# 研究計畫

## 面試準備

### Challenges

- OI值的干擾因子是指??
- ie. Homerange大的物種一般OI會高
- 群居動物也很困擾
- 取樣時間間距?
- 季節影響
- 相機怎麼擺?
- 毛髮要多久check一次?
- 石芝菁做過測試，大約30天還有86%成功率，60天後剩53%
- 糞便的話一週內有約38%成功率
- 遺傳研究的貢獻
- YNP 139(84M+54F) / DSY 4M+4F
- 兩地有顯著遺傳分化 degree of genetic differentiation (F ST = 0.1675, P < 0.01)
- Program STRUCTURE to examine genetic clustering
- homogeneous natureof DSY genetic diversity vs. higher genetic diversity of FBB in YNP
- 阻隔: 地形(1000-2500)；Road; Hunting
- DSY可能是較小的亞族群，兩樣區間的遺傳資訊值得注意，中間有沒有明確阻隔? 有沒有廊道存在?
- Dave 有10篇做認真的abundance & density, 只有2篇認真做population trend，是哪兩篇?
- 泰國的研究
- 每個樣區大約20個樣點，每次做3個sampling occasions，每次3週
- 經費預算
- YNP和DSY樣區，各半年Field work、半年lab work，每樣區一年預算:
- 人事
- 自己的薪水5w*12M= 60w
- 臨時助理3.5w*6M*2人 = 42w
- 臨時工 10w
- 野外調查雜支45w
- 遺傳分析預算45w
- 設備成本: 1.5W x 55 x 3 = 247.5 w
- ==>買相機165台大約250W
- 其他計畫費用大約200W/1年
- SCR的assumptions
1. Demographic closure: there is no recruitment or mortality in the sampled population
2. Geographic closure: there is no emigration or immigration from the state-space. However, SECR allows for “temporary” movements around the state-space
3. Activity centers are randomly disbursed
4. Detection is a function of distance: the encounter probability declines as a function of distance from an individual’s activity center
5. Independence of encounters among individuals
6. Independence of encounters of the same individual
- A challenge in the application of SCR methods is how to efficiently sample areas to allow the most individuals to be detected while also ensuring enough spatial recaptures to estimate the detection kernel (this requires multiple detectors within an individual home range)
- The challenge with applying this type of estimator is that it assumes detection probabilities are similar for all bears within a sampling occasion
- 一定要做simulation，看不同sampling space的relative standard error(相對標準誤，就是%CV)

## 資料蒐集

### Dave
- The opinions of experts and local people, now heavily relied upon for population assessments, are not reliable or sensitive enough for monitoring. Quantitative population assessments are desirable to direct conservation actions toward the most perilous situations, and provide a means to gauge the effectiveness of conservation actions.
- only 10 over a 20 year period used a rigorous method to estimate abundance or density!!

#### Introduction
Data on actual rates of decline would help to direct conservation actions to those species and locations most in need.

Designing and testing methods of population monitoring was chosen by respondents as the top priority for BSG action, followed closely by a prioritization of where conservation actions should be focused.

狹義的監測vs.廣義的監測
Although "monitoring" is strictly defined as tracking changes (or stability) of a population through time (Moussy et al., 2021), we consider monitoring more broadly to include any population assessment (e.g., a single point in time) that has the potential to be used as a baseline for a future comparison, or in a spatial comparison among several populations.

**Crudge et al. 2019 Conservation Action Plan fo Sun Bears**

蒐集來的文獻歸6類:
(1) presence of bears in a specific area, (2) distribution, (3) occupancy, (4) population index, (5) population estimate, and/or (6) population trend (direction or rate of change in distribution and numbers).

#### Results:
吊嘎的proposal可以參考
Burton et al. (2018) was the only occupancy estimate that used genetic sampling (of brown bears), which they combined with camera trapping, reasoning that some bears might visit a site and not leave a sample on barbed-wire hair snares.

