Effects of different land-use on suspended sediment dynamics in Sabah ( Malaysian Borneo ) – a view at the event and annual timescales

Suspended sediment concentrations (SSC) and the duration of high SSC are important for river ecology and water resource conservation. Using annual and storm-event datasets, this paper explores the hypothesis that key suspended sediment variables increase along a land-use disturbance gradient in hilly terrain in Sabah (Malaysian Borneo). Five small (1.7–4.6 km2) catchments of increasing disturbance history – primary forest, old growth virgin jungle reserve, twice-logged forest, multiple-logged forest and mature oil palm – were instrumented from late 2011 with dataloggers and sensors to record river stage, turbidity and rainfall. The oil palm catchment had 4–12 times greater mean dischargeweighted SSC (587 mg L–1), annual sediment yield (1128 t km–2 y–1), median event peak SSC, and duration of SSC above 1000 mg L–1 than in the other catchments. The multiple-logged catchment (last logged around 2004) has SSC characteristics close to values for primary forest, possibly due to increased ground protection against erosion afforded by low understorey regrowth and/or depletion of erodible sediment by multiple logging episodes. Results demonstrate that in hilly terrain even heavily logged rainforest has high value in safeguarding water quality and reducing erosion, whereas oil palm requires careful land management, especially of road runoff and ground cover.


BACKGROUND
Although there have been many studies on the effects of logging and alternative land-uses on suspended sediment yield (Douglas, 1999;Douglas et al., 1992Douglas et al., , 1999;;Walsh et al., 2011), impacts of repeat logging and forest rehabilitation have received little attention.Likewise, impacts of oil palm conversion and associated management practices on suspended sediment in different types of terrain have not been assessed.Roads are known to play a major role in generating sediment and (if poorly aligned) causing landslides in post-logging terrain (Gomi et al., 2006;Negishi et al., 2008;Walsh et al., 2011;Sidle and Ziegler, 2012), but their role in oil palm plantations remains unevaluated.This area of study is important because of downstream effects on sedimentation, water supply, river ecology and flooding.This paper addresses such research gaps by exploring the hypothesis that impacts on suspended sediment variables increase along a gradient of land disturbance.It does this by comparing the suspended sediment dynamics of five small headwater tributaries of increasing magnitude of forest disturbance history (primary forest, old growth-virgin jungle reserve, twicelogged forest, multiple-logged forest and oil palm) in the Brantian, Kalabakan and Segama river catchments in Sabah, Malaysian Borneo.The study forms part of the SAFE (Stability of Altered Forest Ecosystems) Project (Ewers et al., 2011) investigating the comparative impacts and benefits of retaining forest fragments and riparian strips of varying size when a rainforest landscape is converted to oil palm.The hydrological section of this project examines effects on discharge, suspended sediment, natural and pollutionimpacted water chemistry, channel morphology and sedimentation history, but this paper focuses on differences in suspended sediment with land-use and its history.

