Negative Trade-offs Between Community Forest Use and Hydrological Benefits in the Forested Catchments of Nepal's Mid-hills

Widespread community forestry practices in Nepal's mid-hills catchments involve removal of forest products—including firewood, litter, fodder, and medicinal herbs—by the local communities. Uncertainty is growing about how sustainable the management of these catchments is and whether it can meet traditional needs and maintain ecosystem services, particularly water. As part of a broader study on the hydrological effects of community forestry practices, we measured selected soil properties, including saturated hydraulic conductivity (Ks), bulk density (BD,) and soil organic carbon (SOC) across 4 depths (0–10, 10–20, 20–50 and 50–100 cm) in 3 types of community forest sites—broadleaf, pine-dominated, and mixed—in the Roshi Khola catchment of Kavre district. The same measurements were made at a minimally disturbed religious forest site in the catchment that had higher Ks values than the mixed and broadleaf sites, signifying a lower degree of forest use-related disturbance. Likewise, SOC values for the religious forest were significantly higher (P < 0.05) and BD values significantly lower than the pine-dominated and mixed forest sites, particularly at shallower depths (0–50 cm). Importantly, comparison of the median Ks values (16–98 mm h–1) with rainfall intensities measured at the catchment showed the less intensively used pine-dominated site to be conducive to vertical percolation with possible greater contributions to subsurface storage even during high-intensity rainfall events. These results highlight the critical role of forest use practices in landscape hydrology and have implications for the management of the forested catchments in the broader Himalayan region, particularly in relation to the negative local perceptions of the role of pine plantations on declining water resources.


Introduction
During the early half of the 20th century, forest to farmland conversion and high local demand for forest products, including timber, firewood, livestock fodder, and compostable litter, caused significant loss of forest cover in Nepal's mid-hills and gave rise to the widely publicized but contested ''theory of Himalayan environmental degradation'' (Gilmour 1988;Ives 2004;Hofer and Messerli 2006). The alleged effects, mainly episodes of large-scale flooding and landslides (Eckholm 1976), prompted local and international initiatives to reforest the area as a remedial measure that concurrently fulfilled traditional forest needs. For example, a Nepal-Australia forestry project supported the planting of 20,000 ha of the central mid-hills during the early 1990s (Collett et al 1996), while the World Bank proposed planting at annual rates of 50,000 and 10,000 ha until 1990 and 2000, respectively (Sattaur 1987). The reforestation programs largely used fast-growing species of pine, such as Pinus roxburghii, due to the species' high adaptability to the nutrient-poor soils of the mid-hills (Gilmour et al 1990). Importantly, forest development activities increased focus on community involvement, as customary forest policies systematically alienated local forest users (Acharya 2005;Springate-Baginski and Blaikie 2007) leading to the inception of Nepal's community forestry policy in the late 1970s (Cribb and AusAID 2006). At present, more than half (over 2.2 million ha) of the mid-hills catchments contain naturally grown or planted species of broadleaf and pine, more than two-thirds of which are managed by nearly 7 million local users organized as members of the Community Forest User Groups (hereafter CFUGs) (DFRS 2015).
Forestation is commonly associated with improved landscape stability and hydrological conditions through, for instance, improved soil infiltration (Buytaert et  While the time taken for results to be apparent varies from years (Van Noordwijk et al 2003) to decades (Bonell et al 2010), the varied nature of forest management practices confounds the processes, including the ensuing hydrological regime (Farley et al 2004;Bonell and Bruijnzeel 2005;Wohl et al 2012;Julich et al 2015;Ochoa-Tocachi et al 2016;Marín et al 2018). In the lesser Himalayas, where communities rely heavily on local forests for food, fuel, and income (Breu et al 2017;Chakraborty et al 2018), forestry activities are known to affect many aspects of forest functioning. These activities commonly involve regular planting and harvesting of forest products by local communities. For instance, the persistent harvesting of forest litter and understory in southern China negatively affects the soil's structural complexity and supply of organic matter (Brown et al 1995), while cattle grazing diminished soil nutrient availability and soil hydraulic conductivity in forests in southern India (Mehta et al 2008).
In the mid-hills of central Nepal, soil hydraulic conductivity was negatively affected by sustained forest use (Gilmour et al 1987;, consisting of collection of litter, firewood, fodder, and medicinal herbs that typically constitute CFUG activities in the region. However, the likely hydrological effects of forest use are nonuniform across forested catchments because the intensity or regularity of CFUG activities is determined by varied community needs, as well as forest type and condition. For instance, pine forests are frequented less by CFUG members (oral communication, 2016, Rajendra KC,), because pine needles are not as suitable for composting or livestock fodder as broadleaf (Gautam and Edward 2001;KC et al 2015). Further, the evolving nature of forest ecosystems through successional change, for example broadleaf species integrating into pine plantations (Gilmour et al 1990), as reported from parts of the mid-hills (Gautam et al 2002;DFO Kavre 2014b), obscures the poorly understood forest-water relationships in the region. Clearer understanding of these relationships is critical given growing concerns about increased water shortages during the dry season in the mid-hills (CBS 2017;Poudel and Duex 2017) that are frequently attributed to pine plantations (Bhatta et al 2015;Sharma et al 2016). Additionally, the forested areas of the mid-hills catchments, managed mostly by local CFUGs (DFRS 2015), are vital for the local and regional water supply (Rasul 2016), which is significantly affected by the region's highly seasonal climate (~85% of the annual rainfall occurs during June-September; Merz et al 2003).
As part of a larger study to examine the hydrological effects of the community forestry practices in Nepal's midhills, this paper compares selected soil properties from 3 types of unequally used community forest (CF) sites-a broadleaf, a pine-dominated, and a mixed pine and broadleaf forest-with a minimally used religious mixed species forest in the central hill district of Kavre. The specific soil properties are texture, bulk density (BD), soil organic content (SOC), and saturated hydraulic conductivity (K s ). Further, the paper compares the K s results with rainfall intensities measured at the research site to infer the possible hydrological pathways. Finally, the broad implications of the present findings for likely effects on dry season flows are discussed.

