Abundance Estimation of Himalayan Lynx Through Multiple Modelling Approach

Abstract

Abundance is the key parameter of concern to wildlife management and conservation. Different filed and analytical methods are in practice to obtain and analyse data for wildlife abundance estimation. Owing to their low cost, easy management and production oflarge data sets, camera traps have been the tool of choice in field surveys. Data from camera traps have widely been used to estimate population parameters for marked animals using Capture-Recapture approach. Since this approach requires identification of individuals, which limits its usefulness as most of the photo-captured species either totally lack identifiable markings or have unclear markings. Several analytical approaches are emerging to address this limitation so that the population parameters of unmarked animals are estimated. This study tried to compare abundance estimates for the different urunarked approaches using Himalayan lynx as a case study, in the Chitral Gol National Park (CGNP), Khyber Pakhtunkhwa, Pakistan. The species has elusive and nocturnal behaviour and occurs at the low densities. Out of the total 103 grids of 1 km2 over the 77.5 km2 study area, cameras were installed at 30 grids. Due to some technical errors 5 cameras were excluded from the analysis. A database for the pictures :6.'om the remaining 25 cameras were created, to simplify data retrieval and analysis. The modelling approaches used in this study included N -Mixture models, Random Encounter Modelling (REM), Space and Time models and Spatial Count (SC) models. The operational 25 cameras resulted in 1125 camera trap days. Lynx was captured 16 times at 6 trap sites. The N-Mixture model with the Poisson approach estimated abundance at 7.88 individuals for the CGNP. REM estimated an abundance of 2.69 individuals with a density of 3.47/100km2. The abundance estimates for the three approaches of Space and Time Modelling were 5.74±1.78(SE), and 12.52±3.76(SE), 30.5±15.5(SE), for Time-To-Event (TIE), Space-To-Event (STE) and Instantaneous Sampling Estimator (ISE), respectively. The SC model reported an abundance of 6.19±4.68(SD) with density of 811 00ktn2. The population estimates derived from different models ranged between 3-30 lynx in CGNP, which is a huge variation and limits applicability of such results for the management purpose. However, estimates of four models (N-Mixture, TTE, STE, and SC) are closer 6-12 individuals, and seem more realistic for the study area. REM appears to underestimate the population, while ISE has overestimated. We believe these analytical approaches has a promise for investigating populations of umnarked animals, however with small data sets it is difficult to get agreement between them. Increasing sample size and combining data from multiple sources can potentially solve this issue.