Catchment characteristics and land-use history
The five study catchments of differing land-use history (Table I) are located in the upper parts of the Brantian (Virgin Jungle Reserve -VJR, Twice-Logged Forest -LF2, Multiple-Logged Forest -LF3), Kalabakan (Oil Palm -OP) and Segama (Primary Forest -PF) river systems (Figure 1).The catchments were selected to be as similar as possible in terms of catchment area, topography, climate, geology and soil type, though there are some variations (Table I).The VJR, LF2, LF3 and OP catchments are located within the SAFE Project Area while the PF is located within the Danum Valley Conservation Area (DVCA).
The dominant geology of the LF2, LF3, OP and PF catchments comprises Oligocene to Middle Miocene rocks of the Kuamut (and Kalabakan) Formations, which consist of a melange of sedimentary and volcanic rocks, including slump breccia and interbedded mudstones, tuffs, tuffaceous sandstones, shale, conglomerate, chert and limestones (Walsh et al., 2011).The VJR catchment, however, is underlain by Cretaceous to Early Tertiary igneous and metamorphic rocks (mainly gabbro, dolerite, serpentinite, peridotite, dunite and pyroxenite).Topography (in terms of mean slope angle) is very steep in the VJR, steep in the two logged forest catchments and the oil palm catchment, but less steep (though still hilly) in the primary forest catchment (Table I).
The climate is equatorial.At Danum Valley Field Centre (close to the PF catchment), mean annual rainfall and tem-perature for July 1985-June 2016 were 2870.3 mm and 27°C respectively.Rainfall in the SAFE Project area has been recorded since July 2011 (Table II).Mean ratios of monthly rainfall totals at SAFE stations to that at Danum indicate that mean annual rainfalls for 1985-2016 at SAFE are 400-500 mm lower than at Danum, ranging from 2362.3 mm at Old Base Camp (just south of LF3) to 2514.4 mm at New Base Camp (just south of LF2) (Figure 1).Frequencies of heavy daily rainfall are also higher at Danum compared to SAFE Project sites (Table II).
Natural vegetation (Newbery et al., 1999;Singh et al., 2015) is mainly mixed lowland dipterocarp forest, but with some upland ultramafic forest in the more mountainous VJR catchment.Forest quality variables are shown in Table I.The primary forest catchment is in the Danum Valley Conservation Area, a Class One Primary Forest reserve within the Yayasan Sabah Forest Management Area.The Virgin Jungle Reserve catchment lies within a forest area (Figure 1) affected by light logging to make access roads in the past, but where the remaining forest resembles primary forest.The LF2 and LF3 catchments have undergone two or more rotations of selective logging.Both were first logged in the 1970s when an average of 113 m 3 ha -1 were extracted.A second rotation in the 1990s extracted an average of 37 m 3 ha -1 from LF2 (Struebig et al., 2013).The multiplelogged LF3 catchment, however, underwent three rounds of logging over 1990-2004, extracting 26, 22 and 18 m 3 ha -1 , respectively, with trees as small as 30 cm dbh being logged.Revegetation of LF3 since 2004 is dominated by young stands and low-lying understorey (Ewers et al., 2011;Struebig et al., 2013).The OP catchment lies is an area of mature (> 20 years old) oil palm (Elaeis guineensis) in Selangan Batu Estate.It is bench-terraced and has a dense network of roads and access tracks typical of oil palm plantations.There are no riparian forest buffers, as the estate was established before the legislation establishing a mandatory 30 m buffer width (Ewers et al., 2011).

Methodology and data collection
At the stream of each catchment, a gauging station was installed, comprising water depth, turbidity, water temperature and conductivity sensors and a tipping-bucket raingauge   linked to a solar-powered Campbell CR850 datalogger that records data at 5-minute intervals.Water depth (river stage) was converted to discharge using stage-discharge relationships derived using a combination of dilution gauging and (for high flows) the Manning's method.The PF catchment had a 120-degree V-notch weir.Turbidity values (T in NTUs) were converted to suspended sediment concentrations (SSC) using an equation (SSC = 0.674 T) derived previously for streams of the area (Sayer et al., 2006).The small conversion factor reflects the very fine particle size (modal diameter by weight = 0.1 μm) of the suspended sediment -typical of the humid tropics.
Results were explored at annual and storm event timescales.For the annual timescale approach, data from 1st January to 31st December 2012 were used.For the event timescale approach, 20 storms of differing peak discharge rank were selected in each catchment by means of a stratifiedsystematic method using storm data from the period January 2012 to August 2013.In this method, all storm events in which river stage rose by 2 cm or more for at least 15 minutes were ranked.The minimum, first-quartile, median, third-quartile and maximum storms (based on peak discharge) were identified.This yielded five storms and divided the list of storms into four groups.Additional storm events were then selected systematically from each quartile group -four storms from each of groups 1 to 3 and three storms from the uppermost quartile group (as the maximum event had been pre-selected).

Annual timescale
Annual timescale results are summarized in Table III.Mean arithmetic SSC and mean discharge-weighted SSC for the OP catchment (132 and 587 mg L -1 ) were 4-6 times higher than for the PF (34 and 102 mg L -1 ), VJR (24 and 50 mg L -1 ), LF2 (18 and 159 mg L -1 ) and LF3 (14 and 111 mg L -1 ) catchments.The near-pristine forest VJR stream had a higher mean arithmetic SSC but a lower mean discharge-weighted SSC than the logged-forest catchments.Mean arithmetic and discharge-weighted SSCs were not the lowest for the primary forest catchment; this reflected some very high storm SSCs in the primary forest record including the maximum recorded SSC in all catchments (Table III).Annual sediment yield was 4-12 times higher in the oil palm catchment (1128 t km -2 y -1 ) than the other four catchments even though annual water yield was the lowest (Table III).The PF sediment yield value is the average of several years' data derived from previous studies (Douglas et al., 1992;Greer et al., 1996), as gaps in the PF series led to insufficient data for reliable estimation of annual yield.The Q and SSC data were derived from all available 5-minute data over the period 1st January to 31st December 2012; the annual estimates were derived by multiplying a common three-month record (1st July-30th September 2012) by four.The superscripted letters H and L denote the highest and lowest values respectively.

