Estimating Bark Eating Caterpillars Indarbela quadrinotata (walker) in Populus deltoides Using Ranked Set Sampling

The problem of bark eating caterpillar, Indarbela quadrinotata infestation has been observed from variety of horticulture and forest tree species in India. The estimation of infestation of this caterpillar using conventional sampling methods was found difficult because counting the number of caterpillar in each tree is practically not feasible. Ranked set sampling (RSS) is a cost efficient method which provides improved estimators of mean and variance when actual measurement of the observations is difficult to obtain but a reasonable ranking of the units in the sample is relatively easy. In the present study, poplar, Populus deltoides plantation of Western Uttar Pradesh and Uttarakhand was taken for the assessment of Indarbela quadrinotata infestation. The RSS estimator of population mean and variance have been discussed and compared with the corresponding estimators from simple random sampling (SRS). The relative precision (RP) of RSS procedure with respect to the SRS for four different set sizes of k = 3, 5, 7, and 10 has been deliberated. It was seen that RP increase with the increment in k. The method of RSS was found suitable for the assessment of insect pest infestation.


Introduction
The bark eating caterpillar, Indarbela quadrinotata, is economically injurious insect pest. This insect pest infests to many plant species and damages varieties of fruit trees in central India [1] and thus becomes a pest of national significance [2]. Bark eating caterpillar is a polyphagous and comparatively considered of minor importance in forestry; however it is a major pest of Indian gooseberry Emblica officinalis [3,4]. This caterpillar belongs to order Lepidoptera, family Cossidae with bigger larval size, about 35-45 mm, dirty brown in colour and smooth body. Presence of long winding, thick and brownish ribbon-like masses composed of small chips of wood and excreta on the tree is evidence of infestation of this pest [1].
The larva of Indarbela quadrinotata make shelter tunnels inside the trunk, feeding on plant tissues and make plant and stem weaker, resulting in the drying of the branches and finally the mortality of the tree itself [5]. The infestation of pest generally starts in the month of April with the emergence of moths [6]. Female lays egg below the bark or in between cracks and there after larvae bore into the tree continue feeding on the bark for 9 to 11 months, however, the damage caused by this pest reached maximum during rainy season [1,6,7]. The larva of I. quadrinotata made a tunnel into the trunk of the tree for about 20 cm depth and take refuge during the day and eats the bark of the tree during night, making excretal ribbons. Feeding of bark and internal tissues by the larva resulted in inhibiting translocation of cell sap, reduces growth and top dying of the plant [8] and subsequently, death of the heavily infested plants. The pest has been reported to infest 70 plant species across fruits, forests and avenue plantations [9].
The enumeration of the population of active caterpillar I. quadrinotata per tree or estimation of proportion of trees infected by the caterpillar is one of the important challenges for entomologist to know density and distribution of the caterpillar over the tree or tree stands. The caterpillar makes tunnel into the tree and feed bark tissues covering with fecal ribbons. It is compulsory to remove the ribbon made by I. quadrinotatata confirm the active insects inside which requires more resources in terms of effort, cost and time. Therefore, the conventional sampling technique like simple random sampling (SRS) failed here because the actual measurement of active insects present in each tree is not feasible.
An effort for the assessment of I. quadrinotata in poplar trees has been made in the present study with the help of ranked set sampling (RSS) which primarily required ranking of trees based on the number of ribbons and their sizes rather than the actual measurement of the number of active insects in each tree. In this method, the first step is to rank the small set of samples of trees by some rough gauging method (say visual inspection) without the actual measurement. After repeating the process in certain cycles, the unit (tree) associated with each rank order statistic is selected for measurement (see The technique of RSS has been applied satisfactorily in various areas of agriculture and forestry. For example, in estimating yield of dry bark and quinine contents from Cinchona plants [10]; estimation of forage yields in a pine hardwood forest [11]; estimation of shrub phytomass in Appalachinan Oak forest [12]; survey of conifer (Pinus palustris) trees [13] etc.
In the present study, the average number of caterpillars per poplar tree had been estimated by RSS and the same has been compared with the corresponding estimates of SRS. Section 2 deals with material and methods used in the present case study. The framework of RSS has been briefly described and the technique of RSS has been applied in estimating bark eating caterpillars in poplar tree. Section 3 discusses the results and finally the study has been concluded in Section 4.