#### Discussion
Nawaz2008很厲害可以參考，pakistan, collar

辨別個體在亞洲比較困難，Hair在B密度低時成本高。相機也是，尤其當需要B站起來
糞便搜集有時更輕鬆，但ID rate比較低，可能和tropical有關

集中資源!!!
The limitations of each method and the resources required to achieve desired outcomes (objectives) call for careful and coordinated study designs, so as not to waste limited conservation resources on projects that are unlikely to provide useful insights (Buxton et al., 2021).

Observable population responses to conservation actions can then serve to document what works best and improve our use of resources to inform actions most beneficial to bear populations.

### McShea

#### Highlight

- Changes in the distribution of a species can be used to as a metric of conservation status and to identify the loss or gain of isolated populations.
- This mapping process is a primary tool of the IUCN SSC Red List assessment.
- However, modeling the distribution for some species is difficult due to low levels of sampling, low detectability
- Presently the range of all bear species in Asia is delineated primarily using expert opinion.
- BSG **create a rule set to combine current detections with known habitat and home range preferences to improve distribution maps for the purpose of conservation metrics and monitoring. **
- We present a list of recommended actions for improving distribution mapping for bears in Asia

#### Introduction
- documenting the species ’ extent enables the identification and monitoring of isolated populations and can focus conservation efforts on priority areas such as fractures between populations and the current edges of the distribution.
- Cons: One danger in relying on changes in distribution is that they may reflect improvements or deteriorations in the data or methodology, obscuring genuine population changes.
- Asian bears potential range maps using:
- detection/non-detection records
- maximum-entropy methods with presence-only data
- For large areas is problematic
1. Uneven and potentially biased of presnece data, 調查常常聚焦在PA, 過往常用signs, interview等方法
2. observations must be extrapolated to **predict occurrence**, **not potential range**
Model仰賴habitat資訊，可以做出potential range, 但亞洲人類行為(狩獵)嚴重影響B分布
- "Possibly extent" . Here we argue that this range delineation, while useful for pointing to regions where searches for bears should be conducted, is not useful for assessing changes in distribution.
- **To use range comparisons over time for population monitoring, we need a standardized approach to mapping the occupied range, not a potential range defined by habitat**
- **further, the mapping process needs to be sensitive to small changes in distribution.**
- **Here we test a new observation-based process for Asian bears, with a goal of reducing dependency on expert opinions when creating global range maps and measuring a species’ trajectory.**
1. identify suitable habitat
2. acquire and evaluate recent observations
3. use a rule-based process to define the range
4. refine the map with expert opinion

### Fuller

#### Highlight
- Accelerating declines and small population size among Asian bear populations necessitate survey methods that can achieve objectives of monitoring programs, including assessing the status of populations (system state), and providing information on system response to management actions (Yoccoz et al., 2001)
- Designing a population monitoring program for Asian bears presents **challenges** associated with their *low densities and detectability, generally large home ranges, and logistical or resource constraints*.
- The use of an occupancy-based method to monitor bear populations can be appropriate **under certain conditions** given the mechanistic relationship between occupancy and abundance.
- Our simulations detected population declines of 50% with high power ( > 0.80) and low false positive rates (FPR: incorrectly detecting a decline) ( < 0.10) when detectors were spaced at > 0.67 times the home range diameter (home-range spacing ratio: HRSR

#### Introduction

- The IUCN Species Survival Commission describes an **“assess, plan, act”** species conservation cycle for effective conservation (Rodríguez, 2021).

