Continuous water quality monitoring to determine the cause of 
coral reef ecosystem degradation

Edward Laws, University of Hawaii

Abstract

Kaneohe and Waimanalo streams on the windward side of the island of Oahu in the 
Hawaiian Islands have been hardened to prevent flooding.  The hardening process has 
involved elimination of the natural riparian habitat and replacement of the natural stream 
channel with a concrete-lined conduit having vertical walls and a broad, flat bottom.  The 
shallow depth of the water column and the absence of shade have resulted in temperatures 
that average as much as 4-5oC above ambient and rise as high as 32oC during daylight hours.  
Unlike most low-order streams, the hardened sections of both streams are autotrophic, as 
evidenced by elevated pH values and O2 concentrations as high as 150% of saturation.  
Several allochthonous inputs, one from a storm sewer and the other from a natural spring, 
introduced water with anomalously low O2 concentrations and very high nitrate 
concentrations.  The absence of sediments in the hardened sections of the streams precludes 
natural sedimentary microbial processes, including denitrification.  Nitrate concentrations in 
a section of Waimanalo Stream with a natural streambed drop dramatically from values in 
excess of 400 ?M to concentrations less than 10 ?M at the head of the estuary.  Although 
some of this decline is due to dilution with seawater, the concentration of nitrate at the head 
of the estuary is only 10% of the value that could be explained by dilution effects.  Biological 
processes associated with a natural streambed thus appear very important to the functionality 
of the streams and in particular to their ability to process allochthonous nutrient inputs in a 
way that minimizes impacts on the nearshore environment.  Prevention of flooding can be 
accomplished by mechanisms that do not involve elimination of riparian buffer zones and 
destruction of channel habitat.  To maintain water quality and stream functionality, it will be 
important that these alternative methods of flood control be utilized.  Converting natural 
streams to storm sewers is an unenlightened way to address flooding problems.

Executive Summary

Kaneohe and Waimanalo streams on the windward side of the island of Oahu are natural 
streams, significant portions of which have been hardened in the name of flood control.  
Examination of the chemical and physical properties of these streams reveals that the natural 
functionality of the streams has been greatly modified by the hardening process.  The 
hardened reaches are autotrophic rather than heterotrophic; the temperature and pH of the 
water are much higher than ambient; and the absence of sediments has precluded metabolic 
processes such as denitrification that would normally transform nutrients and organic matter.  
More enlightened approaches to flood control could provide protection from flooding and at 
the same time retain stream functionality.

Purpose

The primary objective of the study was to quantify the effects of stream hardening on water 
quality and stream functionality.  

