To identify connectivity among Sumatran tiger populations in Riau Province, Sumatra, I have identified three components, which together will provide a complete view of tiger connectivity; genetic connectivity, structural connectivity and functional connectivity, or animal movement.
Objective 1: Identify whether there is genetic connectivity between tiger populations in central Sumatra
To determine the extent to which tiger populations in Riau are genetically connected or fragmented, I will first conduct scat surveys on foot throughout the province. After establishing a protocol with field teams during the first field season, WWF-Indonesia field teams will collect scat continuously and store samples in DET buffer. Surveys will be conducted opportunistically along trails and near other known areas of tiger use, and within different habitat types. Scat will be collected in all tiger habitat within Riau, assuming appropriate park entrance permits are obtained. Teams will collect scat opportunistically along trails and roads within protected areas throughout the year to increase sample size. I will also attempt to collect scat samples along roads and trails within various plantations in Riau. If collecting tiger scat on foot and by car proves productive and a sufficient number of samples throughout the province are located this way, we will repeat surveys using this method the next year. If collecting samples on foot does not yield samples due to their low detectability by sight, I will enlist the help of a scat detection dog the following year.
I will collect samples from the outside of the scat (Michalski et al. 2011, Stenglein et al. 2010) and store in DET buffer. Other studies have shown DET buffer to have higher amplification success (Wultsch 2013). I will follow the QIAamp DNA Stool Mini Kit protocol (Qiagen, Inc., Valencia, CA, US) with slight modifications as required, to extract DNA from the fecal samples. To verify species of origin of the fecal samples, I will use tiger specific mitochondrial DNA primers such as Tig490, Tig509, and Tig678 (Luo et al. 2004, Mukherjee et al. 2007). A set of microsatellite primers will be used to identify individuals (Williamson et al. 2002, Mondol et al. 2009). The 15mL PCR mix will likely contain 3mL MgCl2, 1.5 mL buffer, 1.2 mL dNTP, 6.32mL H2, 1.5mL primers, 0.3mg BSA, 0.07mL Taq and 1.8mL DNA, but may change due to scat quality, availability of materials, etc. The reaction cycle may be conducted as follows: 3 min at 95o C for denaturation, 35 cycles of 95 o C for 30s, 50 o C for 30s, 72 o C for 30s, and followed by 72 o C for 10min (Williamson et al. 2002). Primers will be fluorescently labeled and run in a multiplex approach to eliminate false negatives (Mukherjee et al. 2007). In the case of smaller felid scat collected opportunistically, the same procedure will be used, but species specific primers will be used, where available. DNA will be sequenced using an automated sequencer (Applied Biosystems 377) and analyzed using Sequencher 5.0 and compared with sequences in GeneBank.
Genetic connectivity will then be determined using multiple methods. First, I will analyze the microsatellite data using the Bayesian clustering methods in Programs Structure 2.x (Pritchard et al. 2000), Geneland (Guillot et al. 2005), BAPS (Corander et al. 2003) and/or TESS (Chen et al. 2007) to determine the most likely number of groups (k) within the population. After defining populations, I will calculate Fst (Wright 1965) to evaluate the degree of genetic differentiation. I will also use assignment tests implemented in GENECLASS and BAYESASS (Wilson & Rannala 2003) to detect migrants between populations.
Objective 2: Model land cover change over time and create predictive future land cover scenarios.
Using land cover datasets created using 30m Landsat imagery by WWF-Indonesia for at least 2007 and 2010 (and possibly dating as far back as 1990) as a basis and using Idrisi’s Land Change Modeler (LCM) (Eastman 2012) I will quantify how land cover in Riau has changed and predict how it is likely to continue to change. Dynamic drivers of land cover change such as roads and settlements will be included as well. The LCM allows the user to select between a neural network or a logistic regression algorithm to model land cover change (Eastman 2012). A matrix of transition probabilities is created using the two base land cover data sets and the LCM uses these probabilities to predict how the land cover will change in a future time step. Transition probabilities may also be altered by the user to determine how different probabilities will impact the landscape. By doing so, I will predict land cover change under a business-as-usual scenario at current development rates and a ‘best-case’ scenario, in which development and encroachment rates are slowed and under a ‘worst-case’ scenario where development and encroachment rates will be increased. I will create transition potential maps for land cover classes using the previous land cover (Fuller et al. 2011, Joshi et al. 2011, Perez-Vega 2011). I will create both hard and soft predictions – that is maps of predicted future land cover and predictions of land cover vulnerability to change.
