Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Chine → 2017 → Spatial Variability of Soil Properties and Spatial Allocation of Salt Discharge Areas for Irrigation Districts

Université de l’agriculture de Chine (2017)

Spatial Variability of Soil Properties and Spatial Allocation of Salt Discharge Areas for Irrigation Districts

黄亚捷

Titre : Spatial Variability of Soil Properties and Spatial Allocation of Salt Discharge Areas for Irrigation Districts

Auteur : 黄亚捷;

Grade : Doctoral Dissertation 2017

Université : Université de l’agriculture de Chine

Résumé partiel
Soil salinization is increasingly severe in Yinchuan North Irrigation District due to lack of water resource. Although existing studies tried to alleviate the soil salinization by land consolidation engineering methods, chemical and biological methods, the problem of soil salinization has not yet been resolved fundamentally to date. This is because the multi-functional spatial allocation for water and soil resources, including cultivated land production, salt discharge reservation and ecology, was still unreasonable. Very few studies have considered soil water and salt constraints, and examined use of spatial allocation of cultivated land and salt discharge area (uncultivated land) of land consolidation to solve the effect of soil salinization. The objective of this study was thus to solve the soil salinization problem by comprehensively considering water and salt constraints and the spatial allocation of water and soil resources. Firstly, on the basis of geostatistics and artificial neural network, we proposed effective methods for predicting soil properties with a high accuracy, thus obtaining regional parameters of soil water and salt movement. Then, based on the optimal management of irrigation and drainage, we investigated the effects of different irrigation and drainage treatments on soil salinity. Finally, we set up scenarios of cultivated land and salt discharge reservation by SahysMod, and proposed the optimal allocations of cultivated land and salt discharge reservation of land consolidation. Results of this study provide a theoretical basis and technical guide for the spatial allocation of cultivated land and salt discharge reservation of land consolidation, and hence have important implications to preventing soil salinization in salt-affected farming areas. The main results and conclusions are in the following.Firstly, geostatistics (e.g. OK and RK) cannot capture regional changes in soil properties due to the nonlinear relationships between soil properties and environmental factors. In addition, artificial neural network only consider the variations of the soil properties caused by correlated influencing factors,regardless of spatial autocorrelation of the surrounding measured data ; or it only consider the spatial auto-correlative information on coordinates (X, Y) of survey points as input, regardless of environment factors. Thus, this study eliminated the aforementioned limiations by combining ordinary kriging with back-propagation network (OK_BP) (i.e., considering the two aspects of spatial variation) and expected to improve the mapping accuracy of soil salinity. Our results showed that OK_BP provided the best mapping accuracy among the four methods (i.e., OK, BP, RK and OK BP). Futhermore, OK BP revealed more details of the spatial variation responding to influencing factors, and provided more flexibility for incorporating various factors in mapping. Therefore, OK BP can serve as an effective method for mapping soil salinity with a high accuracy.Second, due to nonlinear relationships between soil properties and their correlated factors, the relationships can change with location in the same region. This is particularly true in regions of highly heterogeneous landscapes. However, existing study frequently assumed the linear relationships between soil properties and the factors to improve the prediction accuracy, and neglected different effects of these factors on soil properties in localized regions. By considering the effects of topography, soil type,soil texture and land use on soil organic matter, our study used self-organizing map to cluster these factors into different regions and then used ordinary kriging combined with the clustering of a self-organizing feature maps neural network (KCSOM) to predict soil organic matter. Results showed that the use of KCSOM was effective to describe the nonlinear relationships between the soil organic matter and its correlated factors. KCSOM effectively avoided underestimation of the higher values of the interpolation surface and overestimation of the lower values that existed in the other methods

Mots clés : soil property; geostatistics; neural network; soil water and salt movement; spatial allocation of salt discharge area; model;

Présentation (CNKI)

Page publiée le 20 janvier 2018