A study on giant panda recognition based on images of a large proportion of captive pandas

1. As a highly endangered species, the giant panda (panda) has attracted significant attention in the past decades. Considerable efforts have been put on panda conservation and reproduction, offering the promising outcome of maintaining the population size of pandas. To evaluate the effectiveness of conservation and management strategies, recognizing individual pandas is critical. However, it remains a challenging task because the existing methods, such as traditional tracking method, discrimination method based on footprint identification, and molecular biology method, are invasive, inaccurate, expensive, or challenging to perform. The advances of imaging technologies have led to the wide applications of digital images and videos in panda conservation and management, which makes it possible for individual panda recognition in a noninvasive manner by using image‐based panda face recognition method.
2. In recent years, deep learning has achieved great success in the field of computer vision and pattern recognition. For panda face recognition, a fully automatic deep learning algorithm which consists of a sequence of deep neural networks (DNNs) used for panda face detection, segmentation, alignment, and identity prediction is developed in this study. To develop and evaluate the algorithm, the largest panda image dataset containing 6,441 images from 218 different pandas, which is 39.78% of captive pandas in the world, is established.
3. The algorithm achieved 96.27% accuracy in panda recognition and 100% accuracy in detection.
4. This study shows that panda faces can be used for panda recognition. It enables the use of the cameras installed in their habitat for monitoring their population and behavior. This noninvasive approach is much more cost‐effective than the approaches used in the previous panda surveys.

     

Parameter estimation for the distribution of single cell lag times

In Quantitative Microbial Risk Assessment, it is vital to understand how lag times of individual cells are distributed over a bacterial population. Such identified distributions can be used to predict the time by which, in a growth-supporting environment, a few pathogenic cells can multiply to a poisoning concentration level.We model the lag time of a single cell, inoculated into a new environment, by the delay of the growth function characterizing the generated subpopulation. We introduce an easy-to-implement procedure, based on the method of moments, to estimate the parameters of the distribution of single cell lag times. The advantage of the method is especially apparent for cases where the initial number of cells is small and random, and the culture is detectable only in the exponential growth phase.     

Can transposable element copy number distribution parameters be estimated from natural populations of Drosophila melanogaster?

The copy frequency distribution of a transposable element family in a Drosophila melanogaster natural population is generally characterised by the values of the Charlesworths' model parameters α and β (Charlesworth & Charlesworth, 1983). The estimation of these parameters is made using the observed distribution of the occupied sites in a population sample. Several results have been interpreted as due either to the influence of stochastic factors or to deterministic factors (transposition, excision, selection…). The accuracy of this method was tested by estimations performed on samples from simulated populations. The results show that with the sample size usually used for natural population studies, the confidence intervals are too large to reasonably deduce either the element copy number distribution or the values of transposition and excision rate and selective coefficients.     

Cod Gadus morhua L. populations as behavioural units: inference from time series on juvenile abundance in the eastern Skagerrak

Abundance and distribution of cod Gadus morhua in various size intervals and age groups between 2000 and 2005 were followed in coastal trawl surveys. In spite of a reduction in fishing pressure in recent years and high cod recruitment in the Skagerrak region in 2001 and 2003, no recovery could be evidenced. The survey data clearly showed that low cod density areas were not recolonized, even though abundance of juvenile cod remained high for about a year after the recruitment episodes. Increased abundance of fish >400 mm total length was only discernible at some scattered locations where other studies also have suggested local populations still to be present. The intermittent high recruitment has been linked to an inflow of egg and larvae from the North Sea, a theory which also has gained support from genetic studies. It was thus argued that the disappearance of the juvenile cod from the inshore is an effect of a migratory behaviour; the fish of offshore origin eventually leave the coast for the open Skagerrak or the North Sea. These findings support a view on cod populations as essentially behavioural entities, whereas dispersal of early life stages may be less important as a structuring mechanism.     

Confidence Intervals for the Risk Ratio in Cohort Studies under Inverse Sampling

Three simple interval estimates for the risk ratio in inverse sampling are considered. The first two interval estimates are derived on the basis of Fieller's Theorem and the delta method with the logarithmic transformation, respectively. The third interval estimate is derived on the basis of an F-test statistic proposed by BENNETT (1981) for testing equal probabilities of a disease between two comparison groups when the disease is rare. To evaluate the performance of these three methods, a Monte Carlo simulation is used to compare the actual coverage probability with the nominal confidence level for each method and to estimate the expected length of the corresponding confidence interval in a variety of situations. On the basis of the results found in the simulation, we have concluded that the method with the logarithmic transformation is either equivalent to or better than the other two methods for all situations considered here.     