Event timescale
Figure 2 summarizes variations between the catchments in peak SSC, storm sediment yield, end-SSC (which is the SSC recorded at the end of a storm hydrograph) and the duration of SSC above 1000 mg L -1 for the 20-storm datasets.Both median peak SSC and median event sediment yield were highest in OP followed by LF2, LF3, VJR and PF.The end-SSC and the duration of SSC above 1000 mg L -1 variables relate to by how much and for how long, respectively, a stream stays turbid after a storm (and were recorded for each storm event).The OP stream stayed turbid much longer after a storm than all the forested streams in all but the smallest-ranking events, as demonstrated by the lower quartile values of both variables for OP exceeding the even the maximum recorded values for the other catchments (Figure 2).

DISCUSSION AND SYNTHESIS
In the analysis both using the annual (Table III) and event (Figure 2) timescale approaches, values of all assessed suspended sediment variables were greatest for the mature oil palm catchment.Qualitative field observations suggested that these very high sediment variable values for the OP catchment may be due to a combination of (1) bench-terraced slopes characterized by little ground vegetation, a high percentage bare area and rather unstructured, subsoil-dominated soils, all of which lead to low infiltration capacities and enhanced overland flow and slopewash; (2) a higher density of roads and access tracks than in the selectively logged catchments, again promoting overland flow and erosion; and (3) gullying and enlargement of natural valley-side channel systems into which roadside ditches flow.Although it is known that the pre-existing logging road network was greatly added to during conversion to oil palm, the differences in road and track density between the study catchments were not quantified.Also the relative magnitudes of the three observed sediment sources listed above remain unknown.Although the mature oil palm trees provide an extensive but single-layer canopy cover (Fitzherbert et al., 2008) that gives some protection against light and brief rainfall, the rainfall threshold for producing significant streamflow and suspended sediment responses is much lower than under logged and primary forest.Thus rainstorms in 2012 as small as 6.2 mm on 5th July 2012 and 8.0 mm on 16th July produced peak SSCs of 835 and 478 mg L -1 respectively in the oil palm catchment, whereas rainstorms of at least 20 mm were required to generate such SSC responses in the other catchments.Also the high end-SSC and duration of SSC above 1000 mg L -1 values showed that the oil palm stream continues to receive elevated sediment input from slopes long after peak discharge has subsided.This may reflect delayed arrival of road or track drainage from upper slopes, but further fieldwork is needed to assess sediment source dynamics.
Annual sediment yields tended to increase with scale of land disturbance with the exception of LF3 (Table III).Annual sediment yield of the multiple-logged LF3 was 21% lower than that of twice-logged LF2 and only 15% higher than that of the VJR.Also end-SSC and duration of SSC above 1000 mg L -1 values for LF3 were lower than for LF2 (twice-logged) and were very close to the near-pristine VJR (Figure 2).The main reason may be that the dominance of young trees, promoting a dense cover of low-lying shrubs and herbaceous plants (especially ferns and gingers) in LF3 revegetation 7-10 years after the last of many logging cycles, which provides greater protection against erosion than the taller, more advanced regrowth and shaded and hence less dense understorey of LF2.The shorter tree, shrub and herbaceous cover of LF3 intercept raindrops just before hitting the soil thereby reducing the landing velocity and erosive power of the rainfall.A similar phenomenon of reduced erosion rates has been recorded elsewhere in Sabah by erosion monitoring on heavily disturbed terrain in phase with dense nearground vegetation during long-term recovery (Walsh et al., 2011).Also, Negishi et al. (2006) noted the role of fern cover in reducing erosion from abandoned logging roads in peninsular Malaysia.A second possible reason for lower sediment yield of LF3 may be that prior erosion of topsoil during multiple rounds of logging may have depleted the availability of easily erodible sediment in the catchment, though there was no evidence of extensive areas of bare rock or compacted subsoil to support this.
The twice-logged LF2, despite not having been logged since the 1990s, thus ranked as having the second highest values of mean discharge-weighted SSC, annual sediment yield and medians of peak SSC, storm sediment yield, end-SSC and duration of SSC above 1000 mg L -1 after the oil palm catchment (Table II and Figure 2).In the event-scale analysis of 20 storms, however, its high maximum values for recorded suspended sediment variables (Figure 2) were associated with a particularly large localised rainstorm (60.2 mm with a peak hourly intensity of 43.6 mm h -1 ) affecting the catchment on 18th July, 2012.Part of the reason for its high sediment variable values may therefore be due to the chance occurrence of the event in LF2 rather than in other catchments.Similarly the highest SSC (3934 mg L -1 ) in the annual dataset being recorded for the primary forest PF catchment may also reflect the chance occurrence of a particularly large rainstorm at Danum during the study period (Table II).Thus in the more integrative (and thus arguably more objective) event-scale approach, as it draws upon a greater range of storm events, the PF catchment ranked lowest for all suspended sediment variables (Figure 2).
The higher mean SSC in the VJR than in the logged forest catchments may be due to the particularly mountainous nature of the catchment and/or the different geology.Of particular importance may be its high mean channel gradient and slope angle, as elsewhere, gravity has been shown to exert a dominating effect on stream suspended sediment loads and minimise the effects of other influencing factors (Hagedorn and Whittier, 2015).Gomi and Sidle (2003) also stated that sediment dynamics of steep headwater channels differ from those of low gradient channels.In the event-scale results, however, median values of suspended sediment variables are lower than values for both logged catchments.
Comparison of SSC variables, and especially peak SSC, is best done by considering a wider range of individual storm events (as in the event-scale analysis), rather than focussing on the maximum value from the annual timescale dataset.This is because of the highly localised nature of most rainstorms in the area, such that exactly where the maximum storm occurs in a 1-2 year period is spatially random.Thus, using 20 storm events per catchment spanning the range of rainfall-runoff events, as done in this paper, provides a more effective comparison and avoids undue influence of a maximum event in an analysis.Computation of yields, however, is best done with continuously logged data.

