ESM 1 Trajectories of Community Change from 27 Published Papers

1

ESM 1-7 for “Disturbance and trajectory of changein a stream fish community over four decades”, Matthews et al. (Revised 13 March 2013)

ESM 1 –Trajectories of community change from 27 published papers:

Title: Disturbance and trajectory of changein a stream fish community over four decades

Authors: W. J. Matthews, E. Marsh-Matthews, R.C. Cashner, F. P. Gelwick.

Contact: ; Department of Biology, University of Oklahoma, Norman, OK 73019, USA

To evaluate utility of the hypothetical trajectories in Figure 1 of the published text, we reviewed a total of 71 multivariate trajectories appearing in 27 different published papers. Most trajectories were published in figures of those papers as temporal tracks on biplots from various multivariate approaches (CA, DCA, NMDS, RA, etc.), with lines connecting consecutive samples, although a few were tracks on only one axis. These papers covered a wide range of animal or plant taxa, and study types, from observational surveys in the field to studies reporting results of experimental or natural disturbances, and they ranged from a time scale of weeks to decades. But in all cases the authors used the multivariate trajectories to infer something about the time course of change in a community. For each trajectory, we estimated by eye the one of our hypothetical patterns to which it exhibited the closest fit. In a few cases we broke the trajectory as shown into two parts, such as periods with a clear “gradual” change versus a different period with a “saltatory” change followed by recovery.

Of the 71 temporal trajectories we reviewed (Table S1 of ESM 1, below), the majority showed at least some periods of saltatory or abrupt change in the position of the community in multivariate space. Because each multivariate plot we assessed was on a unique scale of axes, it was not possible to make a statistical assessment of “saltatory” versus “gradual”, but the existence of saltatory change was judged subjectively, i.e., if one or more segments in a trajectory appeared by eye to be markedly longer than most others. In 23 of the 71 trajectories we saw evidence of at least one saltatory directional change, followed by at least partial recovery to pre-change conditions (Pattern F of published Figure 1). Another 19 exhibited at least one period of directional saltatory change, without any evidence of return to a former state (Pattern E of Figure 1). Thus, a total of 42 of 71 (= 59%) of the temporal community trajectories showed evidence of saltatory change in at least one interval between samples. Five other trajectories showed evidence of at least one saltatory change, with without any overall directional change in the position of the community in multivariate space (Pattern D of Figure 1). In contrast, the other 29 community trajectories lacked what we judged to be saltatory change, i.e., change was by gradual increments, regardless of whether overall change was directional (15 cases = Pattern B of Figure 1) or not (6 cases = Pattern A of Figure 1), or whether there was any pattern of return toward an earlier state (three cases = Pattern C of Figure 1). With respect to overall directional change or not, 34 trajectories showed evidence, by eye, of movement away from the original community state, without any indication of a tendency to return, whereas 26 showed directional movement away from an original state, but with subsequent return toward the earlier community composition.

Table S1 (ESM 1) Summary of trajectories from 27 published papers:

References:

Adjeroud et al. (2009) Coral Reefs 28:775-780.

Allen & Skagen (1973) Br. Phycol. J. 8:267-287.

Anand & Orloci (1997) Envir. Ecol. Statistics 4:337-344.

Austin (1977) Vegetatio 35;165-175.

Barmuta (1989) Freshwater Biol. 21:223-236.

Bevilacqua et al. (2006) J. Anim. Ecol. 75:908-920.

Bloom (1980) Chapter 6 in: J. Cairnes (ed) Damaged Ecosystems, Ann Arbor Science.

Boulton et al. (1992) Ecology 73:2192-2207.

Brown & eske (1990) Oikos 59:290-302.

Collie et al. (2008) Can. J. Fish.Aquat. Sci. 65:1352-1365.

Currie et al. (2005) Estuarine Coast Shelf Sci. 63:315-331.

Daufresne et al. (2007) Oecologia 151:544-559.

Eby et al. (2003). Ecol. Apps.13:1566-1579.

Fox (1990) Oikos 59:321-329.

Fukami (2001) Oikos 92:215-224.

Gardner et al. (2006) New Zeal. J. Mar. & Freshwater Res. 40:29-42.