Does Mother Nature really prefer rare species or are log‐left‐skewed SADs a sampling artefact?

Intensively sampled species abundance distributions (SADs) show left‐skew on a log scale. That is, there are too many rare species to fit a lognormal distribution. I propose that this log‐left‐skew might be a sampling artefact. Monte Carlo simulations show that taking progressively larger samples from a log‐unskewed distribution (such as the lognormal) causes log‐skew to decrease asymptotically (move towards ?∞) until it reaches the level of the underlying distribution (zero in this case). In contrast, accumulating certain types of repeated small samples results in a log‐skew that becomes progressively more log‐left‐skewed to a level well beyond the underlying distribution. These repeated samples correspond to samples from the same site over many years or from many sites in 1 year. Data from empirical datasets show that log‐skew generally goes from positive (right‐skewed) to negative (left‐skewed) as the number of temporally or spatially replicated samples increases. This suggests caution when interpreting log‐left‐skew as a pattern that needs biological interpretation.     

Co‐occurrence patterns of trees along macro‐climatic gradients and their potential influence on the present and future distribution of Fagus sylvatica L.

Aim During recent and future climate change, shifts in large‐scale species ranges are expected due to the hypothesized major role of climatic factors in regulating species distributions. The stress‐gradient hypothesis suggests that biotic interactions may act as major constraints on species distributions under more favourable growing conditions, while climatic constraints may dominate under unfavourable conditions. We tested this hypothesis for one focal tree species having three major competitors using broad‐scale environmental data. We evaluated the variation of species co‐occurrence patterns in climate space and estimated the influence of these patterns on the distribution of the focal species for current and projected future climates. Location Europe. Methods We used ICP Forest Level 1 data as well as climatic, topographic and edaphic variables. First, correlations between the relative abundance of European beech (Fagus sylvatica) and three major competitor species (Picea abies, Pinus sylvestris and Quercus robur) were analysed in environmental space, and then projected to geographic space. Second, a sensitivity analysis was performed using generalized additive models (GAM) to evaluate where and how much the predicted F. sylvatica distribution varied under current and future climates if potential competitor species were included or excluded. We evaluated if these areas coincide with current species co‐occurrence patterns. Results Correlation analyses supported the stress‐gradient hypothesis: towards favourable growing conditions of F. sylvatica, its abundance was strongly linked to the abundance of its competitors, while this link weakened towards unfavourable growing conditions, with stronger correlations in the south and at low elevations than in the north and at high elevations. The sensitivity analysis showed a potential spatial segregation of species with changing climate and a pronounced shift of zones where co‐occurrence patterns may play a major role. Main conclusions Our results demonstrate the importance of species co‐occurrence patterns for calibrating improved species distribution models for use in projections of climate effects. The correlation approach is able to localize European areas where inclusion of biotic predictors is effective. The climate‐induced spatial segregation of the major tree species could have ecological and economic consequences.     

Comparing Stochastically Restricted Linear Estimators in a Regression Model

Conditions for superiority of the minimum dispersion estimator over another with respect to the covariance matrix are derived when the vector parameter of a regression model is subject to competing stochastic restrictions. The restrictions may also consist both of a deterministic part and a stochastic part.     

COMPARISON OF VERTICAL AERIAL PHOTOGRAPHIC AND GROUND CENSUSES OF STELLER SEA LIONS AT AÑO NUEVO ISLAND, JULY 1990-1993

Counts of Steller sea lion ( Eumetopias jubatus ) pups and non-pups (adults and juveniles) from aerial photographs of rookeries at Año Nuevo Island between 1990 and 1993 were significantly higher than those made on the ground. Based on regression of natural logs of photographic counts versus year, the number of pups declined at a rate of −0.099yr while non-pup numbers declined at −0.315/yr. Examination of ground count data for the same period revealed a significant decline in non-pups (−0.139/yr), but no trend was detected in the ground counts of pups. The regression coefficients from photographic and ground counts of non-pups did not differ significantly. Power analyses using the program TRENDS indicated that detectable rates of change in abundance from four annual surveys were much lower for counts of pups than counts of non-pups where sampling precision was based on fits to linear models.