CONCLUSION
Both the annual-timescale and event-timescale approaches show that suspended sediment transport and erosion are substantially greater in the mature oil palm catchment than in twice-logged and multiple-logged forest catchments.Thus mean discharge-weighted SSC and estimated annual sediment yield using the annual timescale approach were 3-6 times higher in oil palm than in the logged catchments.Results from the more integrative event-scale approach showed that (for storms of similar rank) the oil palm stream stayed turbid for the longest time after a storm event compared with the other four catchments.This is followed by the twice-logged forest, virgin jungle reserve, multiple-logged forest and the primary forest.The multiple-logged forest catchment (8-10 years after last being logged), because of its greater near-ground regrowth, appears to offer more protection against erosion than the twice-logged forest (18-20 years post-logging).This conclusion suggests that even multiple-logged forests should be conserved instead of being converted to oil palm, as they have high value from a land and water conservation aspect.Whether the adverse erosional impacts of oil palm plantations in hilly terrain indicated by these results can be reduced significantly by a combination of road-runoff diversion and cover-crop measures and riparian forest buffers remains unassessed and is a future research priority.Restricting future oil palm plantations to lower relief areas would clearly reduce sediment export significantly.
) and the VJR, LF2 and LF3 are located within the SAFE Project Area.b LAI -Leaf Area Index; AGB -Above Ground Biomass.The average forest cover (primary and secondary) in a 3 km radius surrounding second order sampling points described inEwers et al. (2011).cSingh et al. (2015).d Results from a plot study that may differ in technique compared to the rest(Newbery et al., 1999).

Figure 1 .
Figure 1.Locations of the VJR, LF2, LF3 and OP study catchments.The multiple-logged forest land containing the LF3 catchment (scheduled for conversion to oil palm in the near future) and the VJR, LF2 and OP catchments are all located in the SAFE (Stability of Altered Ecosystems) Project Area.The PF catchment is located in the Danum Valley Conservation Area (DVCA) 20 km north of the northern boundary of the SAFE Project Area and is not shown in the map

Figure 2 .
Figure 2. Box-and-whisker plots showing the minimum, first quartile, median, third quartile and maximum values of key storm-event suspended sediment variables for catchments of different land-use history.Datasets used are the twenty storm events selected for each catchment within the period January 2012 to August 2013 using the event-scale approach (see text)

Table I .
Catchment area, mean slope and forest quality variables of the study catchments a The PF catchment is known as the West Catchment in the Danum Valley Conservation Area.The OP catchment in the Selangan Batu Estate (Benta Wawasan Plantation

Table II .
Rainfall in the SAFE Project (VJR, LF2, LF3 and OP) and at Danum (PF)Annual rainfall in SAFE for the first two years were recorded at the old basecamp (just south of LF3) while latter three years were recorded from the new basecamp (just south of LF2).

Table III .
Specific discharge, suspended sediment concentration (SSC) and estimated annual water and sediment yields for the study catchments