Hughes et al. (1990) J. Biogeography 17:291-306.

Magalhaes et al. (2007) Freshwater Biol. 52:1494-1510.

Muehlbauer et al. (2011) Hydrol. Earth Syst. Sci. 15:1771-1783.

O’Connell et al. (2004) Estaries 27:807-817.

Pegg & McClelland (2004) Ecol. Freshwater Fish 13:125-135.

Pyron et al. (2006) Freshwater Biol. 51:1789-1797.

Rodriguez et al. (2003) J. Veg. Sci. 14:433-440.

Sponseller et al. (2010) Global Change Biol. 16:2891-2900.

Van den Brink & Ter Braak (1999) Envir. Tox. Chem. 18:138-148.

Vieira et al. (2004) Freshwater Biol. 49:1243-1259.

Warwick et al. (2002) Mar. Ecol. Prog. Ser. 234:1-13.

ESM2-- Study area and events:

Title: Disturbance and trajectory of changein a stream fish community over four decades

Authors: W. J. Matthews, E. Marsh-Matthews, R.C. Cashner, F. P. Gelwick.

Contact: ; Department of Biology, University of Oklahoma, Norman, OK 73019, USA

Study area.--

Figure S1 Field survey sites at six fixed location from the lower mainstem (Site 6) to the headwaters (Site 1) of Brier Creek and one tributary, in western Marshall County, Oklahoma

Distances are from confluence of the flowing portion of Brier Creek with the bay now formed as part of Lake Texoma in what was formerly lower Brier Creek, inundated by the reservoir since 1947. Length of sampling reaches is approximate and has varied slightly among surveys depending on availability of water. Lengths include the total reach surveyed at a given site including areas within the reach that may have been dewatered, e.g., when the creek ceased flow in dry periods and comprised isolated pools. Depth = typical maximum depth of deepest pool in reach; width = typical maximum width of widest pool, at base flow. Descriptions of the sites are by WJM from a combination of his field notes from 13 surveys at all sites.

Site 1(34°04.25’ N; 96°49.94’W): Brier Creek , just south of Oklahoma State Highway 99C,17.2 km upstream of confluence with Lake Texoma. Sampling reach 200 m downstream from highway access. Most upstream site. Width 6 m and depth 0.8 m in one upstream pool below a box culvert, then the rest of the reach consisting of small pools or channel 1-1.5 m wide, not more than 0.5 m deep, with mud substrate. Incised approximately 0.5 m in open pasture, canopy 5 percent (in upstream pool).

Site 2 (34°02.50’ N; 96°49.55’W): Brier Creek at bridge on gravel county road13.5 km upstream of confluence with Lake Texoma. Sampling reach is from bridge to approximately 500 m downstream. Width 10 m; depth 1 m; incised approximately 1 m in earthern banks; substrate mud and gravel, with reaches of limestone bedrock. From pool under bridge downstream through a series of wide, shallow pools with small riparian canopy, then one additional pool 1 m deep, then long series of shallow runs (often dry) over limestone bedrock, ending in a wide, shallow sand-gravel bottomed pool exposed to sun and often choked with attached algae.

Site 3 (34°02.44’ N; 96°48.83’W): Un-named tributary of Brier Creek at bridge crossing on north-south county gravel road about 1 km east of main creek, 13.2 km upstream of confluence with Lake Texoma. Sampling reach from approximately 75 m upstream of bridge to as much as 325 m downstream of bridge (wherever water is found); total reach approximately 400 m. Depth = 1m; width = 6m; incised 1-2 m in earthern banks; bottom limestone rubble and cobble or bedrock, with mixed gravel; canopy 25 percent from small riparian trees. Most upstream reach ends at a pool below a sill formed by Cretaceous limestone, then a reach of very shallow pools to and below the county road bridge, then a series of shallow pools and small riffles, no additional deep pools. Site often consists of isolate pools during dry periods, sometimes completely dry.

