複雜含水地層之抽水沉陷行為

Abstract

超抽地下水引起之地層下陷容易發生在含水豐富卻又軟弱之現代沖積地層，該類地層組構複雜，不易評估其壓縮特性，因此需將地層進行適當的簡化並考量現地土層之壓縮反應，以簡便及確實的評估其抽水沉陷行為。 傳統上評估現地土層之沉陷行為以室內土壤力學試驗為主，但對於複雜地層而言，室內試驗成果之代表性及可靠度均不足，唯有倚賴整合型之現場監測方可有效評估地層之實際壓縮性質。本文系統性整合分析濁水溪沖積扇及屏東平原之地面高程測量、分層土層位移及地下水位變化等實測資料，經歸納分析不同地層受地下水位的變動影響之壓縮行為，礫石層(含粗砂)為完全彈性變形、砂土層(中細、砂)為塑性變形、黏土層之壓密特性則符合Terzaghi單向壓密理論。濁水溪沖積扇沿海之彰化與雲林之地層分別處於正常壓密及輕微過壓密狀態，壓縮量主要發生在深度介於60~230公尺之地層(阻水層二及含水層二、三)，其中黏土層及砂土層之體積壓縮係數分別約6.4´10-8 m2/N and 5.7´10-9 m2/N。屏東平原之礫石層分布很廣，其體積壓縮係數經評估約1.0~1.3´10-9 m2/N。 本論文結合現地之經驗模型及土壤力學之材料壓縮理論，建立一套單向度之地層壓縮評估模式，將地層依顆粒尺寸簡化為礫石層(含粗砂)、砂土層(含中細砂)及黏土層(含粉土)等三種地層。本論文將地層之壓縮應變區分為彈性及非彈性部份，並假設其應變量與地下水位變化存在線性關係。接著並將地層概念壓縮模型建立成數值計算模式，應用Crank-Nicholson隱性差分法求解黏土層之依時壓密量，可適用於不同時間期距之土層沉陷計算；礫石及砂土層之壓縮則視為與時間無關之立即性變形。 本論文於數值模式中結合遺傳演算法(Genetic algorithm)，可根據地面沉陷及地下水位變化之實測資料，進行地層參數之優選，優選之地層參數包括黏土層之透水係數、彈性壓縮係數、非彈性壓縮係數，礫石層之彈性壓縮係數及砂土層之非彈性壓縮係數等。 模式經以現地實測資料驗證，模擬計算值與實測值之平均誤差可控制在3~5%以內；且在地層壓縮參數建立後，日後只需地下水位變化資料即可進行地層下陷之趨勢預測，可見模式具有良好之可靠性及實用性。

Usually, compressible multi-aquifer system existed in a recently deposited alluvial formation is very susceptible to land subsidence induced by over-pumping of ground water. Due to the complex hydrogeology, true strata compressibility is often quite difficult to be evaluated. From time to time, it is often necessary to assess the soil compressibility by means of simplifying soil types and considering equivalent field compressive. To find the compressibility and relevant soil properties of strata, the common approach is through field exploration and sampling undisturbed specimens for laboratory soil tests. However, results of laboratory tests revealed that they are unreliable and unrepresentative for complex strata. Hence, the investigation for land-subsidence relies very much on the integrated field monitoring. This study analyzed the integrated data of leveling survey, multi-level strata deformation and groundwater head fluctuation from the Choshui River alluvial fan and the Pingtong plain. From the field evidences of the interaction between the strata compressive deformation and the change of groundwater head, it was observed that the compression of gravel stratum and sandy stratum behave as perfectly elasto-plastic materials; while clayey stratum basically follows the Terzaghi’s one-dimensional consolidation theory. It was also found that the clay in ChangHua County was in a normally consolidated state, while the clay in YunLin County was in an over-consolidated state. The ground settlement was mainly resulted from the compression of sandy stratum within 60-230 m (including aquifer 2,3 and aquitard 2). The coefficients of volume compressibility of the clayey and sandy stratum were back analyzed from the stratum’s compression records; they were 6.38´10-8 m2/N and 5.71´10-9 m2/N, respectively. The coefficient of volume compressibility of the gravel strata commonly present in Pingtung plain was around 1.0~1.3´10-9 m2/N. This study integrated empirical model and compression theory of various types of soils to develop a one-dimensional model of land subsidence due to withdraw of ground water. The proposed model classified the field strata according to grain sizes. Soil strata were divided into three categories: namely, gravel stratum (including coarse sand), sandy stratum (including fine sand) and clayey stratum (including silt). This study further separated the compressive deformation of strata into elastic compression and inelastic compression. The compression deformation was assumed to have a linear relation with the change of ground water level. The concept of the compression model was further utilized to develop a numerical model; an implicit Crank-Nicholson scheme was used to solve the time-dependent consolidation of clayey strata. The compressions of gravel and sandy strata were considered to be instantaneous deformation independent to time. A numerical model incorporating a genetic algorithm (GA) was also developed to search for a set of optimized parameters (including permeability and compression coefficients of various strata) for the characterization strata compressibility based on the compiled data of ground subsidence and the change of groundwater level. Comparing the calculated results and the field data, the average error of the calculated results is within 3-5%. It reveals that the proposed model is an appropriate tool for the prediction of ground subsidence as long as the data of groundwater level is available.