Study area
The study area was the northwestern part of Roshi Khola catchment of Kavre district, Nepal ( Figure 1). The climate varies from subtropical to warm temperate with annual mean temperature of 17 6 0.218C and rainfall of 1330 6 84 mm as shown by the 15-year (2001-2015) records of the Department of Hydrology and Meteorology of Nepal (DHM 2016). The rainfall patterns are highly seasonal, with 60-90% of the annual rainfall occurring during the monsoon period of June to September (Hannah et al 2005;Merz et al 2006). The elevation and aspect influence the microclimate such that the north-facing slopes are moister and cooler than the south-facing slopes (Gautam et al 2003). Typical of Nepal's mid-hills, the soils in the study area are weakly developed and relatively shallow (,100 cm). They are moderate to poorly drained with silt or silt-loam texture and are acidic (pH 4.0-4.3). Forests in the area encompass naturally grown or planted species of broadleaf and pine (mainly Pinus roxburghii), managed primarily by the CFUGs (DFRS 2015). The area under pine forest increased as a result of reforestation programs conducted primarily during the 1980s (Karki and Chalise 1995).
Current study sites were a forested catchment of Indreswar Thalpu (Nga) Community Forest (hereafter referred to as the experimental CF) located 27834 0 10"N, 85830 0 15"E at an elevation of 1710 masl that encompassed stretches of planted pine, natural broadleaf, and mixed forests. After pervasive forest loss, the sites were revegetated naturally and through plantation during the late 1970s and early 1980s, mostly through the auspices of the Nepal-Australia forestry project. In the early 1990s, the forest management responsibilities were officially handed over to the local CFUG (Indreswar Thalpu, Nga) (DFO Kavre 2014a). Thus, organized CFUG activities in those sites have persisted for nearly 30 years (see Table 1).
The current Operational Plan (OP) document of the CFUG (2014/15 to 2024/25) shows that 174 households (total population 790) rely considerably on local forest products. In particular, there is a high annual demand for livestock fodder (.3200 tonnes) and compostable litter (.1600 tonnes), mainly from broadleaf species. However, the annual production levels of the forest are insufficient to meet these demands. This is mainly because of the predominantly agriculture-based lifestyle of the community as well as the significant presence of the pine species in the forest that have low use value for fodder and litter production. Nevertheless, the forest provides a surplus supply of firewood and timber that is occasionally sold in the local markets to earn additional income for the CFUG. This income contributes to funding community development activities, including the construction and maintenance of the local infrastructure, as well as employment (eg wages for the forest watcher).
A religious forest ( Figure 1) with a similar forestry history and soil type was used as a control site that currently undergoes minimal community use because there is a much lower need to obtain forest products for religious purposes.
The individual sites are described in the next section based on the current OP documents, local community insights, and our field assessments, including a forest inventory conducted mostly during January to September 2015. An analysis of the soil profile for each of the forested sites showed the soils to be the fine to fine-loamy derivatives of weathered sandstone, schist, or phyllite of the order Inceptisol (Soil Survey Staff 1994) that transitioned to the parent material at a depth of about 50 to 60 cm. Further details about the location and topography of the sites are provided in Table 2.

