不同單元尺度對土地利用及生態系統服務模擬之影響-以大屯溪流域為例

Abstract

土地利用之組成(composition)及配置(configuration)對維持生態系統及其服務功能有著決定性之影響，除了會影響生態系統之功能，甚至改變生態系結構，進而牽動生態系統之多樣性情況。但當我們透過地理資訊系統工具，進行土地利用變化的觀測或生態系統服務之模擬計算時，不同的網格解析度會詮釋出不同的地景格局樣式，導致不同的生態系統服務計算結果，進而影響到相關土地利用管理政策之制定與保護區域之劃定。尺度變換下所產生之普遍性問題，通稱「可調整地區單元問題」(Modifiable areal unit problem，簡稱MAUP)，也因此，了解尺度問題所帶來的影響以及選擇適當的解析度來進行研究分析十分重要。 本研究以台北大屯溪流域為例，探討不同尺度對土地利用以及CLUE-s(Conversion of Land Use and its Effects for Small Region)土地利用變遷模式模擬結果之影響，並使用InVEST (Integrated Valuation of Environmental Services and Tradeoffs) 生態系統服務模式探討尺度、土地利用與生態系統服務之間的關係。結果發現，尺度於土地利用的影響隨著網格增大，土地利用分類正確率下降，地景嵌塊體會逐漸聚合、聚集與均質化，而土地利用變遷模式所挑選之驅動力因子個數則隨著網格的增大而減少且有所不同，建議進行土地利用分析之最適解析度為所取得資料之最高解析度，最低解析度閾值為100m*100m。尺度的變化對於生態系統服務中的棲地品質影響最大，其次為氮營養鹽與沉積物留存，而出水量則無明顯影響，各生態系統服務之全域空間自相關性隨著網格的增大而遞減，且在空間分布上具有空間自相關性之顯著範圍大多座落在2000m~2500m內，根據本研究結果，建議進行各項生態系統服務模擬之最適解析度為100m*100m，且最低解析度閾值應為150m*150m。本研究結果可作為分析類似集水區面積大小或配置之參考依據，或應用於大型集水區之最小子集水區單元。

Composition and configuration of land use have a decisive influence on maintaining ecosystem structure, services and functions. Such changes can further affect ecosystem biodiversity. When geographic information system tools are applied to investigate changes of land use or to calculate ecosystem services, land use data of different cell size might present different landscape patterns, leading to different ecosystem service results which would affect relative policy making in land use management or ecoregion planning. These common problems due to scaling effects are named as Modifiable Areal Unit Problem (MAUP). Hence, it is important to understand the influence of scaling effects and to select an appropriate resolution for research conduction. The study takes Datum watershed in New Taipei County as study area to investigate how various scales affect land use and land use modeling, Conversion of Land Use and its Effects for Small Region (CLUE-s), and to discuss relationship between scale, land use and ecosystem services by using Integrated Valuation of Environmental Services and Tradeoffs (InVEST). Results showed that as the cell size increased, the accuracy of land use classification decreased and patches of landscape become more aggregated and homogenized. Moreover, the numbers of driving factors chosen by land use change model decreased as cell size increased. It is suggested to use the finest resolution (50*50m) as the most adequate resolution for interpreting land use data, while the higher limit of resolution threshold is 100m*100m. Resolution change has the greatest impact on habitat quality compared to other ecosystem services, followed by nitrogen and sediment retention, and no significant influence on water yield. Moreover, the global spatial autocorrelation on each ecosystem service decreased as cell size increased, significant range of local spatial autocorrelation was mostly located between 2000~2500m. According to the results, it is suggested the most appropriate resolution for modeling ecosystem services is 100m*100m, and the lowest resolution threshold should be 150m*150m. The study provides useful information for analyzing other watersheds of similar size or configuration, and can be applied as the smallest subwatershed unit in a large watershed.