Keiko Fukumoto · Tetsuji Ota · Nobuya Mizoue · Shigejiro Yoshida ·Yukio Teraoka · Tsuyoshi Kajisa

Abstract This study quantified the effect of weeding frequency and weeding schedules on weeding operation time in a sugi (Cryptomeria japonica) plantation stand. A weeding operation time estimation model was developed;then the cumulative weeding operation time after six growing seasons was simulated using the developed model.The developed model included weed height,relative height of weeds to sugi,and initial planting density.The simulated cumulative weeding operation time decreased approximately 6% for each one-treatment decrease in weeding frequency. Under a three-treatment weeding frequency scenario, the simulated cumulative operation time when weeding was conducted during non-consecutive years was longer than that when weeding was conducted during three consecutive years. The results suggest that carrying out weeding treatment during consecutive years is the more effective for reduction of weeding costs. We conclude that weeding schedule as well as weeding frequency must be considered for reduction of weeding operation time.

Keywords Weeding · Operation time · Simulation · Lowcost silviculture · Growth model

Introduction

Weeding is an important practice for sustainable forest management (Dacosta et al. 2011), because weeding promotes growth of planted trees (Mallik et al. 2002; Miller et al. 2003; Homagain et al. 2011), by reducing competition between planted trees and other plants (Wagner et al.2006).In Japan,mechanical treatment with a brush cutter is conducted at least once or twice per year for 5 or 6 years after planting (Tsutsumi 1994). Because weeding using a brush cutter requires repetitive treatment due to rapid weed regrowth (Biring et al. 2003), the method would be the cause of high costs (McDonald et al. 1994; Biring et al.2003;Comeau and Harper 2009;Roy et al.2010).The cost of weeding amounts to approximately 40%of the total cost during the first 10 years after planting in Japan (Yukutake and Yoshimoto 2001).Thus,reduction of the weeding cost is necessary for efficient forest management (Forestry Agency 2015). One option to reduce weeding cost is to reduce weeding frequency (Forestry Agency 2015).

To reduce weeding frequency,knowledge of the optimal weeding frequency and weeding year is required. A key factor for determination of the optimal weeding frequency and weeding year is the weeding operation time, which is the time consumed per weeding. Several studies have assessed the factors that affect the weeding operation time with a brush cutter (Kinjou et al. 2011b; Kitahara et al.2013). Kinjou et al. (2011b) implied that weeding operation time was affected by the height of weeds and the ease with which the operator was able to recognize the planted trees. The authors suggested that the size of planted trees may affect the weeding operation time, because the operator may need to be more careful when the planted trees are small. However, the relationship between the weeding operation time and the weeds’ height and the size of planted trees was not quantified. Similarly, Kitahara et al.(2013) suggested that operation time of weeding was affected by the weeds cover ratio. However, these authors focused on the stand only just after the first growing season.Given that weeding is conducted for 5 or 6 years after planting, relationships between operation time of weeding and other factors in forests should be evaluated from the first to the fifth or sixth growing season.In addition,factors other than the size of weeds and the ease of recognition of the planted trees should be examined. For example, the higher the initial planting density, the more operation time may be required, because operators must circumvent the planted trees. The weeding operation is one of the most physically demanding tasks for operators. Weeding is usually conducted between June and July, during which high temperatures and relative humidity are experienced in Japan. For example, the average temperature and relative humidity at Kagoshima city, which is near the present study site, in July are about and 29.2 °C and 81%,respectively (Japan Meteorological Agency 2017). Moreover, slope aspect may affect the weeding operation time due to differences in solar radiation. High solar radiation may cause physical exhaustion of operators and thus may affect weeding operation time.

As noted above, several factors potentially affect weeding operation time. Thus, to determine the optimal weeding frequency and weeding year, it is useful to know the influence of weeding frequency and weeding year as well as other factors on the weeding operation time. The objective of this study was to investigate the effect of weeding frequency, weeding year and other factors on the weeding operation time in a sugi (Cryptomeria japonica)stand, which is an economically important and common plantation species in Japan. We developed an operation time estimation model to account for weed height, planted sugi height, slope aspect, initial planting density, and relative height of weeds to sugi saplings. Then, the cumulative operation time under different weeding scenarios and initial planting densities was simulated using the developed model.