2.1.
Study area. A total of 12 different sites which includes farmer's plantation areas of poplar, Populus deltoides tree in Haridwar, Saharanpur and Muzaffarnagar districts of the states of Uttarakhand and Uttar Pradesh were selected. The description of these sites along with the area under plantation is given in Table 1. Takahasi and Wakimoto [15] proved that when ranking is perfect, the RSS mean is an unbiased estimator of the population mean and the variance of the RSS mean is always smaller than the variance of the mean of SRS of the same sample size. Dell and Clutter [16] showed that RSS performs better than SRS even in presence of ranking error in the experiment. For more detail, one may refer to [17,18,19,20].

Preference of RSS over SRS.
To compare RSS and SRS, one can find the relative precision (RP) of RSS procedure with respect to the SRS as   represents the average variances within the rank order statistics.
Takahasi and Wakimoto [15] proved that under the perfect ranking, the RP lies between 1 and 2 / ) 1 (  k . The RSS performs better than SRS even in presence of the ranking error [16]. When the actual measurement of the units requires more resources and the reasonable ranking is cheap, the RSS is most effective.
The moderately skewed distribution does not follow any of the above systematic patterns of variances.
Various authors also improved balanced RSS procedure by using the appropriate allocation for each rank order statistics, termed as unbalanced RSS. Kaur et al. [21] and Chandra et al. [22] provided allocation procedures for skewed populations by making allocations in directly proportional to the rank order statistics. For symmetric distributions, Kaur et al. [18] and Chandra et al. [23] provided allocations inversely proportional to the rank order statistics for improving the estimates.
Some recent developments (for mean estimation using RSS [24,25]; for variance estimation [26]; for unequal allocation model in RSS for response estimation of developmental programs [27]) can also be seen in the literature.

Application of RSS in estimating Indarbela quadrinotata in Populus deltoides
tree. In the present study, the concomitant variable, number of ribbons made by Indarbela quadrinotata was chosen for the ranking purpose.
Four small set sizes k = 3, 5, 7 and 10 were used to reduce the possible ranking errors while selecting samples. For illustration of selection procedure, let us consider the set size of k = 5. Five sets of poplar trees each having size of five was selected from the randomly selected sites by the use of simple random sampling without replacement. Such sets of five trees were ranked based upon the number of ribbons made by the insect by visual inspection. Once ranking is done, from the first set, the smallest ranked tree was taken for the actual measurement and other trees were discarded. Similarly, the second smallest ranked tree from the second set was selected for the actual measurement. This process was continued until the largest ranked tree from the last set of five trees was selected for the actual measurement. In this process, 25 trees were randomly selected from five different sets of trees and five were taken for the actual measurement. Now, from these selected five trees, the ribbons from the trees were removed and the numbers of active caterpillars were counted. This process is termed as one cycle. In the present case study, five cycles were taken so as to get 25 trees for actual measurement out of 125 trees. The actual numbers of active caterpillar found for different set sizes are shown in Table 2 to Table 5.

Results and Discussion
The population values were calculated based upon the 125 measured ordered observations (Tables 2 to Table 5). The population mean and population variance (using equation (3) and (4)) was found to be 6.5760 and 7.7462, respectively. The variances  [18,21,22,23]).  The RSS estimator for the population mean using equation (1) and values of RP using equation (5) for different values of k were calculated ( Table 7). The calculation revealed that the RSS estimator is very close to the population mean 6.5760 and the 9 variance is substantially reduced when the set size k increases. The gain in RP of RSS over SRS was found significantly higher and RP increases as the set size k increases.  (1) shown in the Figure 1 for set sizes k = 2 to 10.
Empirically, for all the distributions, the RP increases as the set size k increases.
The present study also suggested the same ( Table 7). The rate of increase in RP becomes higher as skewness of the distribution increases.

Conclusions
The bark eating caterpillar, Indarbela quadrinotata, is a serious pest that damages variety of horticulture and forestry trees in India and is a pest of national significance. It eats the bark of the plant and the heavily infested plants ultimately lead to mortality. To confirm the active insects in the tree, it was compulsory to remove the ribbon made by Indarbela quadrinotata which requires more resources in terms of effort, cost and time.
The assessment of this caterpillar was quite important for taking up remedial measures like control of pest at the egg or larval stages. The conventional sampling technique was not suitable here as the measurement on each and every unit in neither feasible nor desirable. An easier sampling procedure, RSS, which require only ranking of tree after visual inspection, estimated the number of caterpillar with greater precision and at a substantially lower cost. The technique of the RSS can be applied to other similar estimation problems of entomology cases.