- Occupancy is useful to monitor populations across large spatial extents to understand how a species is distributed in space, resulting in an estimate of the true probability that a site is occupied by the species
- Occupancy models allow for imperfect detection: instead of assuming that non-detections represent true absence, occupancy models use repeat sampling, yielding detections or non-detections at each site on each survey, to jointly estimate the occupancy and detection probability (*p*)
- Royle and Nichols (2003) developed an extension of occupancy models (Royle-Nichols model: R-N) that incorporate abundance-induced heterogeneity to estimate the expected site-level abundance ( λ ) from the same detection-nondetection data used for occupancy analyses. Site-level abundances are difficult to interpret when the effective sampling area is unknown or when detection of individuals is not independent
- **Several monitoring objectives can be addressed with occupancy modeling, including geographic range, rate of change in occupancy over time, habitat relationships, and multiple species interactions**
- IUCN Red List of Threatened species includes declines in area of occupancy as part of criteria
- Under particular circumstances, trends in occupancy can be strongly and positively related to abundance
- ** occupancy as an index for abundance relies on the mechanistic relationship between occupancy and abundanc**
- when the sampling unit is a point (and the associated detection area around the point), cells are relevant only to space sampling units sufficiently far apart to achieve independence between sites.
- For low densities and grid cells smaller than the home range size, it is clear that this closure assumption would often be violated.
- When grain size is small relative to home range size using areal sampling, ‘use ’ (i.e., proportion of area occupied) should not be used as an index of abundance
- The most basic occupancy model estimates the expected site-level occupancy probability ( *ψ* ) and the probability of detecting the species given that the species occupies a site (*p*).
- 實際操作
- Areal sampling: individuals records, sign surveys on transects.
- Point sampling: cameras
- trade-off between sampling more sites and applying more effort per site
- **general guidance for rare species**, which would apply to Asian bears, is to survey more sites less intensively.
- Aims of our simulation study
- simulations are important in designing a survey to achieve the desired power to detect a particular effect size (e.g., decline in occupancy between survey periods given an actual proportional decline in abundance).

### Proctor

#### Highlight

- Since an objective of monitoring is to detect declines early enough for management action to impact the population, it is useful to review the utility of available methods to meet management objectives.
- Ideally, monitoring would be accomplished via periodic data-based estimates of population size or density.
- Abundance -->
- typically require repeated sampling of a population containing individuals that can be identified to inform a mark – recapture estimate with open or closed models
- When repeated over years, abundance estimates can estimate trend, but detecting significant changes in population size can be challenging
- new methods to derive abundance or density estimates from sampling when animals are not individually identifiable
- Geographic distribution and occupancy can also be estimated and monitored over time
- Distribution estimates utilize verifiable evidence of species ’ presence over a large sampling area
- Occupancy estimates also use presence data but over a systematic and repeated sampling to estimate detection probability.
- To ensure a monitoring program is successful at meeting its objectives (Legg and Nagy, 2006), field methods, analyses, field team expertize, and finances must be balanced within a viable study plan.

#### Camera
- 辨識胸斑
- 昂貴且吃力，但回報豐富
- 現在有一些軟體可幫忙，主要是machine learning, AI
- 機器學習的一種形式- 深度卷積神經網路Deep convolutional neural networks (CNNs)
- Camera data can also illuminate other aspects of bear ecology and behavior including scent marking, rubbing behavior, scavenging, predation , sex ratio, reproduction, and activity patterns, all of which may be useful in providing context for a monitoring program.
- In summary, cameras traps can be used effectively to achieve all of the monitoring objectives covered in this paper.

#### DNA hair
- nDNA fingerprinting
- a permanent mark, lasting from birth to even after death
- Identifying an effective lure for local bears may benefit from local knowledge on what bears eat and are attracted to.
- the height of the wire is a critical element in the design of hair traps.
- Further challenges associated with this assumption can occur due to degradation of samples by sun and rain as they await collection in the field, leading to failed or partial genotyping; collection frequencies should be considered and even tested with this in mind, particularly in warm moist environments
- compared to sampling DNA from feces (see next section), DNA from hair is usually less degraded
- This technique can be especially advantageous for assessing and monitoring the conservation status of small, isolated populations
- Downstream analyses from DNA analyses have also become prolific and include fragmentation and connectivity, dispersal , HBC.