After creating land cover change prediction maps, I will calculate the area of each land cover class and compare the current and various predicted. I will also convert suitable tiger habitat, i.e. natural forest, to polygons using ArcGIS 10.x and calculate several more landscape metrics to gain a more complete understanding of how the landscape will change. Metrics such as perimeter/area ratio of habitat patches, number of patches across the landscape, mean distance between patches and patch edge/core ratio will be calculated for each landscape map; current land cover and predicted land cover. Fragstats (McGarigal et al. 2012) or similar landscape analysis software may be used to aid in this process. By comparing landscape connectivity change through time, I will identify corridors which are present now but are at high risk of conversion to agriculture in the future. Removing such corridors could potentially have an effect on tiger persistence in the landscape into the future. I will create maps depicting such areas that are at high risk of conversion.
Objective 3: Use fine scale movement data to determine if/how tigers use acacia, eucalyptus, palm oil plantations, logging concessions and how they traverse the landscape protected areas
Capture Protocol – To determine if tigers from the remaining intact portion of the park travel into the human dominated areas of the park and to determine what drives these movements I will place GPS satellite collars on tigers. Before setting traps, collars will be placed in various habitats for 24 hours to determine location error. Box traps, where possible will be used to trap tigers and foot snares (Goodrich et al. 2001) will be used elsewhere. Traps placed in known tiger paths in Tesso Nilo National Park during the first phase of the study. Additional sampling may take place in other protected areas during a later phase of this study. Dairen Simpson, a professional carnivore trapper, will construct foot snares in locations away from cliffs, ledges or open water. Trapping will occur in the dry season to avoid logistic difficulties in the wet season. A trap transmitter will notify us of a triggered trap, and the team will be dispatched to the tiger immediately. An Indonesian veterinarian will be employed to safely execute trapping and handling.
After a tiger has been trapped, we will approach slowly and under cover to the extent the terrain allows, to minimize a stress response from the animal. We will use a combination of ketamine (2.2-2.5 mg/kg) and xylazine (0.8-1.0mg/kg), depending on body size and excitement level (Spelman, 1998) to tranquilize and anesthetize the animal using veterinary projector and dart to the shoulder or hindquarter, whichever is visible and within darting range first. When we approach after adequate induction time (12 minutes) and the animal shows signs of consciousness, we will administer another dose of ketamine (2.2-2.5mg/kg). If the animal does not show signs of consciousness upon distant approach, we will then approach the animal, prodding it with a long probe at the rump first, then the shoulder and neck to make sure the animal is completely immobilized. Once the tiger is immobilized we will first clear airways if necessary, pull the tongue forward, ensure breathing and then remove the foothold trap, cleaning any injuries with sterile alcohol. A sterile ophthalmic lubricant ointment such as Puralube® Vet Ointment will be placed in each eye and we will use a clean cloth to cover the eyes. If not already in left lateral recumbancy, we will reposition the animal. We will check and record vital signs (mucus membrane color, respiratory rate, heart rate, capillary refill time and rectal temperature) and check reflexes (blink reflex, pinnal reflex, tail tone, jaw tone, toe pinch response) every 3 minutes. The animal will also be observed for oral secretions, regurgitation of stomach contents and any signs of gas retention in the abdomen.
To remove the dart from the animal, we will use sterile scalpel to cut a 3/8” nick around the dart so as not to pull off excess tissue. With forceps, any foreign bodies will be removed from the wound and we will then flush the wound with nolvalsan or betadine solution. We will fill the wound with antibiotic cream such as Neosporin, Nolvalsan or Nitrofurasone. We will also place a fly repellent ointment on the wound and in a ten inch diameter around the wound to prevent infestation. The same procedure will be followed in the case of other small open wounds.