Description of sites
Broadleaf forest site: This site was regenerated naturally through community initiatives that primarily involved fencing off the area to restrict traditional forest use. The dominant vegetation consists of Schima wallichii, Castanopsis tribuloides, and Myrsine capitellata along with shrub species, such as Cleyera japonica, Eurya acuminata, Lyonia ovalifolia, as well as Rhododendron arboreum in the higher elevations   Figure 2A). This site undergoes high disturbance due to persistent community use, as the abundance of relatively low-lying (mean height~9 m) broadleaf vegetation is collected for firewood, fodder, and litter ( Figure 2E). There is minimal presence of ground cover on the site except in the less accessible, steeper sections that occasionally have a thin distribution of common grass species.
Pine-dominated forest site: This site is dominated by P. roxburghii, which was planted mostly during the late 1970s to early 1980s. However, due to the proximity to the broadleaf site, the broadleaf species appear randomly along with occasional patches of Nephrolepis fern as ground cover ( Figure 2B). The low species diversity and dominance of pine species on the site has lower value for local users because the pine needles are less suitable as livestock fodder or compostable litter than broadleaf vegetation. Occasional signs of trampling are seen as some local residents visit the site for leisure and, seasonally, to collect wild berries and mushrooms. This level of forest-use intensity entails low to moderate disturbance on the site.
Mixed forest site: This site borders both the broadleaf and the pine-dominated site ( Figure 2C). The broadleaf species, including S. wallichii and C. tribuloides, are mixed with P. roxburghii, although broadleaf species dominate the lower elevations closer to the broadleaf site. While local CFUG members use the site consistently, it also experiences disturbance due to occasional visits for leisure by commuters along the road on its northern boundary ( Figure 2F). The road is unsealed and supports vehicles, mostly during the dry season (October-May). Minor signs of erosion are seen on the site.
Religious forest: The dominant vegetation in this site is a mix of planted and naturally propagated Pinus wallichiana, Quercus semecarpifolia, P. roxburghii, and Alnus nepalensis ( Figure  2D). There is some regeneration and shrub species, including Taxus wallichiana, E. acuminata, and R. arboreum, along with the ground cover of common grass species in steeper sections. Based on the current OP document (2014-2024), the age of the vegetation varies from 5 to 30 years. Although this site is part of the historically degraded national forest, the local community has used the forest for religious purpose since the mid-1980s, until the government formally handed over management duties to the Mukteswar Mahadev religious committee in 2014. Current forest management activities include tree planting, restricted access, and occasional removal of forest products for religious events.

