)相關(guān),與地形濕度指數(shù)、坡長(SlpLen)和比匯水面積(SCA)呈顯著正相關(guān);土壤速效鉀含量與地形濕度指數(shù)和坡長呈顯著正相關(guān),與其他地形因子的相關(guān)性較弱;即地形越緩,土壤中有機質(zhì)、堿解氮和速效鉀的含量越高;土壤有效磷含量與地形因子的相關(guān)性較弱。(2)利用模擬退火算法結(jié)合神經(jīng)網(wǎng)絡(luò)模型對訓(xùn)練集中原始200個土壤樣點的空間分布進行了系統(tǒng)優(yōu)化,對五項土壤養(yǎng)分指標(biāo)都給出了最佳空間布局組合;同時,針對每一個樣點組合給出了與其對應(yīng)的預(yù)測誤差(均方誤差)。誤差結(jié)果表明:土壤pH、有機質(zhì)、堿解氮、有效磷、速效鉀分別最少可用5、6、7、6、5個優(yōu)化后的樣點代替原始樣點,且誤差不高于原始200個樣點的均方誤差。優(yōu)化后的土壤pH、有機質(zhì)、堿解氮、有效磷以及速效鉀的最優(yōu)樣點個數(shù)分別為68、118、87、86、60。(3)以地形因子輔助變量,分別構(gòu)建原始樣點和最佳布局條件下的土壤養(yǎng)分BP神經(jīng)網(wǎng)絡(luò)預(yù)測模型,結(jié)果表明:樣點減少后的預(yù)測模型對土壤養(yǎng)分的解釋能力上升,預(yù)測精度增加,模型復(fù)雜度降低。土壤pH、有機質(zhì)、堿解氮、有效磷和速效鉀的模型擬合度分別提高了16.88%、14.85%、5.29%、104.26%、59.94%;RMSE分別降低了36.20%、41.69%、19.54%、2.67%、0.1%;MAE分別降低了39.49%、53.49%、0.26%、3.52%、10.41%。(4)為了改善傳統(tǒng)采樣方案土壤樣點設(shè)置不合理的情況,對研究區(qū)分坡位確定取樣單元。根據(jù)研究區(qū)地形特征,將地形部位劃分為上坡位、下坡位、溝谷,并根據(jù)優(yōu)化后的樣點分布情況確定各個土壤養(yǎng)分在不同坡位上的取樣單元,結(jié)果表明:各個土壤養(yǎng)分在坡地(上坡位、下坡位)地形上的取樣單元較為接近,溝谷處的取樣單元較大;坡地(上坡位、下坡位)地形上的取樣單元小于溝谷處的取樣單元。綜合各個土壤養(yǎng)分在不同坡位上的取樣單元和野外采樣的實際情況,可將各個坡位上的土壤養(yǎng)分取樣單元平均值作為本研究區(qū)的實際取樣單元:上坡位的取樣單元為2.06 hm2,下坡位的取樣單元為1.81 hm2,溝谷的取樣單元為4.91hm2。(5)利用模擬退火算法優(yōu)化過后的土壤樣點結(jié)合BP神經(jīng)網(wǎng)絡(luò)模型對五項土壤養(yǎng)分指標(biāo)的空間分布情況進行數(shù)字化制圖,其精度較為可靠,空間分布特征符合實際情況,為丘陵區(qū)內(nèi)土壤養(yǎng)分科學(xué)管理、數(shù)字農(nóng)業(yè)、精準(zhǔn)農(nóng)業(yè)的實施提供重要的理論支撐。
[Abstract]:The spatial distribution information of soil nutrients is the basic information for the study of soil resources. It is very important to grasp the variation characteristics of soil nutrients in the study area for agricultural production and environmental simulation. Under the same condition of homologous parent material, climate conditions, farming methods and management measures, the topography is the influence of soil cultivation. The main factor of spatial differentiation is that the acquisition of soil samples takes a lot of cost for the terrain complex hilly areas. It is very important to arrange sampling points to ensure the complete expression of soil nutrient spatial information. The effective and reasonable layout of soil sampling points not only can reflect the spatial information of soil nutrients, but also greatly reduce the soil nutrients. Low cost. This paper takes the typical hilly area (2km2) of Yongxing Town, Jiangjin District, Chongqing as the research area, and studies the spatial variation law of soil pH value, organic matter, alkali hydrolysable nitrogen, effective phosphorus and available potassium content and its relation with the terrain, and uses the simulated annealing method to determine the optimal sampling density and the optimum sampling unit. Under the support of GIS technology, the neural network method combined with topographic factors was used to make predictive soil nutrient mapping. The main results were as follows: (1) the relationship between soil nutrients is close. Soil pH has a significant negative correlation with organic matter, alkali hydrolysable nitrogen and available phosphorus. Soil organic matter has a significant positive correlation with alkali hydrolysable nitrogen and available potassium, soil has a significant positive correlation. There is a significant positive correlation between effective phosphorus and available potassium. Soil nutrients have spatial autocorrelation. Soil alkali hydrolysable nitrogen and organic matter content have strong spatial autocorrelation; soil pH value, soil available phosphorus and available potassium content have moderate degree of autocorrelation. Soil nutrients and topographic factors have significant correlation. Soil pH value and Topographic Wetness The degree index (TWI) showed a significant negative correlation with the horizontal curvature (HORIZC) and the gradient (Slope), that is, soil acidity and alkalinity decreased with the increase of soil moisture content; soil organic matter, alkali hydrolysable nitrogen content was negatively correlated with Gao Cheng (Elevation), horizontal curvature (HORIZC), gradient (Slope) and relative position index (RPI), and topographic humidity. The index, the slope length (SlpLen) and the catchment area (SCA) showed significant positive correlation. The content of soil available K was significantly positively correlated with the terrain humidity index and slope length, and the correlation with other topographic factors was weak. That is, the slower the terrain, the higher the content of soil organic matter, alkaline nitrogen and available potassium, and the weak correlation between soil effective phosphorus content and topographic factors. (2) the spatial distribution of the original 200 soil sample points was optimized by simulated annealing algorithm and neural network model, and the optimum spatial distribution combination was given for the five soil nutrient indexes. At the same time, the prediction error (mean square error) was given for each sample point combination. The error results showed that the soil was soil. Soil pH, organic matter, alkali hydrolysable nitrogen, available P, and available potassium can replace the original sample at least 5,6,7,6,5, and the error is not higher than the mean square error of the original 200 samples. The optimum sample number of soil pH, organic matter, alkali solution nitrogen, available phosphorus and available potassium is 68118,87,86,60. (3) with topographic factors, respectively. The BP neural network prediction model of Soil Nutrient under the original sample and optimal layout was constructed. The results showed that the prediction model of soil nutrients increased, the prediction accuracy increased and the model complexity decreased. The model fitting degree of soil pH, organic matter, alkali hydrolysable nitrogen, available phosphorus and available potassium increased respectively. 16.88%, 14.85%, 5.29%, 104.26%, 59.94%; RMSE decreased 36.20%, 41.69%, 19.54%, 2.67%, 0.1%, respectively, and decreased by 39.49%, 53.49%, 0.26%, 3.52%, and 10.41%., respectively, to improve the layout of the soil samples in the traditional sampling scheme. The sampling units of soil nutrients at different slope positions are determined for upper slope position, downhill position and valley, and the sampling unit on the topographic topography of each soil nutrient on sloping land (upslope position, downslope position) is relatively close, and the sampling unit at the valley is larger, and the terrain on slope (upper and lower slope) is taken. The sample unit of the soil nutrient at different slope position and the actual situation of field sampling can be used as the actual sampling unit of the study area, the sampling unit of the upper slope position is 2.06 Hm2, and the sampling unit of the downslope position is 1.81 hm2. The sampling unit of the valley is 4.91hm2. (5) using the simulated annealing algorithm optimized soil sample point and BP neural network model to digitize the spatial distribution of the soil nutrient index. The precision is more reliable, the spatial distribution features conform to the actual situation, and the scientific management of soil nutrients in the hilly area, the digital agriculture and the essence. The implementation of quasi agriculture provides important theoretical support.
【學(xué)位授予單位】：西南大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：S158