Materials and methods

Study site

The Takakuma Experimental Forest of Kagoshima University was used as the study site.Detailed descriptions of the location were presented by Fukumoto et al. (2017).After clear-cutting of the sugi plantation in May 2006,sugi saplings were planted in 2007. The total planted area was 2.57 ha. In March 2007, we established study blocks in each of which a particular weeding schedule was applied:(1) weeded annually for six growing seasons (i.e.2007–2012); (2) weeded in the second, fourth, fifth, and sixth growing seasons (i.e. 2008, 2010, 2011, and 2012),(3) weeded in the third to the sixth growing seasons (i.e.2009–2012),(4)weeded in the first,third,and fifth growing seasons (i.e. 2007, 2009, and 2011), (5) weeded in the second to the fourth growing seasons (i.e. 2008–2010), (6)weeded in the first to the third growing seasons (i.e.2007–2009), and (7) never weeded (non-weeding). The blocks differed in initial planting density (1500 or 3000 trees ha-1)and slope aspect(south-or north-facing),thus there were 14 study blocks in total (Table 1). The slope in the 14 study blocks was almost identical. Given that differences in slope were limited, we assumed that slope did not affect our analysis. A plot that included approximately 60 sugi saplings was established within each block. The major species of weeds in the blocks were Mallotus japonicus, Rhus javanica var. chinensis, Zanthoxylum ailanthoides, Aralia elata, Machilus thunbergii,and Castanopsis sieboldii (Kinjou et al. 2011a; Fukumoto et al. 2015). In this study, we excluded the non-weeding block from the analysis because weeding was not conducted in non-weeding block.

Measurements

Operation time of weeding

Weeding with a brush cutter was conducted once per year in early July.Three operators worked at each weeding.The same three operators conducted the weeding throughout the 6-year experimental period. We recorded operation time during weeding with the brush cutter, excluding the time spent walking into the study blocks, maintaining tools,refueling, and taking breaks. The operation time per hectare(h ha-1)was calculated for each block area.Given that 23 weeding operations were conducted within each slope aspect (Table 1), in total we recorded operation time data for 46 weeding operations. The operation time data are summarized in Table 2.

Table 1 Timing (years after planting) of weeding, block area, slope aspect, and initial planting density

Table 2 Summary of measurements data

Sugi and weed height growth

We measured sugi tree height and weed height around the sugi trees in each plot from the first to the sixth growing seasons (i.e. 2007–2012). The sugi trees were measured after the growing season every year, and weeds were measured before the weeding was conducted. Detailed descriptions of the measurements are provided in Fukumoto et al. (2017). We summarize sugi and weed annual height growth (m year-1) in Table 2. Weed height measurements were not recorded for some plots in some growing seasons because of understaffing. Although some data were lacking,we were able to use data for 31 weeding operations for which weed height measurements were available. Thus, the data were enough to develop the operation time model.

Data analysis

Operation time model

An operation time estimation model was developed using a hierarchical Bayesian model. The observed operation time of the ith plot in the jth implementation of weeding was assumed to have a normal distribution, as follows:

where Tijis the observed operation time per weeding treatment per hectare(h ha-1),is the expected operation time (h ha-1), and σ2is the variance of the operation time distribution. The explanatory variables included weed height (WH), sugi height (H), slope aspect (S), initial planting density (D), and the relative height of weeds to sugi (WHH; calculated as WH divided by H). WHH was included as an index to express the ease of recognition of sugi. The full model for operation time was expressed as:

where α0- α5are each parameter, and φiis a random parameter for the ith plot. We assigned a normal distribution for the random parameter as follows:

where τ is the variance of φi.We assigned a uniform prior to τ and other parameters.The Markov Chain Monte Carlo(MCMC)method was applied to the parameters estimation using JAGS 3.4.0 software(Plummer 2003)from within R 3.1.3 software(R Core Team 2014)with the rjags package.We ran four chains each of length 30,000 steps,after burnin of 1000 steps and 1/10 thinning.Following our previous study(Fukumoto et al.2017),the Gelman and Rubin ^R and deviance information criterion(DIC) were used to confirm the convergence of MCMC and to select the best model,respectively.

