植物病蟲害嚴重度估計的指標與方法之研究

Abstract

正確及精確地估計病害嚴重度（或稱為'罹病度'）或測量蟲害取食面積是植物保護相關研究的重要基礎，本論文主要探究如何提高植物病害嚴重度指標估計及蟲害取食面積測量方法的正確性及精確性。 植物「病害嚴重度指標」是指以單一數值來說明單株或數個植物樣本之整體病害嚴重度，通常是由順序類別尺度之等級所計算而得，這些尺度的等級是由病害面積(百分比數值形式)劃分而得。傳統上，「病害嚴重度指標」(%)被計算為= [總合 (類別頻率× 每個類別之等級)] / [(全部株樹) × (最大之病害嚴重度指標)] × 100，植物病理學家和其他專業人員利用這些等級之數據，計算得到「病害嚴重度指標」並接續進行處理比較。然而，已有多個研究指出，類別尺度的劃分，對估計病害嚴重程度的正確性及精確性有很大的影響。因此，本論文的第一個目標，是探究不同類別尺度（即病害嚴重度之劃分區間是否等距及區間寬度）以及等級代表數值的選擇（即該類別的等級次序數或組中點值），在「病害嚴重度指標」估計以及後續進行處理比較的影響。 對於「病害嚴重度指標」的估計，上述所提及之傳統計算方式，因並未考慮尺度區間的組中點值，因此特別容易高估「病害嚴重度指標」；而在模擬研究的結果顯示，若評估者的估計是無偏的，則估計「病害嚴重度指標」的最準確方法是考慮尺度區間的組中點值，並使用修正之10％等距劃分病害尺度（即病害嚴重度在50％以下10％等距劃分，加上低病害嚴重度時再加以細分）；若有評估者的偏差下，計算「病害嚴重度指標」估計的準確性主要取決於評估者的偏差相對於實際平均值的程度及方向。因此，減少評估者的評估偏差，是為降低「病害嚴重度指標」估計誤差的關鍵。 當進行比較處理時，影響擬說測驗檢定力之關鍵在於不同類別尺度（即病害嚴重度之劃分區間是否等距及區間寬度），反而不是等級代表數值的選擇（即該類別的等級次序數或組中點值）。一般而言直接利用百分比數值的病害嚴重程度是最理想的，但是使用類別尺度有其實際應用性及可行性時，又必須使用「病害嚴重度指標」表示之，建議選用修正之10％等距劃分病害尺度（即病害嚴重度在50％以下10％等距劃分，加上低病害嚴重度時再加以細分），將在擬說測驗上提供最佳檢定力。 本論文的第二個目標是為蟲害取食面積之嚴重度估計，提供另一個可行的替代方法。傳統上的人工方法是利用方格紙來描繪和測量蟲害取食面積，但這個方法可能耗費較多的時間和勞力；為了提供更有效率的測量方法，本研究藉由ASSESS 2.0以影像分析來測量蟲害取食面積(這裡以水稻瘤野螟為例)，並利用Lin的一致性相關係數（ρc）來確認ASSESS2.0測量結果的正確性和精確性。結果顯示影像分析的測量結果與人工方法（視為黃金標準）具有高度一致性（ρc高於0.983），且影像分析至少減少29％的測量時間；因此，藉由ASSESS 2.0所進行之影像分析，將可推薦作為測量水稻瘤野螟取食稻葉面積的工具。

The accuracy and precision of estimating disease severity or pest damage are among the most important foundation for research on plant protection. This thesis explores how to improve the accuracy and precision of estimating the disease severity index (DSI) and of the measurement of the pest damaged area. The DSI is a single number for summarizing the total effect of the disease on a single plant or a small sample of plants. The DSI has most often been used with data based on a special type of ordinal scale [e.g. the Horsfall-Barratt (H-B) scale] comprising a series of consecutive ranges of defined numeric intervals, generally based on the percent area of symptoms present on the specimen(s). The traditional DSI is calculated as DSI (%) = [sum (class frequency × score of rating class)] / [(total number of plants) × (maximal disease index)] × 100. Plant pathologists and other professionals use ordinal scale data in conjunction with a DSI for treatment comparisons. However, several previous studies have demonstrated that different types of ordinal scale greatly influence the accuracy and precision of estimating disease severity. Therefore, the first objective of this thesis is to investigate the effects on both of the different scales (i.e. those having equal or unequal classes, or different widths of intervals) and of the selection of values for scale intervals (i.e. the ordinal grade for the category or the midpoint value of the interval) on the DSI estimates made for treatment comparison. For the DSI estimates, the finding is that the traditional DSI is particularly prone to overestimation when using the above formula if the midpoint values of the rating class are not considered. Moreover, the results of the simulation studies show that, compared with other methods tested in this study, if rater estimates are unbiased, than the most accurate method for estimation of a DSI is to use the midpoint of the severity range for each class in an amended 10% ordinal scale (an ordinal scale based on a 10% linear scale emphasizing severities ≤50% disease, with additional grades at low severities). As for biased conditions, the accuracy for calculating DSI estimates will depend mainly on the degree and direction of the rater bias relative to the actual mean value. Hence reducing rater-related bias is critical to minimizing the introduction of additional errors into the progress of calculating the DSI. For the treatment comparison, the principal factor determining the power of the hypothesis test is the nature of the intervals, not the selection of values for ordinal scale intervals (i.e. not the mid-point or ordinal grade). Although using the percent scale is always preferable, using an ordinal scale for assessment of disease severity has its practical value with regard to implementation and feasibility. When the DSI is used, it is recommended that an amended 10% category scale should be chosen. This can provide accuracy for estimating the disease severity and optimal power for hypothesis testing. The second objective of this thesis is to provide an alternative for investigating estimates of pest damage severity. Traditionally, the manual approach uses grid paper to depict and measure the pest damaged area, but it could take relatively more time and labor. To provide a more economical measurement method, image analysis via ASSESS 2.0 can be used to measure damaged leaf area - the rice leaf folder as an example. The accuracy and precision of measurement by ASSESS 2.0 are determined with Lin's concordance correlation coefficient (ρc). This result shows that measurement of image analysis is highly consistent (ρc higher than 0.983) with the manual approach, which is regarded as the gold standard. In addition, the image analysis can reduce measurement time by at least 29%. Thus, image analysis via ASSESS 2.0 can be recommended as a tool for measuring leaf damaged area of rice leaf folder.