Site 4 (34°01.56’ N; 96°49.33’W): Brier Creek on private property, 11.8 km upstream of confluence with Lake Texoma. Sampling reach 320 m downstream of broken concrete low-water bridge on abandoned county road, now closed to public. Width 10 m; depth 1.3 m; incised 1-2 m in earthern banks; canopy approximately 30 percent. Site 4 consists of a series of rock-cobble bottomed pools, then one gravel bottomed pool about 1.3 m deep, then a long reach of gravel-rubble shallows, ending in another gravel/mud-bottomed pool to about 1 m deep.

Site 5 (33°59.88’ N; 96°49.68’W): Brier Creek at Oklahoma State Highway 32 bridge,8.4 km upstream of confluence with Lake Texoma. Sampling reach from 100 m upstream of bridge to 250 m downstream of bridge, total reach 350 m. Width 12 m, depth 1.3 m; incised 2-3 m in earthern and sandstone banks; canopy approximately 60 percent from large riparian trees. Sampled reach includes one long upstream pool, then a gravel riffle, a large, deep pool under the bridge, then a series of gravel and bedrock riffles and narrow bedrock pools, with one additional deep pool near the lower end of the reach. Site 5 has perennial flow in all but driest periods, small seep springs are just upstream of this site.

Site 6 (33°57.17’ N; 96°50.49’W): Brier Creek at Powell Road bridge, 2.9 km upstream of confluence with Lake Texoma. Sampling reach from bridge to 400 m south (downstream). Most downstream site. Width 12 m; depth 1.5 m. Sampling reach includes one wide pool below the Powell Road bridge, three additional large, wide pools interspersed by flowing gravel riffles, and several long and narrow pools further downstream. Stream is incised within earthern banks approximately 3-5 m high, with near complete canopy of large riparian trees; large woody debris common at this site, some remaining from the 1981 flood.

Events.--

Table S2 (ESM2) Interval between surveys, event and documentation, number of events in interval (N), magnitude of events in interval (M), and Bray-Curtis distance (BCD) value for the interval.

Flood and drought events in Brier Creek were documented and their magnitude estimated from a combination of (1) direct observations and photographs of many of the events by WJM; (2) references in publications as noted; (3) rainfall records for Madill, Oklahoma (local NOAA weather observer to 1993; Oklahoma Mesonet station for Madill, 1994 to present); and (iv) Palmer Z short-term drought index values for the South-Central Oklahoma Climate Division. From 1979 to 1995 WJM lived near Brier Creek (Madill, Oklahoma, or OU Biological Station), and regularly visited the creek to document the effects of any large rainfall events with field notes and/or photographs. After 1995 WJM and EMM visited the creek and nearby streams during the severe droughts in 1998, 2000, and 2006, to document sizes of remaining pools in one or more stream reaches (e.g., Matthews and Marsh-Matthews, Copeia 2006:296-710; Marsh-Matthews and Matthews 2010, page 466 in Gido and Jackson, AFS Symposium 73). For periods when visits were not possible we determined occurrence from rainfall records.

Because some events were documented primarily on the basis on local rainfall, it is important to estimate the relationship between rainfall and discharge in Brier Creek. Brier Creek does not have a long-term USGS gauge, but it was gauged at State Highway 32 (our Site 5) from 1970 to 1974 as part of a small-stream gauging program (USGS Open File Report 81-824, 1981, “Rainfall-Runoff Hydrograph and Basin Characteristics Data for Small Streams in Oklahoma”. This report included 15 events from 1970 to 1974 in which discharge at the Brier Creek gauge exceeded 1,000 cfs (USGS uses the English system), along with the local rainfall for that day in 15 minute increments. We converted discharge in cfs to m3/sec, and plotted discharge against rainfall for that date, up to the time of maximum discharge (Figure S2, below):

Figure S2 Discharge (m3/sec) versus rainfall in one day (cm), at Brier Creek Site 5, 1970-1974

Normal, base-flow discharge of Brier Creek at or near Highway 32 is 0.1-0.2 m3/sec during much of the year (Power and Stewart, 1987, Am. Midl. Nat. 117:333-345), so it is apparent that a single day rainfall exceeding 3 cm can at times cause a major increase in discharge, i.e., erosive flooding. The relationship is quite variable, depending on antecedent rainfall, condition of soil, or time of year, but the overall correlation between daily rainfall and discharge was significant (r = 0.469, P = 0.039 one-tailed test). Therefore, for intervals when no direct observations were available, we considered a rainfall event exceeding 5 inches (12.7 cm), or exceeding 4 inches (= 10.2 cm) in one day to be a probable flood event and scored it as such in the table below.