We calculated the root mean square error (RMSE) and coefficient of determination (R2) from the observed and predicted operation times(Fig. 1).The marginal R2and the conditional R2(Nakagawa and Schielzeth 2013) were also calculated to evaluate the model following Fukumoto et al.(2017).

Weeding simulation

Cumulative operation time over the six growing seasons under the different weeding scenarios was simulated using the selected model. For the simulation, in addition to the operation time model, we used the sugi and weed height growth models developed by Fukumoto et al. (2017) as follows:

Detailed descriptions of the models are provided by Fukumoto et al. (2017). Briefly, the explanatory variables include WH, H, WHH and S. The value 1 was assigned to S when the slope faced south; otherwise, zero was assigned. We assumed that the initial planted sugi and weed heights were 0.6 m and 0 m, respectively, based on our measurements. The simulated weeding frequencies ranged from one to six depending on the weeding treatment during the six growing seasons.We assigned the value zero to WH for estimation of sugi and weed height growth,when the weeding was conducted. For each weeding frequency,there are several schedules for the weeding year. Thus, all possible weeding schedules for each weeding frequency were simulated in this study. The number of schedules for one,two,three,four,five,and six weeding frequency are 6,15, 20, 15, 6, and 1, respectively. In total, there are 63 possible weeding schedules.

We simulated cumulative operation time under the different weeding frequencies under both initial planting densities (i.e. 1500 and 3000 trees ha-1) on north- and south-facing slopes. Then, Hijand WHijwere calculated with Eqs. 3 and 4, and we substituted the values obtained into Eq. 2.To evaluate the effect of the weeding schedules,we focused on the three-treatment weeding frequency,following a previous study (Fukumoto et al. 2017). The cumulative operation time and sugi height under both initial planting densities and both slope aspects showed the same trend. Thus, we assumed that the initial planting density was 3000 trees ha-1and a north-facing slope when we focused on the three-treatment weeding frequency(Fig. 2).

Fig. 1 Relationship between predicted and observed values of weeding operation time per operation. a The predicted operation time calculated only from fixed factors. b The predicted operation time calculated from both the fixed and random factors

Fig.2 Summary of the relationship between simulated mean cumulative operation time and weeding frequency for a a north-facing slope,and b a south-facing slope. Error bars indicate standard deviations

Results

Model selection and goodness of fit

The model including WH,WHH and D was selected as the best model based on DIC (Table 3). The coefficients of WH,WHH and D had positive values.The marginal R2and RMSE values were 0.65 and 1.10,respectively.In addition,the conditional R2and RMSE values were 0.93 and 0.18,respectively. The R2and RMSE values suggested that the selected model well expressed the predicted weeding operation time per one operation.

Weeding scenario

The predicted mean cumulative operation times on a northfacing slope with annual treatment under initial planting densities of 1500 and 3000 trees ha-1was 156 and 171 h ha-1, respectively (Fig. 2a). On a south-facing slope, the predicted mean cumulative operation time withannual treatment under initial planting density of 1500 and 3000 trees ha-1was 147 and 162 h ha-1, respectively(Fig. 2b). The mean cumulative operation time under an initial planting density of 3000 trees ha-1was longer than that of 1500 trees ha-1. The mean cumulative operation time with a three-treatment weeding frequency under both initial planting densities and slope aspects was about 88 to 106 h ha-1.Mean cumulative operation time decreased by approximately 6% for each one-treatment reduction in weeding frequency. Thus, if the weeding frequency was halved, the cumulative operation time was not halved.

