【摘要】：精細(xì)的土壤屬性信息是開(kāi)展環(huán)境管理、監(jiān)測(cè)、建模及精細(xì)農(nóng)業(yè)等工作必需的基礎(chǔ)數(shù)據(jù),而目前可用的土壤信息是土壤調(diào)查得到的土壤圖,它們的圖斑包含的土壤信息類(lèi)型有限且可能融合其他類(lèi)型,并且屬性離散,難以滿足社會(huì)對(duì)高精度土壤信息的需求。但以往的土壤制圖研究中,已經(jīng)收集了大量的包含土壤養(yǎng)分、土壤類(lèi)型、成土環(huán)境的土壤采樣數(shù)據(jù),并且繪制了蘊(yùn)含土壤專(zhuān)家豐富經(jīng)驗(yàn)的傳統(tǒng)土壤圖,這些寶貴的歷史資料可以建立土壤景觀模型并在精細(xì)土壤信息獲取中被充分利用,縮減研究成本的同時(shí)提高研究效率和準(zhǔn)確性,然而土壤信息的獲取往往是通過(guò)地理空間分析獲取。為提高傳統(tǒng)土壤制圖成果中土壤信息的精度,本研究以地統(tǒng)計(jì)學(xué)理論、土壤景觀建模為基礎(chǔ),在地理空間分析中,對(duì)土壤環(huán)境要素提取與融合、土壤采樣布局設(shè)計(jì)與優(yōu)化、土壤景觀模型建立與完善、土壤綜合信息概括與展示四個(gè)方面進(jìn)行探索。以湖北省鐘祥市為例,先后探索的內(nèi)容為在土壤景觀關(guān)系的基礎(chǔ)上設(shè)計(jì)土壤采樣方案、在研究區(qū)內(nèi)路網(wǎng)空間分布的基礎(chǔ)上優(yōu)化土壤采樣的空間布局、在地形單元?jiǎng)澐值幕A(chǔ)上預(yù)測(cè)土壤有機(jī)質(zhì)的空間分布,得到的結(jié)論如下:(1)研究區(qū)內(nèi)的土壤景觀關(guān)系在沒(méi)有人為干擾的情況下是固定不變的,因此通過(guò)比較不同采樣方案中由對(duì)應(yīng)的采樣點(diǎn)數(shù)據(jù)求得的土壤景觀相關(guān)系數(shù)矩陣的相似性,可以驗(yàn)證研究區(qū)內(nèi)土壤景觀關(guān)系存在固定關(guān)系模式可能性。以研究區(qū)域A內(nèi)的8個(gè)地形因子和5個(gè)常規(guī)土壤養(yǎng)分為研究對(duì)象,利用空間分級(jí)統(tǒng)計(jì)方法將地形因子劃分為5個(gè)級(jí)別,采用Pearson相關(guān)性分析各分區(qū)上的樣點(diǎn),并對(duì)不同采樣布局下地形因子與土壤養(yǎng)分間的Pearson系數(shù)進(jìn)行相似性分析。不同采樣方案下地形因子與土壤養(yǎng)分間的Pearson系數(shù)相似性程度在99%以上,說(shuō)明研究區(qū)內(nèi)存在固定的關(guān)系模式。綜合大量環(huán)境因素而確定的基于地形因子分級(jí)的采樣方案A預(yù)測(cè)精度較高,以其關(guān)系模式代表固定關(guān)系模式評(píng)價(jià)其余采樣方案的合理性是可行的。本節(jié)設(shè)計(jì)的基于網(wǎng)格的3種傳統(tǒng)隨機(jī)采樣方案中,含有3661個(gè)采樣點(diǎn)的方案C與采樣方案A獲得的關(guān)系模式的相似性最接近,因而比其他方案樣點(diǎn)布局更合理。(2)城鎮(zhèn)化的發(fā)展會(huì)使區(qū)域內(nèi)的交通網(wǎng)絡(luò)更加發(fā)達(dá)而趨于穩(wěn)定,可為野外土壤采樣提供便捷的交通條件。以湖北省鐘祥市東部區(qū)域(研究區(qū)B)為研究對(duì)象,基于路網(wǎng)設(shè)置5種采樣尺度進(jìn)土壤采樣,采樣過(guò)程中,采樣數(shù)據(jù)可能會(huì)因?yàn)橐恍┮蛩囟a(chǎn)生誤差,采樣點(diǎn)的空間布局也可能不是十分合理,影響最終研究結(jié)果的精度,因此需要剔除采樣數(shù)據(jù)中誤差較大的采樣點(diǎn)。運(yùn)用模擬退火算法對(duì)各樣點(diǎn)的空間布局分別進(jìn)行優(yōu)化,以獲取基于路網(wǎng)的土壤采樣優(yōu)化布局。在此基礎(chǔ)上,對(duì)地形因子和優(yōu)化后樣點(diǎn)的有機(jī)質(zhì)建立多元線性回歸模型,同時(shí)建立基于神經(jīng)網(wǎng)絡(luò)的多層感知機(jī)模型,并用此模型精度與多元線性回歸模型精度進(jìn)行對(duì)比。結(jié)果表明:利用道路網(wǎng)制定土壤采樣方案是可行的,優(yōu)化后的采樣點(diǎn)布局能夠準(zhǔn)確獲取土壤景觀知識(shí),并且優(yōu)于原始樣點(diǎn)的精度。本研究利用道路空間分布格局、歷史樣點(diǎn)、數(shù)字高程數(shù)據(jù)等可利用資源設(shè)計(jì)采樣方案,為減少采樣成本、提高采樣效率、展現(xiàn)有機(jī)質(zhì)空間分布格局提供了有效手段與理論依據(jù)。(3)以鐘祥市(研究區(qū)A)土壤有機(jī)質(zhì)空間分布為例,克服傳統(tǒng)地形分類(lèi)方法中僅依據(jù)單一指標(biāo)(如高程)的缺點(diǎn),綜合由30m精度數(shù)字高程模型生成的地形因子,依據(jù)其在不同地形條件下的層次組合規(guī)律構(gòu)建地形分類(lèi)規(guī)則,精確地劃分為13種典型地形單元,并運(yùn)用普通克里金法對(duì)不同地形單元內(nèi)的土壤樣本插值,獲得相應(yīng)區(qū)域的土壤有機(jī)質(zhì)空間分布。通過(guò)組合各地形范圍下的結(jié)果,以獲取蘊(yùn)含地形因素影響的有機(jī)質(zhì)空間分布。研究發(fā)現(xiàn),地形起伏較大的地形單元的預(yù)測(cè)精度與全局預(yù)測(cè)結(jié)果精度相似度達(dá)0.75,而地勢(shì)平緩區(qū)域內(nèi)的預(yù)測(cè)精度大幅度提升,比全局預(yù)測(cè)結(jié)果精度提升了16.39%,因此基于地形單元的空間預(yù)測(cè)能精確有效地獲取土壤有機(jī)質(zhì)空間特征。利用地形分區(qū)獲取較高精度的有機(jī)質(zhì)空間分布,進(jìn)一步探討了有機(jī)質(zhì)地統(tǒng)計(jì)學(xué)研究中地形的協(xié)同影響。
[Abstract]:Fine soil attribute information is the basic data necessary for environmental management, monitoring, modeling and precision agriculture. At present, available soil information is the soil map obtained from soil survey. Their patches contain limited types of soil information and may merge with other types, and their attributes are discrete, which is difficult to meet the high precision of society. However, in the past soil mapping research, a large number of soil samples including soil nutrients, soil types and soil forming environment have been collected, and traditional soil maps containing rich experience of soil experts have been drawn. These valuable historical data can be used to establish soil landscape models and obtain fine soil information. In order to improve the precision of soil information in traditional soil mapping results, this study is based on Geostatistics theory, soil landscape modeling, geospatial analysis, and soil ring. Four aspects are explored: extraction and fusion of environmental factors, design and optimization of soil sampling layout, establishment and perfection of soil landscape model, generalization and display of soil comprehensive