Soil sampling and analysis
Multiple field visits and meetings with the local forest users were held to obtain an in-depth understanding of the forested sites. Samples were collected during March to mid-April 2015 from 6 (5 for the religious forest) representative points located approximately 20-30 m apart in each site along an approximate ''S'' shape ( Figure 1B). As applied in other parts of the mid-hills, the sampling strategy ensured that the sampling sites were not clustered and were distributed evenly (Shrestha et al 2007). The sampling equipment (EijkelKamp Agrisearch Equipment, the Netherlands) comprised chromium-plated stainless steel rings (100 cm 3 ) fitted to an Edelman auger. This was used to draw minimally disturbed core samples from 4 depths ( Importantly, obtaining representative measures of K s is difficult because it is naturally highly variable (Zimmermann et al 2006) and is affected by the methods of measurement (Paige and Hillel 1993;Reynolds et al 2000;Hangen and Vieten 2018;Zhang et al 2019). As such, our sample size may not be sufficiently large to account for such variations, so the K s data presented here need careful interpretation. The samples were drawn from the midrange of each depth, except for the deepest layer (50-100 cm), where the depth to the parent material affected the sampling decision. The samples were analyzed at the laboratory facilities of the Kathmandu University, located approximately 5 km from the experimental CF. Texture was determined by the soil hydrometer method (Gee and Bauder 1986), BD by the core method (Blake and Hartge 1986), SOC by the dry combustion method (Nelson and Sommers 1982), and pH using a glass calomel electrode probe in a soil water ratio of 1:1 (McLean 1982).
Saturated hydraulic conductivity (K s ) was determined using the constant head method based on the Darcy equation given as where V ¼ volume of water flowing through the soil sample, L ¼ sample length, A ¼ cross-sectional area of the sample, t ¼ time taken, and H 2 -H 1 ¼ hydraulic head difference. The K s measurements and apparatus design are based on procedures described by Klute and Dirksen (1986: 694-696). The apparatus comprised a rack to hold 4 core samples that incorporated a constant head maintained by a common water supply. Water was siphoned individually to the soil cores, and the percolated volume was recorded every 10 minutes until 3 constant measurements were obtained. The core method used here is relatively simple, cost-effective, and reliable, particularly in complex landscapes (Ilek and Kucza 2014) such as these.

Rainfall intensity
The rainfall data used to infer the dominant hillslope hydrological pathways were recorded at a nearby location (about 270 m from the experimental CF; Figure 1A, Weather station 1) during the respective monsoon periods of 2015 and 2016. Rainfall was recorded using a tipping-bucket rain gauge (Onset Computer Corporation, USA) at 30-minute intervals.
A rainfall event was categorized as an event that measured a minimum of 5 mm in total and occurred after a dry period of at least 3 hours from the preceding event . For each event, the maximum 30-minute (I 30max ) and 60-minute (I 60max ) rainfall intensities (expressed as equivalent hourly rainfall intensities) were determined by computing the maximum rainfall over the corresponding periods (Ghimire et al 2013).

Data analysis
A nonparametric Kruskal-Wallis test (Kruskal and Wallis 1952) for nonnormal data was used in R (version 3.4.0) to statistically compare the results of the selected soil properties (BD, SOC. and K s ) of the various forest types. Dunn's multiple comparison test (Dunn 1964) with Bonferroni correction was further used to compare the results across 4 depths. A difference was considered significant when P , 0.05.
The K s values obtained were used to infer the likely hydrological pathways with respect to rainfall intensities based on the daily rainfall data collected as described. In doing so, median surface and subsurface K s values for each forest type were compared with the selected percentiles of maximum rainfall intensities (eg over 30 minutes, I 30max ) to estimate the rainfall at the soil surface (Gilmour et al 1987;Bonell et al 2010;Ghimire et al 2013). This is important because the K s distribution and rainfall intensities strongly influence the hydrological pathways in areas with concentrated rainfall such as these (Zimmermann et al 2006;Germer et al 2010).

Results and discussion BD and SOC measurements as influenced by the intensity of forest use
The BD measurements, as expected, generally increased with depth for all sites ( Figure 3D). In particular, median values for the mixed forest site were significantly higher than those of the religious forest at the 3 upper depths (P , 0.007, 0.01, and 0.003 at 0-10, 10-20, and 20-50 cm, respectively). While site attributes could account for this difference, the higher values for the mixed forest suggests increased foot trafficrelated compaction that occurs due to the site's proximity to the road ( Figure 2F). Although the contribution of increased foot traffic is difficult to categorize here, BD as a measure of compaction increases with increased frequency and intensity of forest management activities (Osman 2013), including thinning (Tarpey et al 2008) and harvesting (Whitford and Mellican 2011). Thus, the generally higher median values for the broadleaf site compared with the religious forest could reflect the degree of community use.
The median values of SOC ( Figure 3C) ranged from approximately 2% to 8%. The maximum, 11%, was obtained from the religious forest and the minimum, 1%, for the pinedominated site. As expected, SOC decreased with depth. The values for the religious forest were significantly higher than those for the pine-dominated site at all corresponding depths, suggesting the reduced contribution of pine needles to SOC compared with that of accumulated litter in the religious forest. The SOC levels for the pine-dominated site are consistent with those in other parts of the mid-hills (Shrestha and Singh 2008;Aryal et al 2013), including China (Yang et al 2010) and India (Sharma et al 2011).
In Nepal's mid-hills, inherent factors affecting the SOC levels include forest type, climate, and topography (Bajracharya and Sherchan 2009). These are further affected by community forestry practices, including removal of biomass. The higher SOC levels for the religious forest compared with other sites indicate reduced removal of litter, fodder, or firewood, allowing higher biomass accumulation and decomposition. Similar effects, such as increased SOC levels and associated nutrient availability due to prolonged length of litter retention, have been reported in other parts of the mid-hills (Schmidt et al 1993) and globally, including parts of South and North America. In these cases, the persistent removal of aboveground organic matter reduced soil carbon (Hofstede et al 2002;Powers et al 2005). Conversely, the retention of harvesting residue conserved organic matter and improved site quality and productivity in south Australia (Hopmans and Elms 2009)

K s measurements and inferred hillslope hydrological pathways
The median K s values generally remained higher for the pine-dominated and religious forest (in the shallower depths), likely indicating the lower degree of anthropogenic disturbance related to forest use in these sites (Ziegler et al 2004;Zimmermann et al 2006). The median values ranged from approximately 16-98 mm h -1 with maximum for the pine-dominated and minimum for the mixed forest sites ( Figure 3B). While the values were generally lower for more intensively used sites, that is, the broadleaf and mixed forests, the consistently higher median values for the pinedominated site were significant at 3 depths (P , 0.001, 0.004, and 0.003 at 0-10, 10-20, and 50-100 cm, respectively) compared with the mixed forest site. Similar results showing an inverse relationship between K s and disturbance have been reported in other tropical landscapes (Zwartendijk et al 2017) and in parts of the mid-hills, using in situ methods comprising a constant head well permeameter combined with ring infiltrometers (Gilmour et al 1987) and disc permeameter . The mid-hills studies showed that forestation improved soil infiltration, particularly in the less-disturbed natural forests. This is believed to improve hydrological outcomes in tropical landscapes (Ilstedt et al 2007) through reduced compaction and increased macroporosity due, for instance, to increased SOC levels (Lal 1988;Neary et al 2009). Notably, however, the mixed and broadleaf forest sites of the present study had lower K s values, despite the higher SOC levels, compared with the pine-dominated site (Table 3). This underlines the critical role of anthropogenic disturbance on soil hydraulic conductivity, as has been found in other parts of the lesser Himalayas (Bonell et al 2010).
As Figure 3B shows, the median K s values did not vary significantly among sites at the deepest layer (50-100 cm) but varied widely at the shallower depths (0-50 cm). This is probably a function of the vegetation cover and forest use rather than inherent site qualities, such as soil type. Interestingly, a comprehensive analysis of the global database on these relationships (Jarvis et al 2013) reported that soil texture has only a weak effect on soil hydraulic conductivity, particularly at shallower depths (,30 cm), compared with SOC, BD, and land use. Thus, due to those variations, as well as the presence of an impeding layer, the shallower depths largely govern hydrological pathways causing water to pond or flow laterally, depending on rainfall intensity ( Figure 4A). For instance, overland flow or ponding is probable in the mixed forest site with maximum 30-minute (I 30max ) or 60minute (I 60max ) rainfall intensities of 47.8 mm h -1 and 34.6 mm h -1 , respectively. The intensities were derived from a total of 103 rainfall events recorded during the monsoon periods of 2015 and 2016 that highlight the significant contribution of monsoonal rainfall to the annual totals ( Figure 4B). Specifically, the monsoonal totals were 874 mm (97% of the June-December rainfall) in 2015 and 1030.9 mm (82% of the annual rainfall) in 2016. These values are comparable to the long-term measurements of 944 mm (76% of the annual totals) recorded close to the present study area (Weather station 2, Figure 1A). The frequency and distribution of the maximum 30-minute (I 30max ) and 60-minute (I 60max ) rainfall intensities are presented in Figure 4C.
However, the observed patterns at the mixed forest site are unlikely to be the dominant flow path because the median values of I 30max (11.2 mm h -1 ) and I 60max (7.6 mm h -1 ) suggest vertical percolation with overland flow or ponding probable only beyond the 80% percentile (21.8 mm h -1 ) of I 30max . In fact, percolation to varying depths occurs at all other sites, even under the maximum of I 30max , until ponding or lateral flow occurs ( Figure 4A), with the pine-dominated site allowing percolation to the deepest layer (20-50 cm).
Even though higher rainfall intensities for shorter intervals, for example a maximum hourly equivalent of 88.8 mm h -1 to 130 mm h -1 for 5-minute intervals, have been used to infer the hydrological pathways in other parts of the midhills (Gilmour et al 1987;Ghimire et al 2013, studies have recognized such rainfall patterns to be less dominant in the region.

Hydrological implications of the sustained community forestry practices
Forest-use practices strongly influence the hydrological outcomes of many tropical and subtropical landscapes (Ilstedt et al 2007), which often support the traditional lifestyle of many local communities. In Nepal's mid-hills, the forest fodder and litter are mixed with animal dung, which constitutes the primary source of soil enrichment (Pilbeam et al 2005;Giri and Katzensteiner 2013), including improvement of N, P and K levels (Balla et al 2014). While all forest types of the mid-hills (national, private, or community forests) supply these products, community forests alone contribute more than 50% of the litter supplied (Adhikari et al 2007).
Yet, the hydrological effects of the sustained removal of these forest products are uncertain, even though the resulting reductions in soil microbial activity (Ding Ming et al 1992) and SOC levels are known to reduce rainfall infiltration (Franzluebbers 2002). Further, the corresponding increase in compaction, resulting from higher foot traffic and trampling, exacerbates the situation because it impedes soil hydraulic conductivity (Startsev and McNabb 2000). The lower K s values of the broadleaf forest site are indicative of this effect as it undergoes higher foot traffic due to the greater use value of the forest products, while the mixed forest site has higher foot traffic due to its proximity to the road and has correspondingly low K s values. This could hamper the replenishment of soil and groundwater reserves, contributing to reduced dry-season flows in the area, even though water use by vegetation is an important consideration in evaluating these effects . Indeed, removal of litter and woody debris has been found to cause increased soil loss and runoff (Hartanto et al 2003), while the compaction related to forest use accelerates erosion, shallow landslides , and floods (Alaoui et al 2018). Moreover, a recent study (Upadhayay et al 2018) showed that community forestry practices induce higher sediment loss, than that lost from agricultural land in Nepal's mid-hills catchments. Such a situation confounds reported land use-related social and environmental consequences in the region (Gardner and Gerrard 2003;Jaquet et al 2016) and is ironic because much of the forest in the mid-hills was established to curb sediment loss.

Conclusion
Increased forestation is widely believed to improve hydrological conditions, particularly in tropical and subtropical environments. However, land-use history and prevalent management regimes, such as community forestry practices, might have a greater effect on forest-water relationships, as shown by the results of this study.
Specifically, broadleaf and mixed forest sites showed higher compaction (BD) and lower hydraulic conductivities (K s ) than the minimally used religious forest, which is likely the result of the higher foot traffic and increased trampling associated with greater use of the sites by CFUG members. The K s values of the broadleaf and mixed forest sites were lower, despite their higher SOC values, than the pinedominated site, even though higher levels of SOC improve FIGURE 4 (A) Potential hydrological pathways for the various study sites after a 30-minute rainfall intensity of 47.8 mm h -1 ; (B) monthly rainfall distribution during the study period; (C) frequency distribution of maximum 30-minute (left) and 60-minute (right) rainfall intensities recorded at the study sites in the Roshi Khola catchment of Kavre, Nepal. soil infiltration of forested sites. With growing debate about the role of pine plantations on reduced dry season streamflows in parts of the mid-hills of Nepal, this preliminary study suggests that a more nuanced understanding of the impact of community forestry on catchment hydrology is needed. It also highlights the need for increased research, particularly in view of the prevailing community forestry practices in the broader mid-hills region.

A C K N O W L E D G M E N T S
We are grateful to the members of the Indreswar Thalpu CFUG and the Mukteswar Mahadev religious committee for their help in our work; we received invaluable support to conduct our field activities as well as background information about the local landscape. We thank the people of Malpi and Dhungkharka community for their support, especially Bharat Mijar, Hari KC, Balan Mahat, and Prem Timilsina for the logistic help during the fieldwork. We thank the District Forest Office of Kavre for the valuable information about the community forestry practices in the research area and the overall support of our project. We thank the anonymous reviewers for the constructive comments on the paper.