Table 3 Summary of parameters and credible interval incorporated in the selected model

We then focused on the three-treatment weeding frequency under an initial planting density of 3000 trees ha-1on a north-facing slope. Weeding treatment in years 1, 2,and 6 required the longest operation time (112 h ha-1),whereas weeding treatment in years 1,2,and 3 required the shortest operation time (93 h ha-1; Fig. 3). These values represented 65% and 54% of that of the 6-year treatment,respectively. The difference in operation time between the two scenarios was 14 h ha-1. The operation time required was increased when weeding was not conducted in consecutive years (e.g. treatment in years 1, 2, and 6, or years 1,3,and 6,or years 1,5,and 6).In contrast,when weeding was conducted in consecutive years, the cumulative operational time decreased (e.g. treatment in years 1, 2, and 3,or years 2, 3, and 4).

Discussion

Fig. 3 Predicted cumulative operation time under a three-treatment weeding frequency on an initial planting density of 3000 trees ha-1

For efficient forest management,weeding operation time is an important factor because the operation time directly relates to cost. This is, to our knowledge, the first attempt to investigate the effect of weeding frequency and weeding schedules as well as other factors on weeding operation time. Weeding operation time was affected by WH, WHH and D (Table 3). Given that the coefficients of WH and WHH were positive, the higher the weed height or the higher the weed height relative to sugi height required a longer operation time. The possible reason is that weeds obscured the planted trees,hence the ease of recognition of the planted trees affected the operation time (Kinjou et al.2011b).The results mean that keeping the weed height low is important to reduce the weeding operation time. However, even if the weed height is low, a longer operation time is required if the planted tree height is low;therefore,the planted tree height as well as the weed height is important for reduction of the weeding operation time.

Initial planting density also positively affected weeding operation time (Table 3). One possible reason is that operators have to circumvent more planted trees when the planting density is higher.The other possible reason is that the interval between adjacent trees is narrow at a high initial planting density.Thus,a sensitive weeding operation may be needed under a high initial planting density.Recent studies indicate that low-density planting is effective with regard to reduction of the initial planting cost(Sasaki et al.2009; Ota et al. 2013) and enhanced planted tree growth(e.g. Fukuchi et al. 2011). We revealed that low-density planting is effective for reduction of the weeding cost as well as reduction of the initial planting cost. However, it should be noted that the reduction in weeding time achieved with low-density planting is much smaller than that attained by reduction of weeding frequency (Fig. 2).

Cumulative operation time decreased approximately 6%for each one-treatment reduction in weeding frequency.Even if the weeding frequency was halved,the cumulative operation time was not similarly halved (Fig. 2). When weeding is not conducted in a given year, the operation time for that year is zero. Thus, the cumulative operation time decreased with the reduction of weeding frequency.However,the operation time of weeding is increased in the year immediately following that when weeding is not conducted. This is because WH and WHH increase in the absence of weeding and have positive effects on the subsequent weeding operation time. Thus, the cumulative operation time does not decline commensurate with the decrease in weeding frequency. Both the operation time reduction due to non-weeding in a given year and the increase in operation time resulting from growth of WH and WHH must be considered. With a three-treatment weeding frequency, the cumulative operation time tended to increase when weeding was not conducted in consecutive years(e.g.treatment in years 1,2,and 6,or years 1,3,and 6, or years 1, 5, and 6) (Fig. 3). Weeding should be conducted in consecutive years to reduce weeding cost.

This study demonstrated that weeding frequency and weeding schedule affect weeding operation time. Weeding frequency as well as weeding schedule also affects the growth of planted trees (Fukumoto et al. 2017) and mortality of planted trees(Kobe et al.1995).To determine the optimal weeding frequency and weeding schedule,not only the cost (i.e. weeding operation time), but also the growth and mortality of planted trees should be considered. Thus,further studies accounting for weeding operation time,growth of planted trees, and planted tree mortality are required.

Conclusion

We clarified the factors that affect weeding operation time.We conclude that WH, WHH and D positively affect weeding operation time. We also conclude that weeding schedules as well as weeding frequency affect the weeding operation time.With regard to weeding operation time,our findings suggest that weeding should be conducted in consecutive years. Further studies accounting for weeding operation time, growth of planted trees, and planted tree mortality are required.