2020-08-012024-05-18https://scholars.lib.ntu.edu.tw/handle/123456789/702472摘要：植物多樣性評估系統應具備精準度之特性，更需建立於機率理論。此系統應提供植物物種豐富度和物種多樣性空間分佈兩項關鍵資訊以助於永續森林經營。本計畫總體目標為結合變動機率取樣法及在地專家知識以開發由在地知識引導之植物多樣性快速評估系統。當地專家先快速評估所有林分之植物多樣性，建立其多樣性共變量，再以不等機率與共變量成正比，選取樣點進行物種普查，此方法應能有效評估植物多樣性。本計畫探討之重點變動機率取樣法為：List、3P及LPM 取樣法。本計畫之三個具體目標將於三年內完成。計畫執行期間將建立九個資料庫並模擬超過17,000個組合，其中重點在於模擬在地專家快速評估方法。第一年目標為運用模擬結果及九個物種豐富度推估模式評定三個取樣法於估計植物物種數量之精準度；第二年目標為運用模擬結果及四個克利金法評定三個取樣法於估計植物物種多樣性空間分佈之精準度；第三年目標為透過現場試驗了解物種知識變化及三個取樣法之間的連接性。試驗地設立於福山森林動態樣區內。本計畫開發之評估系統可結合當地原住民社區共同進行植物多樣性評估，使得在地知識得以應用與保存，並增加社區就業機會及參與森林管理過程。<br> Abstract: Assessment of plant diversity is important because it benefits human society. A plant diversity inventory system should be cost-effective, precise, accurate, and probability-based for practical and scientific justification. It needs to generate two pieces of critical information for sustainable forestry: species richness and spatial distribution of species diversity. Species richness is internationally recognized as an indicator of sustainability in forest planning. Mapping spatial distribution of species diversity helps decision makers in land use planning. Past research has seldom explored the question of which sampling designs lead to more accurate and precise estimation of species richness and mapping distribution of species diversity. Variable probability sampling designs with sample selection probability proportional to a covariate are theoretically more efficient. An excellent choice of covariate is rapid assessment of species richness and abundance by experts from local communities, who are familiar with their forests. Using human judgement in sampling is commonly regarded as scientifically unjustifiable. However, local expert knowledge is increasingly recognized to benefit diversity inventory. The overarching goal of this project is to develop an innovative sampling strategy based on variable probability sampling with rapid assessment by local experts. A forest is first split into many small parcels. A local expert rapidly assess diversity in all parcels. Parcels are selected with probability proportional to rapidly assessed diversity for census. The variable probability designs of interest are: list sampling (LIST), 3P sampling (3P), and Local Pivoted Method (LPM). This project has three specific objectives to be achieved over three years. Objective of Year 1 is to assess accuracy and precision of LIST, 3P, and LPM in estimating species richness in a forest. A simulation study with nine datasets is set up. Simulated factors include rapid scan approach, species knowledge, covariate, sampling design, and sample size with a total of more than 17,000 combinations. Nine species richness estimators are applied to predict species richness and assess performances of sampling designs. Objective of Year 2 is to assess accuracy and precision of LIST, 3P, and LPM in mapping spatial distribution of species diversity. Simulated combinations from Year 1 are used. Four kriging methods are applied to map distribution of species diversity and assess performances of the sampling designs. Objective of Year 3 is to assess performances of LIST, 3P, and LPM under field conditions. A field trial is conducted at the 25-ha Fushan Forest Dynamics Plot. Ten students are recruited to test whether increasing knowledge in species leads to more efficient sampling of species diversity. Industries and small woodlot owners can benefit from the proposed variable probability sampling designs with rapid assessment by local experts. The inventory system can help stakeholders meeting conservation requirements of a certification scheme to access premium markets. Moreover, it is expected to lower cost of diversity assessment. Other benefits to local communities include increasing values of local knowledge, employment opportunities, and direct involvement in forest management activities.生物多樣性評估森林資源調查森林取樣法在地專家知識取樣原理共變量取樣法物種豐富度永續林業不等機率取樣法。biodiversity assessmentforest resource inventoryforest samplinglocal expert knowledgesampling theorysampling with covariatesspecies richnesssustainable forestryunequal probability sampling.