Study On Establishing The Early-Warning Model Of The Soil Salinization Hazard In Ugan-Kuqa River Delta

As a process of the ecological environment degradation, soil salinizationbecome a disaster to a certain extent, the serious consequences of this disaster isdeclining or losing the biological or economic productivity and complexity of land,reduction or nothing of yield of crops, posing a serious threat on the agriculturesustainable development. The Ugan-Kuqa River Delta Oasis is one of the serious soilsalinization areas in our country. The blindly expansion of cultivated land, heavilyexploitation the surface water to irrigation and unreasonable use of land resources,resulting in the rise of the groundwater level and declining soil fertility. Theworsening soil salinization not only restricts agriculture and economic development,but also affects food production and ecological environment in study area. Soilsalinization is the result of a variety of natural conditions and socio-economicconditions in the different spatial and temporal scales interactions and coupling, fromthis sense, the soil salinization warning belong to the scope of disaster managementwarning.Warning is mean in before disasters or catastrophes and the other dangers whichneed to watch out for happen, report an emergency signal of dangerous situations tothe relevant departments based on the summary of the law or observed precursor, andavoid hazards in the case of without the knowledge or lack of preparation, thus toreduce the harm caused by the loss of behavior. Therefore, the basic framework of thesoil salinization disaster warning system including the procedures of clarifyingwarning meaning, finding warning source, analyzing warning sign, predicatingwarning degree and removing warning objects in this study. According to the manyyears of practical investigations, Ugan-Kuqa oasis soil salinization warning degreecan be divided into four categories: no warning, light warning, middle warning andserious warning.With the rapid development of computer and information technology, expansionthe3S technology's applications and the boom re-emergence of artificial intelligence technology, the research of early warning applied in ecology, environment,agriculture and climate and made a huge progress. Consider the actual situation of thestudy area, obtained data and the reference to the experience of previous studies, weselected of BP neural network model, RBF neural network model and fuzzy logicmodel to establish the soil salinization hazard prediction model.1. Clarifying warning meaning, select the warning factors, establish earlywarning indicators. By the disaster warning source analysis of soil salinization, thesalinization evaluation unit was generated and information was extracted with theapplication of GIS spatial information management, the geostatistical analysis and3Danalysis.(1) Take a normal distribution test, semi-variance model analysis togroundwater level and salinity data of38observation well in study area, finallythrough the Kriging spatial interpolation result, obtained the groundwater level andsalinity data of71sampling points.(2) Potential evapotranspiration is combinedaffected by temperature, wind speed, sunshine hours, relative humidity, precipitationand other factors, so selection of potential evapotranspiration instead of climaticfactor's effect to sample points is reasonable. In order to improve data accuracy andreliability, take the Kuqa, Xingha, Shaya area as a center, selected18counties aroundthem, and used the Penman-Monteith equation to calculate the potentialevapotranspiration by50years climate data, and after the spline interpolation, obtainthe required sample point information.(3) By analysis of satellite images of the studyarea and the relevant information, the sampling points are mainly concentrated in thealluvial plains, so in this paper we take the slope value instead of the steepness of thesurface unit. With the use of ArcGIS software's3D Analysis tool, extract groundelevation and calculate slope extraction.(4) The impact of human activities on soildegradation embodied in the land use types. Therefore, by the used of the land usemap of the study area, we extract land-use types of the71sampling points withArcGIS software' attribute query function, and take quantitative conversion.2. Estabilish the soil salinization disaster evaluation model. The soil salinizationformed is the result of the role of climate, hydrology, parent material, topography andother natural factors and reclamation and irrigation and other human factors.(1) By the used of gray correlation degree model, we analysis the soil salinization throughsoil salinity as the parent sequence, other influencing factors as subsequences. Theresults show that the size of the correlation between soil salinity and other factors: thepotential evapotranspiration>land use type>soil conductivity> groundwaterlevel>groundwater salinity> TDS> slope value> PH value.(2) Based on the arid soilsalinization research of many experts at home and abroad, the results of theconsultation of experts and the results of gray system model, establish a hierarchicalmodel and structural matrix by Analytical Hierarchy Process (AHP). We evaluatedthe natural and geographical factors, soil internal factors and human affected activitieswhich affecting soil salinization, and finally calculate the weight of each warningfactor.(3) Referencing the research of risk evaluation model and early warningforecasting model proposed by domestic and international researchers, combiningwith land desertification hazard early warning principle, considering the formationand development of soil salinization impact factor, we establish a simple and practicalsoil salinization disaster warning evaluation model. Finally, considering the oasisactual situation, we proposed control measures and improvement measures. This willhelp to evaluate and analyze the dynamic changes of soil salinization disaster; caneffectively reduce the losses caused by disasters.3. Build an early warning model. With the use of artificial intelligencecomputing and fuzzy mathematics method, the feasibility of early warning wereanalyzed, after repeated adjustment of network structure and parameters, soilsalinization hazard prediction model were established based on BP neural network,RBF neural network, fuzzy mathematics. The results show that the accuracy of thesethree methods was acceptable and suitable to soil salinization hazard warning. Thestudy provides an effective and viable new ways to analyze and predict the soilsalinization, it is a complement to the traditional dynamics monitoring of soil salinity.