Guam Coconut Rhinoceros Project Technical Report

Netted Panel Traps to Test if CRB are Deflected

Aubrey Moore, Roland Quitugua and Ian Iriarte

Summary

It has been hypothesised that CRB adults are sometimes physically deflected when arriving at panel traps. This hypthesis was tested in a field trial by wrapping alternate traps in tekken. If beetles are bouncing off traps, we would expect to see more captured in those wrapped in tekken. However, we observed no significant difference in trap capture rate. Unwrapped traps caught 0.30 beetles per trap-day, wrapped traps caught 0.26 beetles per trap-day (p = 0.56; permutation test).

The beetles caught in traps wrapped in tekken were located as follows:

  • 34 entangled in tekken on side of net
  • 31 entangled in tekken covering the cone at the top of the trap
  • 8 'at large' in the receptacle at the botton of the trap
  • 3 entangled in tekken covering the funnel at the top of the trap

Methods

Two transects of panel traps were established in southern Guam. Alternate traps were wrapped in tekken. Traps were visited weekly and captured beetles were counted. For traps wrapped in tekken, the location of each beetle was recorded. During each visit, traps were moved to the next station within each transect.

Analysis

Data were stored in a CSV file, 'trapCatch.csv', and analysed using the IPython notebook, 'Netted Panel Traps Experiment to See if CRB are Deflected'.

In [1]:
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

Get data from CSV file

In [2]:
# netted=1: panel trap wrapped in tekken
# netted=0: panel trap not wrapped

df = pd.DataFrame.from_csv('trapCatch.csv', parse_dates=[1])
df['trap_rate'] = df['beetles']/df['days']
df.tail()
Out[2]:
startDate endDate days netted beetles notes net_top net_side net_bottom in_trap trap_rate
trapSite
P2 2015-04-17 2015-05-24 7 1 3 NaN 1 1 NaN 1 0.428571
P3 2015-04-17 2015-05-24 7 0 4 NaN NaN NaN NaN NaN 0.571429
P4 2015-04-17 2015-05-24 7 1 1 NaN NaN 1 NaN NaN 0.142857
P5 2015-04-17 2015-05-24 7 0 1 NaN NaN NaN NaN NaN 0.142857
P6 2015-04-17 2015-05-24 7 1 7 NaN 6 1 NaN NaN 1.000000

Apply permutation test for significance of difference between means

mean1 = mean daily trap rate for unwrapped panel traps
mean2 = mean daily trap rate for panel traps wrapped in tekken

In [3]:
class permtest:
    
    def __init__(self, xs, ys, ntrials=10000):
        """ xs an ys are numpy arrays
        """
        self.ntrials = ntrials
        self.xs = xs
        self.ys = ys
        self.meanxs = np.mean(self.xs)
        self.meanys = np.mean(self.ys)
        self.diff = np.abs(self.meanxs-self.meanys) 
        self.simdiff = [] 
        self.estimate_p()
        self.plot()
        
    def estimate_p(self):
        n = len(self.xs)
        zs = np.concatenate([self.xs, self.ys])
        self.simdiff = []
        for j in range(self.ntrials):
            np.random.shuffle(zs)
            self.simdiff.append(np.abs(np.mean(zs[:n]) - np.mean(zs[n:])))
        self.p = np.sum(self.diff < self.simdiff) / (self.ntrials*1.0)
    
    def plot(self):
        fig = plt.figure()
        ax1 = fig.add_subplot(111)
        ax1.hist(self.simdiff)
        ax1.set_ylim(0,1.1*ax1.get_ylim()[1])
        ax1.vlines(self.diff, 0, ax1.get_ylim()[1], color='red', linewidth=3)
        s = '''
        Result of permutation test with %d trials 
        mean1 = %f
        mean2 = %f
        diff = %f
        p = %f
        ''' % (self.ntrials, self.meanxs, self.meanys, self.diff, self.p)
        ax1.set_title(s)
        ax1.set_xlabel('Absolute difference between means')
In [4]:
xs = df['trap_rate'][df['netted']==0].values
ys = df['trap_rate'][df['netted']==1].values
mytest = permtest(xs, ys)

Position of beetles caught by netted panel traps

In [5]:
df.drop(['days', 'netted', 'beetles', 'trap_rate'], 1).sum()
Out[5]:
net_top       31
net_side      34
net_bottom     3
in_trap        8
dtype: float64

BibTex record for this document

@unpublished {Netted-Panel-Traps, title = {Netted Panel Traps to Test if CRB are Deflected}, year = {2015}, type = {CRB Technical Report}, author = {Aubrey Moore and Roland Quitugua and Ian Iriarte} }
In [ ]: