Effects of forest harvesting and woody debris removal
on two Northland streams, New Zealand 

BRENDA R. BAILLIE et al
New Zealand Journal of Marine and Freshwater Research v.39, n.1, March 2005 1mar2005

BRENDA R. BAILLIE
   Forest Research
   Private Bag 3020
   Rotorua, New Zealand
   email: brenda.baillie@forestresearch.co.nz

KEVIN J. COLLIER
JOHN NAGELS
   National Institute of Water and Atmospheric Research Limited
   P.O. Box 11 115
   Hamilton, New Zealand

Abstract 

The short-term effects of Pinus radiata forest harvesting to the stream edge followed by stream-cleaning (removal of woody debris from the stream channel), on instream light levels, stream temperature, dissolved oxygen concentrations, and aquatic invertebrates were assessed in streams draining partly (25% clear-cut) and totally (100% clear-cut) harvested catchments, compared with nearby indigenous forest and mature pine plantation reference sites. There were marked increases in in-stream light levels and water temperatures following forest harvest and stream-cleaning at both sites. In-stream light levels increased from 8–13% to 60–90% and maximum monthly water temperatures increased on average by 5.6°C in the partly harvested and 3.6°C in the fully harvested catchment. Dissolved oxygen levels decreased at both sites shortly after harvest (94%–71% saturation in the partly harvested catchment; 72%–37% saturation in the totally harvested catchment), increasing to 75% and 81%, respectively, 1 year later. Although aquatic invertebrate mean density and taxa richness increased at both sites following harvest, the relative abundance of sensitive mayfly, caddistly, and stonefly species decreased and community composition changed to one dominated by Chironomidae (midges) or Mollusca. Impacts relative to pre-harvest conditions were not as marked in the totally harvested catchment, possibly because of pre-existing elevated stream temperatures and high levels of sand and silt. Any downstream protection provided by the forested headwaters of the partly harvested catchment was soon lost after the stream entered the clear-cut area, although these forested headwaters may provide a potential source of aquatic invertebrates for re-colonisation in the future as water quality and habitat recover. Our results suggest that: (1) pre-existing constraints on habitat quality can influence the magnitude of harvesting impacts; and (2) length of stream edge harvested may be a better indicator of impact on some aspects of stream ecology, such as lighting, stream temperature, dissolved oxygen, and aquatic invertebrate community composition, than percentage of catchment harvested. This study also highlights the importance of considering the hydro-logical and landscape context for mitigating harvesting and wood management impacts on stream ecosystems. 

Keywords harvesting; forestry; stream; aquatic invertebrates; temperature; dissolved oxygen; light levels; substrate; woody debris; New Zealand

Forest harvesting along stream edges has the potential to significantly impact physical, chemical, and biological processes in the stream ecosystem (Campbell & Doeg 1989; Harding et al. 2000; Sabater et al. 2000; Thompson 2001). Of primary importance is the removal of riparian vegetation, which can increase the amount of light reaching the stream channel (Davies-Colley & Quinn 1998). Increases in solar radiation to small shallow streams typically elevate stream temperature (Graynoth 1979; Collier et al. 1997), and can increase primary production, mainly through increased periphyton abundance which can modify nutrient spiralling processes (Graynoth 1979; Sabater et al. 2000; Thompson 2001). These factors, along with inputs of logging slash and increased loadings of fine organic matter in the stream channel, can lead to decreases in dissolved oxygen minima (Ponce 1974; Robertson 1999; Collier & Bowman 2003). In addition, harvesting and associated roading activities have been shown to elevate water flows (Campbell & Doeg 1989; Fahey & Rowe 1992) and alter substrate composition through increased sediment deposition on the streambed, either via channel bank erosion or sediment carried by surface run-off (Graynoth 1979; Harding et al. 2000; Kreutzweiser & Capell 2001). These cumulative effects of forest harvesting operations can alter the composition and abundance of aquatic invertebrate communities (Graynoth 1979; Robertson 1999; Harding et al. 2000; Thompson 2001; Collier & Bowman 2003; Death et. al. 2003). Many impacts of pine forest harvesting are minimised or eliminated when riparian forest cover is retained (Harding et al. 2000; Thompson 2001).

Plantation forestry in New Zealand is dominated by Pinus radiata D. Don, a native of western North America, which now accounts for 22% of New Zealand's total forest cover (NZFOA 2002). Most harvest operations in New Zealand clear-cut forests to the edge of streams, many of which are subject to frequent high flows. Woody debris entering the stream during harvest operations is often removed ("stream-cleaned") to reduce the risk of debris dam formation, and mobilisation of wood during high flow which can cause damage to stream habitat and downstream infrastructures such as culverts, bridges, and fences (Baillie 1999). Stream-cleaning also removes any shade provided by woody debris covering the stream channel.

Most woody debris removal studies have involved removal of wood that has been present in the stream for some time, usually resulting in changes in channel bed stability, increased sediment and organic matter export from the stream system, and reduced pool frequency (Bilby & Likens 1980; Bilby 1984; Smith et al. 1993). Less is known of the effects of stream-cleaning on water quality or lotic ecology, although recent studies suggest that the magnitude and duration of these impacts vary with degree of channel disturbance and stream type (Collier & Bowman 2003; Death et al. 2003). To investigate the short-term impacts of clear-cut harvest to the stream edge and post-harvest wood removal, we studied two streams in Northland, New Zealand. We focused on the effects on in-stream light levels, temperature, dissolved oxygen, and substrate to understand mechanisms leading to changes in aquatic invertebrate communities relative to pre-harvest and indigenous forest stream conditions. Temperature and in-stream light levels were also compared with a stream in a mature stand of P. radiata.

All sites were located in Mahurangi Forest, Northland, New Zealand (Fig. 1). This area has a subtropical climate, dominated by westerly winds (Hessell 1988). Mean annual rainfall is between 1600 and 2000 mm; maximum rainfall is in winter and localised heavy rainfalls are common. Mean annual air temperature is between 14.0 and 14.5°C. Geology is predominantly noncalcareous sandstone and mudstone (Waitemata Group), with weakly to strongly leached yellow-brown earths (Ministry of Works and Development 1975; Department of Scientific and Industrial Research 1961). Hill slopes range from moderately rolling to very steep. Before the establishment of P. radiata in this forest, the predominant land cover was gorse and manuka scrub with smaller areas of indigenous forest and reverting grassland.

Sites 1–3 drained in a south or south-easterly direction (Fig. 1), whereas site 4 flowed north-westerly. Sites were either 1st or 2nd order streams, and varied in catchment area and channel size (Table 1). Site 1 was in indigenous forest and was sampled on all occasions to provide a measure of temporal variability under unimpacted conditions. Site 3 provided a partly harvested comparison and site 4 a fully harvested catchment (Table 1). Maintaining continuity of temperature data was difficult at these sites because of frequent high flow events, so site 2 (a smaller mature pine site, 19 ha) provided an additional reference site for temperature and light levels only. A number of additional sites were selected before harvest to provide spatial replication but unpredictable harvest schedules limited the number of harvested streams to two.

For the purposes of this study, year 1 refers to August 1999 to July 2000; year 2, August 2000 to July 2001; and year 3, August 2001 to July 2002. Forest harvesting began in the lower part of the catchment at site 3 in year 1; trees were clearfelled and extracted from the ridge top to the stream edge using a yarder system. By the end of year 1, trees had been harvested on both sides of the stream in the vicinity of the study site and harvested on one side of the stream only upstream from the study site. By the end of year 2, the remainder of the lower catchment area had been harvested (45 ha in total, 25% of the catchment area, 47% of the length of the main stem of the stream). Immediately after harvest, the stream was manually stream-cleaned; all large woody debris (LWD) and the majority of small woody debris (SWD) were removed from the stream channel and deposited on the banks. Most of this material was logging slash which had been in the stream channel for a short period of time (weeks to months).

The entire catchment up to the stream edge was harvested at site 4, predominantly using yarder operations although c. 10% of the area was harvested using ground-based machinery. By the end of year 1, 83% of the catchment had been harvested, with a further 11% in year 2, and the remainder in year 3. The stream channel was stream-cleaned for c. 300 m of its length from the bottom of the catchment where the study reach was located. Large stems suspended or partly suspended across the channel were left; the remainder of the woody debris, predominantly composed of recently deposited logging slash, was removed.

Sampling was conducted over a 3-year period and all variables were measured before harvest (pre-harvest). Stream lighting and dissolved oxygen (DO) were measured shortly (within 1 month) after harvest (post-harvest), and all variables were re-measured in the summer following harvest (post-harvest 1) (Table 2). Measurements were made along a 100 m section of stream channel located in the lower part of the catchment at each site. Site 1, the indigenous forest reference site, was sampled in all years. One-off measurements at sites 1 and 2 provided reference light levels at the indigenous forest and mature pine forest sites, and stream temperature was monitored continuously throughout the 3-year period at all four sites.

Table 1 Characteristics of the stream study sites in Mahurangi Forest, Northland, New Zealand.

Site: 				1	    2	    	    3	      	    4
Grid reference (NZMS 260 R09)	590 388     576 384   	    576 382   	    536 362 
Stream order*			2nd 	    1st 	    2nd       	    2nd
Catchment area (ha) 		114 19      179 118
Vegetation 			indigenous  Pinus radiata   60% indigenous  Pinus radiata
							    40% Pinus radiata
Bankfull width (m) 		2.6 	    3.4 	    6.5 	    3.6
March baseflow (litres/s) 	9 	    2 		    12 		    9 

*Strahler 1957, based on NZMS 260 1:50000 maps.

To measure changes in stream lighting following riparian harvest and stream-cleaning, in-stream light levels were assessed at each site using two LAI-2000 canopy analysers, following the method of Davies-Colley & Payne (1998). A reference canopy analyser was positioned in open light conditions close to the stream site. The second canopy analyser was deployed at the water surface to measure stream light. Readings were taken successively at 10%, 30%, 50%, 70%, and 90% of the stream width at 5 m intervals along the stream channel until 20 point readings were recorded. These readings were averaged for each site and expressed as a % of light from the open reference site. This procedure was repeated at bank-full height. Descriptive statistics were used to compare changes in light levels before and after harvest at sites 3 and 4 with the two reference sites.

Dissolved oxygen was measured simultaneously at each site using Multiparameter water quality sondes (DataSonde 3, Hydrolab Corp. Texas, United States). Pre- and post-harvest measurements were timed for the height of the summer season; post-harvest measurements within 1 month of harvest. Duration of measurement was 5–7 days and DO readings and temperature were logged at 15-min intervals. These temperature readings were only used in the interpretation of the DO data. Separate temperature data collected from continuous temperature monitoring (see below) were used in the interpretation and analysis of stream temperature.

At each site, surficial substrate composition was determined by measuring the intermediate axis of 10 particles across 10 evenly spaced transects along the 100 m reach (Leopold 1970) and classifying them into 10 inorganic particle size classes and two woody debris size classes following Gordon et al. (1992). Although this method may overestimate the absolute frequency of larger particles, which have a greater surface area and therefore a greater chance of being sampled (Leopold 1970), it nevertheless should provide a useful indication of the relative amounts of different sized substrate elements available for benthic invertebrate colonisation.

Stream temperature was monitored continuously using StowAway Tidbit temperature loggers (operating range -5 to +37°C). Loggers were checked for accuracy in containers of crushed ice and water before deployment in the field. Temperature loggers were deployed at the downstream end of the 100 m transects at each site. Stream temperature was logged at 10-min intervals until autumn in year 3 when intervals were increased to 30 min. Descriptive statistics of daily and monthly data and paired t-tests were used to compare temperature at the two harvested sites (sites 3 and 4) before and after harvest with temperature at the two reference sites (sites 1 and 2).

To assess aquatic invertebrate community composition, five Surber samples (0.1 m², 0.25 mm mesh) were collected from locations randomly selected from x–y coordinates (metres along reach x % distance across transect) along the 100 m reach at each site. Because of the random selection of sampling sites, invertebrates were collected from a range of habitat types, but most were from runs. Samples were preserved in the field in 70% isopropyl alcohol, and whole samples were hand-picked and identified in the laboratory using Winterbourn (1973), Chapman & Lewis (1976), Winterbourn et al. (2000), and B. Smith (NIWA, Hamilton, New Zealand, unpubl. key for Hydrobiosis). Most identifications of aquatic insects and molluscs were to species or genus, whereas other groups were identified to the lowest practical taxonomic level, often Phylum (e.g., Platyhelminthes, Nemertea).

Table 2 Timetable of physico-chemical and biological measurements. (DO, dissolved oxygen; post-harvest, shortly (within 1 month) after harvest and stream-cleaning; post-harvest1, in the summer following harvest.) Stream temperature was monitored continuously at all four sites.

	Year 1 					Year 2 					Year 3
Site 1 	light, DO, substrate, 			DO, substrate, 				DO, substrate,
	aquatic invertebrates 			aquatic invertebrates 			aquatic invertebrates

Site 2 	light

Site 3 	pre-harvest: light, DO,			post-harvest: light, DO 		post-harvest1: light,
	substrate, aquatic invertebrates						DO, substrate, aquatic invertebrates

Site 4 	pre-harvest: light, DO,			post-harvest1: light, DO,
	substrate, aquatic invertebrates	substrate, aquatic invertebrates
	post-harvest: light, DO

Fig. 2 Lighting at the water level and at bank level as a percentage of open light conditions at sites 3 and 4, before (pre), shortly after (post), and 1-year after harvest (post 1) compared with sites 1 and 2, the indigenous and mature pine forest sites.

Distributions of the data were investigated using normal probability plots, and data were log-transformed where appropriate to improve normality (total invertebrate densities). The effects of sampling occasion on invertebrate density, taxa richness, and %EPT (Ephemeroptera, Plecoptera, and Trichoptera) were investigated. The algal-piercing hydroptilid caddisfly Oxyethira albiceps was excluded from calculations of %EPT as high densities are indicative of degraded conditions. Community evenness (Pielou) and diversity (Margalef) were also calculated from the DIVERSE routine in Primer v.5.2.2 (Plymouth Marine Laboratory, United Kingdom). Differences between sites before harvest and within the native site over time and pre-and post-harvest differences within sites were investigated using Analysis of Variance, followed by an LSD test.

Patterns in community composition were examined with non-metric multidimensional scaling (NMS) using Primer based on the Bray-Curtis distance measure of percent abundance data. This analysis was conducted on data from individual Surber samples from each date to investigate spatial and temporal patterns within and among sites. The two-dimension solution for the ordination provided a stress value of 0.17. "Stress" is a measure of the departure from monotonicity in the relationship between dissimilarity in the original p-dimensional space and distance in the reduced k-dimensional ordination space (McCune & Grace 2002). Solutions with stress values in the range 0.1—0.2 are common in ecological studies, and provide a useable representation of the two-dimensional relationship among samples, although the details of the plot should be interpreted cautiously (Clarke & Warwick 2001). Differences between a priori defined groups (sites, pre- and post-harvest) were investigated using Analysis of Similarities (ANOSIM). Spearman rank correlations were conducted between scores for the first two ordination axes and the relative abundances of taxa that exceeded 5% of total numbers overall.

Site 1 (indigenous forest) had the lowest incidence of light reaching the stream channel (<5%; Fig. 2). Light levels were slightly higher (4—7%) in the mature pine control (site 2) and 8—13% at sites 3 and 4 pre-harvest. Light levels at site 3 increased after harvest to 62% at the water level and 90% at bank level, and there was little change 1-year post-harvest. The post-harvest incidence of light reaching the stream channel at site 4, which was narrower than site 3 (Table 1), increased to 30% at the water level and 44% at bank level, but light levels had decreased 1 year after harvest (Fig. 2). Before harvest, riparian vegetation contributed most of the shade to the water surface (89—99%) at all sites and there was little difference in light levels at the water surface and the upper channel banks. After harvest to the stream edge, followed by stream-cleaning, light levels at the water surface were 28% and 14% lower than at the bank top for sites 3 and 4, respectively. Channel banks provided most of the shade to the water surface, particularly in the narrower channel at site 4 but surrounding topography, and in the example of site 4, remnant mature pine and regenerating riparian vegetation, also contributed.

Fig. 3 Dissolved oxygen levels at A, site 3 and B, site 4 before (pre), shortly after (post), and one year after harvest (post1). Indigenous forest site (site 1) provides reference data; no reference data were available at site 3 for I-year after harvest.

Before harvest, DO levels at site 3 averaged 94% saturation, similar to site 1 (Fig. 3). Diurnal dissolved oxygen patterns appeared to be driven by temperature with low DO concentrations coinciding with peak water temperature, and little primary productivity (algae and macrophytes) was evident in these streams. Shortly after harvest, DO levels at site 3 dropped to an average of 71% saturation compared with 98% saturation at site 1, and were still primarily driven by temperature. By 1-year post-harvest there was little change in average DO levels at site 3 (75% saturation, no data available for site 1) but there was a strong diurnal amplitude of c. 18% saturation, indicating that this was driven by primary photo-synthetic processes rather than temperature. At site 4, pre-harvest DO levels averaged 72% saturation and were 24% lower than DO levels at site 1, indicative of strong instream oxygen demand (Fig. 3). Shortly after harvest, DO levels dropped even further to an average of 37% saturation, compared with 88% saturation at site 1. There was little diurnal activity and very low primary productivity, indicating the dominant role of respiration in driving DO patterns at this site. Dissolved oxygen levels the following summer (1-year post-harvest) had increased to an average of 81% saturation, 17% below site 1 (Fig. 3). There was a strong diurnal amplitude in DO patterns in the stream indicating that primary production was increasing DO concentrations to 100% saturation during the day, and respiration and/or temperatures apparently controlling reductions overnight.

Medium-large and large gravels dominated inorganic channel substrate in the indigenous forest reference site (site 1) throughout the study (Table 3).

Sand/silt comprised <20% of substrate elements, and moderate amounts of woody debris occurred in most years. Before harvest, substrate composition at site 3 was dominated by sand/silt and particles in the range small-medium gravel to small cobble. There was a 15% increase in the percentage abundance of small gravel and a decline in boulder abundance after harvest at this site, but otherwise differences in substrate size distribution were less than 10% compared to pre-harvest. Site 4 contained substantial amounts of sand/silt in the stream channel before harvest, much higher than that recorded at sites 1 and 3 (Table 3). Sand/silt levels decreased by 20% after harvest whereas wood increased in relative abundance.

To compare stream temperature between the harvested and reference sites, the indigenous forest site (site 1) was used as the reference site in year 2, because of insufficient data at site 2; and in year 3, the mature pine site (site 2) was used as the reference site because of insufficient data at site 1. During the three years of monitoring, mean monthly stream temperatures were similar at two reference sites (Fig. 4). Where comparable monthly data were available for sites 1 and 2 (n = 10, 8, and 7 for years 1–3 respectively), mean annual temperatures differed by 0.1°C in years 1 and 3 and by 0.3°C in year 2.

Table 3 Substrate composition at sites 3 and 4 before and 1 year after harvest compared with site 1, the indigenous forest reference site.

				          Site 1          Site 3     	    Site 4   .
				Year 1  Year 2	Year 3 	Before After 	Before	After
Sand/silt <2 mm 		18.8 	11.2 	11.0 	20.0 	18.0 	69.4 	49.0
Small gravel 2—8 mm 		3.8 	2.0 	12.0 	0.0 	15.0 	5.6 	11.0
Small-medium gravel 		10.5 	10.2 	11.0 	14.3 	10.0 	3.7 	8.0
Medium-large gravel 16—32 mm	29.3 	21.4 	19.0 	11.4 	3.0 	4.6 	2.0
Large gravel 32—64 mm  		19.5 	16.3 	13.0 	19.0 	16.0 	0.0 	0.0
Small cobble 64—128 mm 		4.5 	7.1 	6.0 	13.3 	23.0 	0.0 	0.0
Large cobble 128—256 mm		0.0 	1.0 	0.0 	6.7 	15.0 	0.0 	0.0
Boulder >256 mm 		0.0 	0.0 	0.0 	10.5 	0.0 	0.0 	0.0
Bedrock 			0.0 	1.0 	0.0 	1.9 	0.0 	0.0 	0.0
Small wood <100 mm 		9.8 	20.4 	18.0 	0.0 	0.0 	13.0 	23.0
Large wood >100 mm 		3.8 	9.2 	10.0 	2.9 	0.0 	3.7 	7.0

Fig. 4      A, mean; B, minimum; and C, maximum monthly stream temperatures at the Mahurangi Forest sites, Northland, New Zealand over 3 years. Horizontal lines indicate duration of catchment harvesting, arrows indicate when harvesting occurred alongside the study reach; -->. site 3, —  --> site 4.

Before harvest, mean annual stream temperatures were 12.7°C, 13.4°C, 13.8°C, and 13.9°C across sites 1–4 respectively. Following harvest, by year 3, mean annual temperatures at sites 3 and 4 had increased to 14.5°C and 15.1°C, respectively, and were 1.2°C and 1.8°C higher than the unharvested mature pine site (site 2). Where comparable monthly data were available, these increases were significant (paired t-test; P < 0.001, n = 10 for both sites). Although mean monthly stream temperatures increased at sites 3 and 4 following harvest, there was little variation in minimum monthly stream temperatures among sites during the three years of monitoring (Fig. 4). The lowest minimum temperature recorded was 5.4°C at sites 1 and 3 in year 2. Site 4 maintained consistently higher mean and minimum monthly temperatures throughout the 3-year period, probably because of its northerly aspect. Maximum monthly temperatures increased at sites 3 and 4 following harvest and by year 3 were averaging 20.2°C and 18.6°C at sites 3 and 4, 5.3°C and 3.6°C higher than the unharvested mature pine site. For comparable months, these differences were significant (paired t-test; P = <0.001, n = 10 and P = <0.001, n = 10 respectively). The highest maximum temperatures recorded in the two harvested sites were 26.3°C at site 3 in January, year 3 and 23.3°C at site 4 in December, year 2.

Mean diurnal temperature range during the summer months of January and February in year 1 were similar at sites 1, 2, and 4 (1.8°C, 1.4°C, and 1.5°C, respectively) and higher at site 3 (2.5°C), reflecting the wider channel at this site. Mean diurnal range increased at sites 3 and 4 following harvest and by year 3 averaged 7.1°C and 4.1°C, respectively, compared with 1.1°C at the mature pine reference site (site 2).

During the 3 years of monitoring at site 1, the indigenous forest reference site, there was no significant difference in aquatic invertebrate density, taxa richness, %EPT, community diversity (Margalef), and evenness (Pielou) (P = 0.26, 0.75, 0.25, 0.75, and 0.63 respectively). Aquatic invertebrate densities were lowest at site 1, the indigenous forest reference site (mean <800 0.1 m-2) and taxonomic richness and %EPT were stable at c. 34—38 taxa and 38—45%, respectively (Fig. 5). Before harvest, aquatic invertebrate density, taxa richness, %EPT, Margalef diversity, and Pielou evenness at site 3 were similar to site 1. However, with the exception of evenness

all these metrics were significantly lower at site 4 where sand and silt dominated the substrate (P < 0.05). Mean density increased and %EPT and Pielou evenness decreased significantly (P < 0.05) at both sites following harvest (Fig. 5) and although taxa richness increased at both sites following harvest, it was significant at site 3 only (P = 0.03; site 4, P = 0.07). At site 3, Margalef diversity increased after harvest (P = 0.03) but there was no significant change at site 4 (P = 0.07). All community metric values investigated were significantly different (P < 0.05) to site 1 in the same year for site 3 post-harvest, whereas at site 4 post-harvest values for density, %EPT, and Pielou were significantly different (P < 0.05) (Fig. 5).

The faunal composition (% abundance) at site 3 before harvest was very similar to site 1 (Table 4) where Ephemeroptera dominated, followed by Mollusca (mainly Potamopyrgus antipodarum) and chironomids. Faunal composition at site 4 before harvest was dominated by an even mix of mayflies, molluscs, oligochaetes, and other taxa (mainly Copepoda and Acari). Following harvest (1-year post-harvest), there were marked reductions in the percentage of mayflies at both sites, and increases in the percentage of Chironomidae at site 3 and Mollusca at site 4. The non-metric multidimensional scaling analysis indicated a broad spread among native forest samples along axis 2 that was not related to year of sampling, suggesting considerable heterogeneity in community composition over a small spatial scale within the reach (Fig. 6). Within-site variation appeared less at the pine forest sites, with site 4 clearly distinct in terms of community composition from site 3 or the native forest site before harvest. Sites 3 and 4 moved closer together in ordination space following harvest, but there was little difference at site 3 compared to pre-harvest community composition. ANOSIM indicated significant differences among all sites and dates (P < 0.01) except for the native sites where there was no significant difference among dates. Spearman rank correlations with the most abundant taxa indicated significant inverse correlations between axis 1 scores and relative abundance of the leptophlebiid mayflies Zephlebia and Deleatidium (r_s = -0.71 and -0.45, P < 0.001 and 0.05, respectively), and positive relationships with densities of the chironomid Tanytarsus (r_s = 0.41, P < 0.05) and the hydrobiid snail P. antipodarum. Axis 2 was also inversely correlated with Zephlebia densities (r_s = -0.41, P < 0.05) and positively with P. antipodarum densities (r_s = 0.64, P < 0.001).

Fig. 5 Aquatic invertebrate community density, taxa richness, percent Ephemeroptera, Plecoptera, and Trichoptera (%EPT), Margalef diversity and Peilou evenness, at sites 3 and 4 before and 1-year after harvest compared with site 1, the indigenous reference site sampled on three occasions. Error bars are SE.

Fig. 6 Non-metric multidimensional scaling plot using percentage abundance data for all samples collected at the native site (S1) in 2000, 2001, and 2002, and at the pine forest site 3 (S3) and site 4 (S4) pre- and post-harvest. Ellipses encompass pine forest samples sites collected on each date.

Removal of riparian vegetation during harvest operations and subsequent removal of most of the woody debris from the stream channel increased light levels reaching the streambed, elevated water temperatures, decreased DO levels, and altered the composition of the macroinvertebrate community. The responses of the two Northland streams to harvesting is similar to that in other studies in New Zealand (e.g., Graynoth 1979; Harding et al. 2000; Thompson 2001; Collier & Bowman 2003; Death et

al. 2003) and elsewhere (Hall & Lantz 1969; Campbell & Doeg 1989; Growns & Davis 1991; Davies & Nelson 1994), although the response magnitude varies.

Solar radiation is the most important factor influencing temperature regime in most small streams, although other factors such as the source of flow, interaction with the hyporeic zone, and ground water hydrology also contribute (Ward 1985; Story et al. 2003). Channel width, bank height, proximity and height of riparian vegetation, and the height of surrounding topography will all influence the degree of water surface exposure to direct solar radiation. Although riparian forest removal can have a strong influence on thermal regimes in small streams, in clear-cut catchments heating of subsurface and groundwater sources may potentially contribute to temperature increases. In small streams, post-harvest wood covering the channel can mitigate the effects of riparian vegetation removal on water temperatures by providing shade (Collier et al. 1997). In these two Northland streams, harvesting to the stream edge, followed by stream-cleaning, increased light levels at the water surface by 50% at site 3 and 22% at site 4, similar to post-harvest stream lighting changes recorded elsewhere (Davies-Colley & Quinn 1998; Robertson 1999; Thompson 2001). In the narrower stream at site 4, channel banks, rapidly regenerating riparian vegetation, and a residual area of mature pines all contributed to the lower post-harvest light levels recorded at this site.

Table 4 Percentage composition of the major aquatic invertebrate groups at sites 3 and 4 before and 1-year after harvest compared with site 1, the indigenous reference site (three sampling occasions combined). Note that, although Plecoptera were present, their relative abundance was low and for the purposes of this graph they have been lumped under the "Other" category.

		Site 1*	  Site 3 	Site 4
% abundance 		Pre   Post 	Pre   Post
Ephemeroptera 	36.1 	42.2   7.1 	18.3   3.8
Trichoptera 	3.6 	 3.6  13.1 	 1.8   0.9
Coleoptera 	8.7 	 1.5   0.8 	 0.8   1.1
Chironomidae 	12.6 	20.1  51.2 	 8.1   9.4
Other Diptera 	2.4 	 1.0   0.4 	 8.1  10.6
Mollusca 	27.4 	16.9  18.4 	18.1  61.5
Oligochacta 	3.5 	 7.0   7.8 	19.7   8.2
Other 		5.7 	 7.6   1.2 	25.2   4.5

	* 3 years pooled

The pre-harvest temperature regimes observed in this study are typical of small forested streams with predominantly overland flow, low summer base flow, and warm summer air temperatures (Graynoth 1979; Harding et al. 2000; Johnson & Jones 2000; Quinn & Kemp 2000). Clear-cut harvest to the stream edge can have a marked impact on stream temperature in these types of streams, where maxi-mum temperatures up to and exceeding 22–25°C, and diurnal fluctuations of 10°C to 11.5°C, have been recorded (Brown & Krygier 1970; Graynoth 1979; Collier et al. 1997; Harding et al. 2000; Johnson & Jones 2000; Quinn & Kemp 2000). This compares with maximum temperatures ranging from just over 14.0°C to 20.5°C, and diurnal fluctuations of 0.5°C to 2°C in cooler, spring-fed streams in the central North Island of New Zealand, following harvest (Pruden et al. 1990; Collier & Bowman 2003). Temperature responses in Northland are typical of those in streams in northern New Zealand with flow dominated by run-off. These streams lack the thermal stability and cooler temperature regime of spring-fed streams, and are more susceptible to impacts from loss of shade following riparian harvesting and channel wood removal. At the harvested sites, summer temperatures are sometimes reaching levels considered stressful to some of the more sensitive aquatic invertebrates (Quinn et al. 1994; Cox & Rutherford 2000).

Dissolved oxygen levels in streams can decline because of lower saturation at warmer temperatures, increased respiration from the breakdown of fine material remaining in the channel, and decreased reaeration rates where debris left in the channel impedes water flow (Ponce 1974; Hall & Lantz 1969; Pruden et al. 1990). Impeded water flow was not an issue at these sites as all the wood was removed from the stream channel at site 3 and only key structural pieces that were present before harvest were left behind at site 4. Initially, elevated temperature was the primary driver behind depressed oxygen levels at both harvested sites. However, fine organic material can create significant biological oxygen demand and reduce DO levels independent of temperature demand (Hall & Lantz 1969; Ponce 1974; Pruden et al. 1990; Collier & Bowman 2003) and high respiration from decomposition of pine needles remaining in the stream channel at site 4 following stream-cleaning was a likely contributor as well. In the Northland streams, it took up to 1 year after harvest for diurnal fluctuations in DO levels, initially driven by temperature changes, to become dominated by photosynthetic activity, most likely algal growth at both sites and some macrophytes at site 4. DO levels of 57–70% have been recorded in streams following harvest, although one study found no differences, except one site with high organicloadings where DO levels dropped to 75% (Pruden et al. 1990; Robertson 1999; Collier & Bowman 2003). Most of the DO levels recorded in this study were above the 50% saturation considered detrimental for the long-term survival of sensitive aquatic invertebrates (Nebeker 1972). Only site 4 recorded levels below this, and only shortly after harvest.

Harvesting and associated roading operations often increase sediment inputs into the stream (Graynoth 1979; Campbell & Doeg 1989; Kreutzweiser & Capell 2001; Collier & Bowman 2003; Death et al. 2003) and increased fine material can fill interstices in the substrate, altering the habitat for aquatic organisms (Ryan 1991). The Northland streams did not show significant increases in sand/ silt following harvest. However, site 4 already had pre-existing high levels of sand/silt.

The change in benthic invertebrate community composition following harvest is consistent with findings in most other studies in New Zealand that have examined the effects of land use on aquatic invertebrates (Graynoth 1979; Pruden et al. 1990; Harding et al. 2000; Thompson 2001; Collier & Bowman 2003; Death et al. 2003). Although results often vary depending on stream type and harvesting practices, particularly post-harvest wood treatment (Pruden et al. 1990; Robertson 1999; Thompson 2001; Collier & Bowman 2003), there is commonly a decrease in relative abundance of sensitive mayfly, caddisfly, and stonefly species and an increase in Diptera, Oligochaeta, Chironomidae, Mollusca, and more tolerant species of caddisflies such as Oxyethira albiceps. Indeed, relative abundances of common mayflies provided a good indicator of harvesting impacts, as has been observed elsewhere in New Zealand (K. J. Collier unpubl. data) and overseas (Davies & Nelson 1994).

Post-harvest increases in total invertebrate density largely reflected higher numbers of chironomids and molluscs, and were probably in response to increased growths of algae and degraded habitat conditions. Significant increases in taxa numbers and diversity may have been in response to a greater diversity of food resources (algae and allocthonous organic matter), and at site 3 at least to the moderating effect of the upstream native forest which would have provided a source of colonists. Growns & Davis (1991) found that differences in benthic invertebrate abundance and richness of clear-felled and undisturbed sites were obscured by differences between sites within streams 8 years after harvesting of Eucalyptus forest in south-western Australia. All sites in our study were clearly distinct in terms of community composition even before harvest, and site 4 had markedly lower density, %EPT, and diversity than the other sites reflecting the high levels of sand/silt present in this stream. Despite these differences, the response trajectories to harvesting were similar at both sites indicating that disturbance impacts can override within-site variation, at least soon after clear-felling.

CONCLUSIONS

The aim of this study was to investigate the short-term response of stream ecosystems to clear-cut harvesting followed by woody debris removal at two contrasting sites in Northland, New Zealand. These results demonstrate that impacts of harvesting to the stream edge followed by stream-cleaning, on light levels, temperature, DO, benthic substrate, and aquatic invertebrate communities can be influenced by prevailing in-stream conditions before harvest. Site 4 was less sensitive to harvest impacts because of pre-existing higher stream temperature and high levels of sand/silt. Regardless of the difference in percentage of catchment harvested, both sites experienced similar changes in water quality following harvest relative to an indigenous forest reference site, and a change in invertebrate community composition to one dominated by midges or molluscs. In fact, higher stream temperatures were recorded in the larger, partially-harvested catchment (site 3) indicating that stream size as well as pre-harvest habitat conditions were more important than percentage catchment harvested for some aspects of water quality. Whatever moderating influence the native headwaters and mature pines were conferring on stream temperature at site 3, this was soon lost once the shallow stream moved into open, high light conditions. These findings suggest that the percentage of stream edge harvested is more likely to be a better indicator of harvesting impacts on factors such as light levels, temperature, DO, and aquatic invertebrate composition, than percentage catchment harvested. However, intact forested headwaters are a potential source of aquatic invertebrates for downstream stream re-colonisation and may enhance recovery rates of sensitive taxa once stream habitat and water quality become suitable. There has been no significant recovery of these streams toward pre-harvest conditions, 1–2 years after harvest.

Shallow, low-order streams such as these, fed mainly by overland flow, are susceptible to temperature and DO changes when riparian vegetation isremoved. Harvesting and stream-cleaning have less pronounced effects on streams with more stable, cooler flow regimes, such as those in the central North Island fed by spring flows (e.g., Collier & Bowman 2003). It has been suggested that retention of moderate amounts of woody debris in small, spring-fed streams can moderate harvesting impacts (Collier & Smith 2003), but in these Northland streams retention time of woody debris is likely to be low and this would pose a risk to the downstream environment, given the frequency of high flow events observed during the trial period. The retention of riparian areas along the stream edge would assist in mitigating most of the adverse impacts observed in this study (Graynoth 1979; Thompson 2001). This comparison highlights the importance of considering the hydrological and landscape context when developing management plans for mitigating harvesting impacts on stream ecosystems.

We thank Carter Holt Harvey Forests for providing sites; Rob Davies-Colley, Kerry Costley (NIWA), and Graham Coker (Forest Research) for assistance with the shade measurements; Paul Lambert of NIWA for identification of the invertebrate samples; Mark Kimberley (Forest Re-search) for assistance with the statistical analysis; and Jon Harding, Canterbury University, John Stark, Cawthron Institute and two anonymous referees for reviewing and improving the manuscript. This research has been funded by the Foundation for Research, Science and Technology.

Baillie, B. R. 1999: Management of logging slash in streams of New Zealand—results of a survey. Project Report 85. Liro Forestry Solutions, Rotorua, New Zealand.

Bilby, R. E. 1984: Removal of woody debris may affect stream channel stability. Journal of Forestry 82(10): 609-613.

Bilby, R. E.; Likens, G. E. 1980: Importance of organic debris dams in the structure and function of stream ecosystems. Ecology 61(5): 1107-1113.

Brown, G. W.; Krygier, J. T. 1970: Effects of clear-cutting on stream temperature. Water Resources Research 6(4): 1133-1139.

Campbell, I. C.; Doeg, T. J. 1989: Impact of timber harvesting and production on streams: a review. Australian Journal of Marine and Freshwater Research 40: 519-539.

Chapman, M. A.; Lewis, M. H. 1976: An introduction to the freshwater Crustacea of New Zealand. Auckland, Collins.

Clarke, K. R.; Warwick, R. M. 2001: Change in marine communities: an approach to statistical analysis and interpretation, 2nd ed. United Kingdom, Ply-mouth Marine Lab.

Collier K.; Baillie, B. R.; Bowman, E.; Halliday, J.; Quinn, J.; Smith, B. 1997: Is wood in streams a dammed nuisance? Water & Atmosphere 5(3): 17-21.

Collier, K. J.; Bowman, E. J. 2003: Role of wood in pumice-bed streams I: impacts of post-harvest management on water quality, habitat and benthic invertebrates. Forest Ecology and Management 177: 243-259.

Collier, K. J.; Smith, B. J. 2003: Corrigendum to "Role of wood in pumice-bed streams II. Breakdown and colonisation" [Forest Ecology and Management 177: 261-276]. Forest Ecology and Management 181: 375-390.

Cox, T. J.; Rutherford, J. C. 2000: Predicting the effects of time-varying temperatures on stream invertebrate mortality. New Zealand Journal of Marine and Freshwater Research 34: 209-215.

Davies, P. E.; Nelson, M. 1994: Relationships between riparian buffer widths and the effects of logging on stream habitat, invertebrate community composition and fish abundance. Australian Journal of Marine and Freshwater Research 45: 1289-1305.

Davies-Colley, R. J.; Payne, G. W. 1998: Measuring stream shade. Journal of the North American Benthological Society 17(2): 250-260.

Davies-Colley, R. J.; Quinn, J. 1998: Stream lighting in five regions of North Island, New Zealand: control by channel size and riparian vegetation. New Zealand Journal of Marine and Freshwater Re-search 32: 591-605.

Death, R. G.; Baillie, B.; Fransen, P. 2003: Effect of Pinus radiata logging on stream invertebrate communities in Hawke's Bay, New Zealand. New Zealand Journal of Marine and Freshwater Re-search 37: 507-520.

Department of Scientific and Industrial Research, New Zealand 1961: Geological Map of New Zealand 1:250,000. Sheet 2A-Whangarei. New Zealand Geological Survey.

Fahey, B. D.; Rowe, L. K. 1992: Land use impacts. In: Mosley, P. M. ed. Waters of New Zealand. New Zealand Hydrological Society. Pp. 265-284.

Gordon, N. D.; McMahon, T. A.; Finlayson, B. L. 1992: Stream hydrology: an introduction for ecologists. Chichester, United Kingdom, Wiley. 526 p.

Graynoth, E. 1979: Effects of logging on stream environments and faunas in Nelson. New Zealand Journal of Marine and Freshwater Research 13: 79-109.

Growns, I. O.; Davis, J. A. 1991: Comparison of the macroinvertebrate communities in streams in logged and undisturbed catchments 8 years after harvesting. Australian Journal of Marine and Freshwater Research 42: 689-706.

Hall, J. D.; Lantz, R. L. 1969: Effects of logging on the habitat of coho salmon and cutthroat trout in coastal streams In: Northcote, T. G. ed. Symposium on Salmon and Trout in Streams. University of British Columbia, Vancouver, Canada. Pp. 355-375.

Harding, J. S.; Quinn, J. M.; Hickey, C. W. 2000: Effects of mining and production forestry. ln: Collier, K. J.; Winterbourn, M. J. ed. New Zealand stream invertebrates: ecology and implications for management. New Zealand Limnological Society. Pp. 230-259.

Hessell, J. W. D. 1988: The climate and weather of the Auckland region. Wellington, New Zealand Meteorological Service. 34 p.

Johnson, S. L.; Jones, J. A. 2000: Stream temperature responses to forest harvest and debris flows in western Cascades, Oregon. Canadian Journal of Fisheries and Aquatic Sciences 57 (Suppl. 2): 30-39.

Kreutzweiser, D. P.; Capell, S. S. 2001: Fine sediment deposition in streams after selective forest harvesting without riparian buffers. Canadian Journal of Forest Research 31: 2134-2142.

Leopold, L. B. 1970: An improved method for size distribution of stream bed gravel. Water Resources Research 6(5): 1357-1366.

McCune, B.; Grace, J. B. 2002: Analysis of ecological communities. Gleneden Beach, Oregon, United States, MjM Software Design.

Ministry of Works and Development, Water and Soil Division 1975: New Zealand Land Resource Inventory Worksheets, Warkworth Sheet N34. Wellington, Government Printer.

Nebeker, A. V. 1972: Effects of low oxygen concentrations on survival and emergence of aquatic in-sects. Transactions of the American Fisheries Society 4: 675-679.

NZFOA 2002: NZ forestry facts and figures 2002/2003. Wellington, New Zealand Forest Owners Association and Ministry of Agriculture and Forestry.

Ponce, S. L. 1974: The biochemical oxygen demand of finely divided logging debris in stream water. Water Resources Research 10(5): 983-988.

Pruden, C.; Coker, C.; Harding, J. S. 1990: Survey of water quality and invertebrate fauna in headwater streams, Matea Plateau Kaingaroa Forest. Forest Research Institute Contract Report FWE 90/8. 28 p.

Quinn, J. M.; Kemp, C. L. S. 2000: Effects of plantation forestry on visual water clarity and stream temperatures at Tairua Forest: review of seven year's monitoring incorporating the 2000 annual report. NIWA Client Report CHH00202/2.

Quinn, J. M.; Steele, G. L.; Hickey, C. W.; Vickers, M. L. 1994: Upper thermal tolerances of twelve New Zealand stream invertebrate species. New Zealand Journal of Marine and Freshwater Research 28: 391-397.

Robertson, G. 1999: Forestry stage impacts on stream ecology in the Wakapuaka River catchment, Nelson. Unpublished BSc Hons (Ecology) thesis, University of Otago, Dunedin, New Zealand.

Ryan, P. A. 1991: Environmental effects of sediment on New Zealand streams: a review. New Zealand Journal of Marine and Freshwater Research 25: 207-221.

Sabater, F.; Butturini, A.; Marti, E.; Munoz, I.; Romani, A.; Wray, J.; Sabater, S. 2000: Effects of riparian vegetation removal on nutrient retention in a Mediterranean stream. Journal of the North American Benthological Society 19(4): 609-620.

Smith, R. D.; Sidle, R. C.; Porter, P. E.; Noel, J. R. 1993: Effects of experimental removal of woody debris on the channel morphology of a forest, gravel-bed stream. Journal of Hydrology 152: 153-178.

Story, A.; Moore, R. D.; MacDonald, J. S. 2003: Stream temperatures in two shaded reaches below cutblocks and logging roads: downstream cooling linked to subsurface hydrology. Canadian Journal of Forest Research 33: 1383-1396.

Strahler, A. N. 1957: Quantitative analysis of watershed geomorphology. Transactions of the American Geophysical Union 38: 913-920.

Thompson, R. M. 2001: Effects of exotic pine forestry on stream community attributes. Unpublished PhD thesis, University of Otago, Dunedin, New Zealand.

Ward, J. V. 1985: Thermal characteristics of running waters. Hydrobiologia 125: 31-46.

Winterbourn, M. J. 1973: A guide to the freshwater Mollusca of New Zealand. Tuatara 20: 141-159.

Winterbourn, M. J.; Gregson, K. L. D.; Dolphin, C. H. 2000: Guide to the aquatic insects of New Zealand, 3rd ed. Bulletin of the Entomological Society of New Zealand 13. 102 p.