One comment needs to be made regarding peak discharge values reported (above) from the USGS gauge, and peak discharge estimates made by us or our collaborators at the same site, based on non-gauged observations. Peak discharges reported by USGS on four occasions from 1970 to 1974 exceeded 100 m3/sec (converted from their cfs by a factor of x 0.0283), whereas the peak discharges calculated from flow measurements and stage observations x cross section estimates by our collaborators were much lower, in spite of the events being well-documented as highly erosive, torrential floods. Power and Stewart (1987) reported for 14 June 1983 an observed (by WJM) stage rise of 3.45 m, and a calculated peak discharge of 49.2 m3/sec. The 1983 flood coincided with reported rainfall in nearby Madill, Oklahoma, of 4.02 inches (= 10.2 cm). Harvey (1987,Trans Am Fish Soc116:851-855; ) reported for 5-6 June 1985 a stage rise of 3.5 m and a calculated peak discharge of 55 m3/sec, at which time rainfall reported for Madill was 4.35 inches (= 11.0 cm). Both of the floods reported by Power and Stewart (1987) and by Harvey (1987) were major erosive events, with severe scour of stream substrates, and bedload movements. We cannot specifically reconcile the differences in magnitude of discharge reported by UGSS (from relatively modest rainfall) and the discharges calculated by our collaborators, and the calculated discharge for floods in 1983 and 1985 may indeed have underestimated the discharge in the stream. But it is clear from either source that there is a substantial relationship between rainfall and flood events for Brier Creek.

We classified four events as extreme relative to all others from 1980-2008, including:

Drought (and extreme heat), summer 1980: This drought, which affected all of southern Oklahoma (Palmer Z-Index, South-Central Oklahoma Climate Division = - 3.90) was accompanied by extremely high temperatures, as documented by Ross et al. (1985, American Naturalist 126:24-40). Rainfall in Marshall County was far below average (Figure 1A of Ross et al. 1985), with only 0.47 inches (1.2 cm) during all of July and August 1980. Daily maximum temperature was far above average during that period, averaging 40.6 oC in July and 39.9oC in August (Ross et al. 1985). Although unofficial, WJM observed an actual temperature of 47oC in Madill, Oklahoma. Matthews et al. (1982, Southwestern Naturalist 27:216-217) recorded death of fish (orangethroat darters, Etheostoma spectabile) in Brier Creek from high temperatures when flow ceased at Site 5 and water reached 39oC. This drought eventually broke with a 3-day rain of 9.46 inches (24 cm) September 12-14, 1980, which resulted in flooding in the creek.

Flooding, October 1981: From 12-14 October 1981, official rainfall in Madill was 18.34 inches (46.6 cm) followed on October 16 by another 3.73 inches (9.5 cm). This total rain of 56 cm almost doubled any other rain event during the 40 years in our study, and includes the greatest daily maximum and monthly maximum for rain in the history of Marshall County ( These rains caused massive flooding of Brier Creek, substantially overtopping the banks and flooding the entire valley near Site 6, and leaving debris lines well up into the riparian forest (as documented by Ectachrome slides taken by WJM shortly thereafter). The wooden bridge at Site 6 (Powell Road) was washed away, and replaced subsequently by a new concrete structure. Many whole trees were transported downstream by this flood, and deposited within the reach of Site 6, where they remained intact for many years (Ectachromes by WJM in 1985), and to the present some of the larger logs remain submerged within Site 6). Although there have been numerous other severe erosive floods in Brier Creek during the 40 years, this event clearly exceeded all others in volume of rainfall and discharge in the creek. Unfortunately, no gauge was active at this time, so the actual maximum discharge is not known, but this event was clearly greater in magnitude then the floods documented in Power and Stewart (1987) and Harvey (1987) with almost three times the rainfall in comparison to those other major events.