#### DNA scat
- Scat-sourced mitochondrial DNA (mtDNA) is often used for species identification because it is more abundant (many copies)
- low levels of template (nDNA) may yield unreliable individual identification
- the basic assumption of unambiguous individual identification is more difficult to meet with fecal than hair-derived DNA.

#### Telemetry
- sample of female bears is monitored and that their observed age-specific reproduction and survival are representative of the population
- A principal challenge of this method is capturing and radio-collaring a sufficient sample of adult female bears and monitoring them over a long enough period of time (e.g., ≈ 10 adult females over a decade)
- Estimating trend via vital rates is only one of many reasons that bears are captured and radio-collared.
- May be particularly useful in helping to inform density estimates derived from unmarked camera trapping data

### Dana

- "Effective" monitoring is critical to measure and direct conservation efforts
- 2 metrics: population abundance(local density) or groth(λ)
- Methology
- Spatially explict capture-recapture (SCR), the current gold standard for density estimation.
- Open population SCR(OPSCR), monitor chmanges in density overtime.
Reliant on "marked" individuals, are currentyl the only methods with enough power to reliably detect even moderate to major (20-80%) decline.
- 佈署研究以前，建議...
- conducting simulations, power analyses, and assumption checks
- assess precision by reporting dispersion measures such as coefficients of variation
- 這些原則對密度低、分布廣的物種都有效

#### Introduction

- Conservationists frequently must prioritize **actions**
- Rely on monitoring data to indicate and evaluate
- when and where action is necessary
- outcome
- conservationists may need to know if a population is in decline or recovering, identify areas that contain sustainable populations, or evaluate if protection of a specific area resulted in increases in the bear population.
- --> monitor population change
1. Why is abundance or density the preferred parameter for monitoring population trends?
- Estimating how many individuals are in a population or area at a certain point in time (abundance, N)
2. Overview of density estimation methods for Asian bear populations

#### SCR
- Optimal placement of traps can be approximated by a design where the number of individuals (n) sampled equals the number of total recaptures (r) with precision of estimates
- Ngoprasert et al. (2012), who used a 1.5 – 2 km trap spacing, but learned subsequently from simulations that a 4-km spacing (and hence larger study area) would have improved the precision of their density estimates.
- Simulation studies are critical in assessing designs and easily implemented using secrdesign or writing functions to employ in oSCR R package

#### OPSCR
- Conservationists are often more interested in determining the direction and rate of population change (hereafter trend) than simply population size.
- Open, non-spatial capture-recapture (CR) population models (which allow births, deaths, immigration, and emigration) have been used to estimate trends in bear demography and relate these changes to environmental factors
- (OPSCR) models include, at a minimum, submodels for recruitment, mortality, and between season movement.
- Instead of stringing together static snapshots of populations with non-spatial SCR models, OPSCR models allow the combined estimation of spatial population dynamics and can therefore be used to make population forecasts and project outcomes of different scenarios
- OPSCR models require multiple seasons of data (usually multiple years for large carnivores such as bears

#### No marks methods
- Small violations of these assumptions (detailed extensively in Gilbert et al., 2021) can result in strong biases and reduced precision, rendering the resulting estimates useless at best, and misleading at worst
- The greatest hindrance to the use of camera traps to estimate density using unmarked estimation methods is the reduced precision typical of estimates from most empirical studies (Cappelle et al., 2021; Palencia et al., 2021) compared to accepted standards for wildlife management (CV ≤ 0.20; Williams et al., 2002) and those achieved by empirical SCR studies
- While telemetry studies are expensive, invasive, and difficult to implement, many methods of monitoring may benefit from home range estimates, movement rates, vital rates, and other information that can be gathered from collared bears.

#### USCR
- USCR+ 有加其他資訊e.g., telemetry data
- Spatial mark – resight (SMR), 只有部分個體辨識
- If sufficient ancillary data are available (e.g., kernels estimated from telemetry data or a sample of identifiable bears), USCR + could be an effective method to address several monitoring needs

#### Tourani 2020做了一個Hierarchical modeling，整合individual identitical / non-identitical的資料