【參考文獻】

相關(guān)期刊論文 前10條

1 王強;張莉莉;馬友華;張承祥;;微地形土壤養(yǎng)分空間變異特征及養(yǎng)分管理研究[J];安徽農(nóng)業(yè)大學(xué)學(xué)報;2016年06期

2 張浩;彭月月;李斌;李啟權(quán);王昌全;李冰;向金友;易蔓;;宜賓地區(qū)土壤養(yǎng)分空間變異及其豐缺狀況分析[J];中國煙草科學(xué);2016年04期

3 宋賢沖;項東云;郭麗梅;鄧小軍;曹繼釗;;貓兒山森林土壤養(yǎng)分的空間變化特征[J];森林與環(huán)境學(xué)報;2016年03期

4 王苗苗;陳洪松;付同剛;張偉;王克林;;典型喀斯特小流域不同植被類型間土壤養(yǎng)分的差異性及其空間預(yù)測方法[J];應(yīng)用生態(tài)學(xué)報;2016年06期

5 劉鵬舉;夏智武;唐小明;;基于DEM和坡面特征的坡位生成方法[J];北京林業(yè)大學(xué)學(xué)報;2016年02期

6 楊福芹;馮海寬;李振海;高林;楊貴軍;戴華陽;;基于赤池信息量準(zhǔn)則的冬小麥葉面積指數(shù)高光譜估測[J];農(nóng)業(yè)工程學(xué)報;2016年03期

7 韓宗偉;黃魏;羅云;張春弟;祁大成;;基于路網(wǎng)的土壤采樣布局優(yōu)化——模擬退火神經(jīng)網(wǎng)絡(luò)算法[J];應(yīng)用生態(tài)學(xué)報;2015年03期

8 楊建虎;常鴻莉;魏琪;;黃土高原小流域土壤養(yǎng)分空間特征及其與地形因子的相關(guān)性[J];西北農(nóng)林科技大學(xué)學(xué)報(自然科學(xué)版);2014年12期

9 孫孝林;王會利;寧源;;樣點代表性等級采樣法在丘陵山區(qū)土壤表層有機質(zhì)制圖中的應(yīng)用[J];土壤;2014年03期

10 李啟權(quán);王昌全;岳天祥;李冰;張新;高雪松;張毅;袁大剛;;基于定性和定量輔助變量的土壤有機質(zhì)空間分布預(yù)測——以四川三臺縣為例[J];地理科學(xué)進展;2014年02期

相關(guān)碩士學(xué)位論文 前3條

1 盛慶凱;基于支持向量機的土壤養(yǎng)分制圖研究[D];西南大學(xué);2013年

2 蘇曉燕;采樣設(shè)計對土壤有機質(zhì)含量空間預(yù)測精度的影響研究[D];南京師范大學(xué);2012年

3 史利江;基于GIS和地統(tǒng)計學(xué)的土壤養(yǎng)分空間變異特征研究[D];上海師范大學(xué);2006年

，

本文編號：1950572