We will collect hair samples and blood samples for genetic analysis and comparison with fecal samples collected in the field. One fecal sample will be stored in 95% alcohol and one will be stored in DET. After fitting the GPS collar to the animal, we will administer 0.125mg/kg yohimbine intravenously to reverse the effects of xylazine. We will then retreat to a vehicle a safe distance away to ensure a safe and complete recovery and we will record time to head lifting, sternal recumbancy and standing recovery.
If the animal ceases breathing, air will be administered via an AMBU bag and intubation after applying 2% lidocaine spray on the laryngeal folds. We will administer Dopram (doxapram HCl) at 0.5-1mg/kg intravenously repeated as necessary every five minutes, with a maximum administered amount of 2mg/kg. If the animal experiences asystole a dose of atropine 0.5-5mL will be given to the heart with a 22 gauge needle and repeated every five minutes if necessary. In case of seizure, we will administer 2.5-5mg/kg midazolam intramuscularly, depending again on body size. Although not expected, if a life threatening injury is sustained during capture or if the animal does not regain heartbeat, we will administer 100mg/kg of sodium pentobarbital intravenously and monitor and record vital signs for ten minutes.
We will use foot snares designed to exclude smaller carnivores, but in the case of box traps, if we do catch a smaller carnivore in the trap, we will immobilize it using a 5:1 ketamine to xylazine mixture, with dosage varying depending on body size (i.e. 5mg/kg ketamine + 1mg/kg xylazine, for calm felids or 10mg/kg ketamine + 2mg/kg xylazine – 20mg/kg ketamine + 4 mg/kg xylazine for felids >35 pounds or are excited). If a smaller felid stops breathing, we will administer 5.5-11mg/kg of Dopram intravenously every 15 minutes, if necessary. The same protocol will be followed as for tigers.
Spatial Analysis – The GPS satellite collars will be placed in different land cover types for at least 24 hours to assess their spatial accuracy. For each land cover type, the mean location error will be calculated (Horne 2007). GPS data will be collected at regular intervals using African Wildlife Tracking GPS collars, 5 hours apart for the majority of the year. For one month during both rainy and dry seasons, the GPS collars will be remotely changed to collect fixes more frequently as collar batteries allow This finer scale movement data will be used to build Brownian Bridge (Horne et al. 2007) movement models (BBMM) for each tiger to examine the individual paths tigers use, their time spent in each habitat type, and to determine whether roads and other anthropogenic features affect movements. Movement models will be overlaid with current land cover data in ArcGIS 10.x. To identify which land cover types tigers move through and are likely to spend time in, I will regress the raster cell based BBMM utilization distribution probabilities on landscape covariates such as land cover, distance to water, distance to roads and distance to elevation (Sunarto et al. 2011, Horne et al. 2007, Lewis et al. 2011). Using this method, we may be able to predict which areas tigers are more or less likely to travel through across the landscape. If high resolution satellite imagery is available, we may be able to correlate movement probabilities with different stages of plantation growth, as well.
If movement data is limited, I will create a habitat suitability model for tigers in Riau using Maxent. Inputs to the model will include tiger scat and track GPS locations serving as the ‘tiger presence’ locations, and landscape covariates such as distance to road, distance to water and forest cover. Covariates included in the model will be selected based on Sunarto et al. (2011) findings. I will use 25% of the tiger presence locations for model testing, and retain 75% of the locations for model training. The output from Maxent will then be converted to a raster grid in ArcGIS. This habitat suitability model will then be used to model connectivity using Circuitscape (McRae and Shah 2009). Protected areas within the landscape will serve as the ‘habitat patches’ for corridor modeling purposes; connectivity between these patches will be modeled. I will also model future habitat suitability for tigers in Riau using the ‘projection’ settings in Maxent. Input will include predicted land cover resulting from the Land Change Modeler (see Objective 2).