- 期刊
灰色預測GM(1,1)模型應用於粒徑分佈的評估
The Evaluation of Grey Predictive GM (1,1) Model Applied into Soil Particle Distribution
摘要
目前在國內外的文獻中，對於評估土壤粒徑分佈模型之相關研究實在不多見，唯見Skaggs等人於西元2001年(Skaggs et al., 2001)提出的土壤粒徑評估之經驗模型，此模型在沉泥(Silt)組成百分比大約70%時會出現極大的預測誤差，因為目前Skaggs等人提出的粒徑分佈預測模型之預測精準度仍有很大需要改進的空間，本研究從本研究場址中的222個土樣利用灰色預測GM(1,1)模型（簡稱GM(1,1)模型）來進行土壤粒徑分佈的評估，並提出此模型不適用的粒徑分佈條件及與Skaggs經驗模型進行預測精度上的比較。本研究發現GM(1,1)模型除了在砂質(Sand)土壤之評估較Skaggs經驗模型差之外，其餘的土壤種類皆優於Skaggs經驗模型。本研究在利用均勻係數由小而大分類各種土樣之過程中發現均勻係數在大於300時，相當適合GM(1,1)模型之評估，而且隨著均勻係數的增加，各個土壤粒徑分佈評估模型之累積誤差值有減少的趨勢。本研究亦發現Skaggs經驗模型在土壤樣本的沉泥(Silt)百分比超過70%時會出現極大的誤差，此結果與Skaggas等人(Skaggs et al., 2001)所得之結果相同。
並列摘要
Nowadays, there are few studies abroad and in Taiwan on models for estimating soil particle-size distributions. Only Skaggs et al. (2001) presented the empirical model of soil particle-size distribution. This model has a large prediction error when about 70 percent of the soil is silt because the model needs to be improved. In this study, 222 soil samples were selected from several sites and the Grey Predictive GM(1,1) model (GM(1,1) model) was used to estimate the particle-size distribution of the soil samples. The GM(1,1) model can be compared with Skaggs empirical model to point out the unfitted conditions of the GM(1,1) model and their accuracies. In this study, the GM(1,1) model was found to be better than Skaggs empirical model except with sand soil. The coefficient of uniformity from the smallest to the largest was used to distinguish the different types of soil and found that when the coefficient of uniformity was greater than 300, the GM(1,1) model made good estimations. Also found was that when the coefficient of uniformity increased, the accumulative absolute error of both the soil particle-size distribution models decreased.