集水區土砂生產量之估計研究

Abstract

台灣現有水庫多興建於河川集水區中、上游，常因集水區地形陡峻、地質脆弱，或因集水區內遭人為濫墾、濫伐，導致表土沖蝕嚴重，河水挾帶大量泥砂入庫，水庫容積日減，極不利於水庫之運轉操作。根據歷年水庫淤砂量測量結果，全台灣水庫平均每年淤積量高達1,460萬立方公尺，約為一座明德水庫之總蓄水容量，顯見台灣現有水庫之淤積情況已十分嚴重，急待謀求解決之道。 近年來國內水利事業單位已積極落實河川集水區管理工作，並採用水土保持界現有之沖蝕公式加以推估集水區之土砂生產量，而影響河川集水區之泥砂來源及生產量之因素複雜，難以制訂具前瞻性之集水區土砂採取策略。故集水區之土砂來源及土砂生產量之準確估計，已成為今日刻不容緩之重要課題。 因為臺灣地質構造複雜破碎，土石材料豐富、加上主流坡度陡峻及降雨量集中等因素，致使臺灣河川之輸砂量頗大，故土石、泥砂之治理亦成為河川治理之重要考慮因素。在河川治理規劃階段，有關水理演算模式之輸砂邊界條件難以評估，故仍有賴對於河川集水區泥砂產量之合理估計，方能解決此一難題。 本研究乃針對以上議題，以竹、苗地區之大埔水庫為研究標的，進行集水區之土砂生產量估計研究；研究方法先評估、選定影響集水區土砂生產量之特徵因子，藉由水文統計分析及遙測衛星影像分類處理，分別求出各特徵影響因子與集水區土砂生產量之線性關係式，再以主成份分析及迴歸統計法求得適用本集水區之土砂生產模式。 本研究之成果顯示：本研究建立之土砂生產模式，以集水區累積降雨量自變數、集水區累積逕流量自變數、或集水區累積降雨量及累積逕流量組合自變數模式等三種為較佳模式。本研究所採用之多年次遙測衛星影像分類處理程序，對於集水區土地利用與地表覆蓋因子之變化量估算，顯示此法可以獲得有用之地表資訊，將來可考慮應用於各種地理、水文、環境研究領域。本研究採用之主成份分析，除可減少迴歸分析中自變數之個數，相對增加迴歸條件式與迴歸精度，並可有效解決複迴歸分析之自變數間相關性之問題。本研究建立之大埔水庫土砂生產模式，可提供做為水庫營運及排砂操作之參考依據。此外遙測衛星影像分類成果將來藉多年次之影像差異分析，可作為集水區管理之重要依據。

The existed reservoirs in Taiwan are mostly constructed at the middle and upper reaches of the rivers. Frequently due to the steep topography and frail geology or due to improper cultivation and deforestation, they have resulted in serious erosion to the top-soil. The river water carries abundant of sand to the reservoir; subsequently, the latter’s capacity is reducing day by day, which is very disadvantageous to the operation of reservoir. According to the measured results to the sand-depositing volume of reservoir in the past years, the average deposit volume of the reservoirs throughout entire Taiwan has reached up to 14.6 million cubic meters, which is nearly equivalent to the total storage capacity of Ming-Te Reservoir. It is obviously indicating that the deposit situation of existed reservoirs in Taiwan has been very serious and urgently needs to search resolutions. Recently, the domestic hydraulic organizations have aggressively implemented the works of river watershed management and adopted the existed erosion formula used by the soil conservation field to estimate the soil loss at watershed. However, the factors of sediment origin and loss volume, that will impact the river watershed, are very complicated and also very difficult to formulate a prospective strategy of sediment extraction at watershed. Therefore, the precise estimation of sediment origin and loss at watershed has become an important issue with great urgency, nowadays. Moreover, owing to the factors that the geologic structure of Taiwan is complicated and fragile and the unconsolidated materials are plentiful, adding that the gradient of main rivers is steep and the rainfall is concentrated in short periods of time, they have resulted in great sediment yields. Consequently, the measures for mitigating sediment yields have also become important factors to be considered in administrating rivers. At the planning stage of river management, it is hard to obtain parameters of boundary conditions for deriving a sound hydrographic model for understanding the sedimentation and sediment transportation process. Therefore, it is highly required to have an approach to obtain an estimation of sediment yields. This research, to the aforesaid issues, is using the Ta-Pu Reservoir at Hsinchu and Miaoli districts as the study area to proceed to an estimating research of soil loss volume in river watershed. The research measure is firstly evaluating and selecting the influence factors that will affect the soil loss volume at watershed. Subsequently, through the hydrological statistical analysis and remote sensing satellite image processing, we can obtain the linear relation between each influence factor and soil loss volume at watershed respectively and then reach the model of soil loss applicable to this watershed through principal component analysis and regression analysis. Models with different combinations of parameters are scrutinized in this study. It is revealed that there are three models established in this research showing good prediction of sediment yields. They are the models applying factors of accumulated rainfall variable at watershed, accumulated runoff variable at watershed, and an assembled variable model of accumulated rainfall and accumulated runoff at watershed. The multi-temporal remote sensing satellite image processing procedure adopted in this study shows a very good time-series estimation of the land use and land cover. Similar procedure can be applied in extracting thematic information such as physiographic, hydrologic and environmental features. The principal component analysis adopted by this research, besides reducing the amount of variables in regression statistical analysis, can relatively increase the correlation coefficient for the regression conditions. In addition, it can also effectively solve the problem of the correlation of variables of compound regression analysis. The model of soil loss for Ta-Pu Reservoir as established in this study can be applied as the referential bases of reservoir operation and sediment yield mitigation. Furthermore, the result of remote sensing satellite image processing through multi-temporal image change analysis can be further applied as an important basis for the administration of watershed.