information. The spatial distribution of soil organic matter was predicted on the basis of topographic unit division by optimizing the spatial distribution of soil sampling. The conclusions are as follows: (1) The relationship of soil landscape in the study area is fixed and unchanged without human disturbance. The similarity of soil landscape correlation coefficient matrix can validate the possibility of the existence of a fixed relationship model in the study area. Eight topographic factors and five conventional soil nutrients in the study area A were studied. The topographic factors were classified into five levels by using spatial classification statistical method, and each score was analyzed by Pearson correlation analysis. The Pearson coefficients between topographic factors and soil nutrients under different sampling arrangements were analyzed. The similarity of Pearson coefficients between topographic factors and soil nutrients under different sampling schemes was more than 99%, indicating that there was a fixed relationship model in the study area. Among the three traditional random sampling schemes based on grids designed in this section, scheme C with 3661 sampling points has the closest similarity to the relational pattern obtained by sampling scheme A because its relational pattern represents the fixed relational pattern. (2) The development of urbanization will make the transportation network more developed and stable, and provide convenient transportation conditions for field soil sampling. Sampling data may cause errors due to some factors, and the spatial layout of sampling points may not be very reasonable, which affects the accuracy of final research results. Therefore, it is necessary to eliminate sampling points with large errors in sampling data. On this basis, a multivariate linear regression model is established for the topographic factors and the organic matter of the optimized sample points, and a multilayer perceptron model based on neural network is established. The precision of the model is compared with that of the multivariate linear regression model. The optimized layout of sampling points can accurately acquire soil landscape knowledge, and is superior to the accuracy of the original sampling points. This study uses the available resources such as road spatial distribution pattern, historical sampling points and digital elevation data to design sampling scheme, which provides an effective way to reduce sampling cost, improve sampling efficiency and exhibit the spatial distribution pattern of organic matter. (3) Taking the spatial distribution of soil organic matter in Zhongxiang (study area A) as an example, to overcome the shortcomings of traditional topographic classification methods which only depend on a single index (such as elevation), the topographic factors generated by the 30m precision digital elevation model are synthesized, and the topographic classification rules are constructed according to their hierarchical combination rules under different topographic conditions. The land is divided into 13 typical topographic units, and the spatial distribution of soil organic matter is obtained by interpolating soil samples in different topographic units with ordinary Kriging method. The accuracy similarity between the element prediction and the global prediction is 0.75, while the prediction accuracy in the gentle terrain region is greatly improved, which is 16.39% higher than that of the global prediction. Therefore, the spatial prediction based on topographic units can accurately and effectively obtain the spatial characteristics of soil organic matter. Spatial distribution further explores the synergistic effect of topography in organic texture statistics.
【學(xué)位授予單位】：華中農(nóng)業(yè)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類(lèi)號(hào)】：S159

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 趙明松;張甘霖;王德彩;李德成;潘賢章;趙玉國(guó);;徐淮黃泛平原土壤有機(jī)質(zhì)空間變異特征及主控因素分析[J];土壤學(xué)報(bào);2013年01期

，

本文編號(hào)：2203601