Approach

Water samples were taken at the locations shown in Fig. 1-2.  Waimanalo Stream was sampled 
along the Kahawai tributary and below the confluence of Kahawai and Waimanalo Streams near 
the mouth of the stream where it discharges into Waimanalo Bay.  Most of the Waimanalo 
Stream stations were sampled a total of 10-12 times at roughly 3-4 week time intervals during 
the period February-October, 2002.  Kaneohe Stream sampling was carried out at roughly three-
week intervals during the period June-November, 2002.  Most Kaneohe Stream stations were 
sampled a total of nine times.  
Waimanalo stations 2-5 and 7 lie along a hardened section of the stream that extends for a 
distance of approximately 0.8 km upstream and immediately downstream of Kalanianaole 
Highway.  Station 1 lies immediately upstream of the hardened section.  Station 6 is the effluent 
from an underground storm sewer that discharges beneath the Kalanianaole Highway bridge.  
Stations 7-9 lie at the beginning, midpoint, and end, respectively, of a stream restoration project 
carried out by the Waimanalo Watershed Project.  Station 10 is at the head of the Waimanalo 
Stream estuary.  In the Kaneohe Stream study, station 1 is located in a natural stream channel 
with no upstream hardening.  Station 5 is the effluent from a spring that seeps into Kamooalii 
Stream near the Likelike Highway culvert.  Station 10 is immediately downstream of the 
hardened section of the stream in the head of the Kaneohe Stream estuary.  The remaining 
stations are located along the hardened section of Kamooalii/Kaneohe Stream.
Water samples were collected in 250-ml plastic bottles and immediately placed in an ice 
chest.  Measurements of temperature, pH, oxygen concentration, and turbidity were made in the 
field.  Temperature was recorded to the nearest 0.1oC with a thermometer calibrated at 0oC (ice 
bath) and 100oC (boiling water).  Oxygen concentrations were recorded with a YSI model 58 
dissolved oxygen meter. pH was recorded to the nearest 0.1 using an IQ Scientific model 3000 
portable pH meter.  In the laboratory, the water samples were filtered through pre-weighed glass 
fiber filters (Whatman GFF) with a nominal porosity of 0.7 ?m.  The filters were dried in a 
drying oven at 105oC to constant weight.  The filters were weighed on a Mettler model H20T 
analytical balance to the nearest 0.01 mg.  Duplicates were run on random samples as a check on 
precision.  Blanks were run by filtering 250 mL of distilled water through a filter.  The weight of 
material collected on the filters ranged from a few milligrams to several tens of milligrams.  The 
blank correction was less than 0.1 mg.  The concentration of total suspended solids (TSS) was 
calculated from the difference in the weights of the filter before and after filtering.
The filtrate from the suspended solids filtration step was transferred to plastic bottles and 
processed for nutrient concentration measurements.  The filtrates were frozen if not immediately 
analyzed.  Concentrations of nitrate + nitrite (hereafter, nitrate), phosphate, and silicate were 
measured on the filtrate using colorimetric techniques on a Technicon Instruments 
AutoAnalyzer.  The procedures used for the colorimetric assays adhered to those described in 
APHA (1998).  Limits of detection were 0.5 ?M for silicate and 0.1 ?M for nitrate and 
phosphate.  Concentrations of total dissolved nitrogen (TDN) and total dissolved phosphorus 
(TDP) were determined by first oxidizing the filtrates with an Ace-Hanovia ultraviolet light 
photo-oxidation unit and then assaying for nitrate and phosphate, respectively.  Concentrations of 
particulate nitrogen (PN) and particulate phosphorus (PP) were calculated by assuming that the 
TSS contained 0.35% nitrogen and 0.11% phosphorus by weight (Laws and Ferentinos 2002).  
Concentrations of total nitrogen (TN) and total phosphorus (TP) were then calculated as TDN + 
PN and TDP + PP, respectively.

Findings

Before comparing results between stations, we first examined the distribution of data at each 
station to determine which parametric or nonparametric method would be best suited for the 
analysis.  HIDOH water quality criteria (HIDOH 2000) for nutrients and suspended solids are 
based in part on geometric mean concentrations, the rationale being that the data are best 
described by a log-normal distribution function.  If the data followed a log-normal distribution, 
the medians and geometric means would be identical.  To test this hypothesis, we compared the 
geometric means and medians of the nutrient and TSS concentrations at the Waimanalo (Table 1) 
and Kaneohe (Table 2) sampling stations.  Median silicate concentrations exceeded geometric 
mean silicate concentrations at 19 of 20 stations, a highly improbable result by random chance (p 
= 4x10-5 by a two-tailed test assuming a 50% chance that the median is greater than the 
geometric mean).  Median nitrate concentrations exceeded geometric mean nitrate concentrations 
at 17 of 20 stations, again a highly improbable result by random chance (p = 0.0026).  In the 
silicate and nitrate comparisons, the median values exceeded the geometric mean values by as 
much as a factor of 2.  Using the same logic, however, there was no significant difference 
between median and geometric mean concentrations of TN, TP, and TSS.  The best agreement 
between geometric means and medians occurred for TSS and TP, where the median equaled or 
exceeded the geometric mean at 8 and 10, respectively, of the 20 stations.  If the median and 
geometric mean were identical, one would be expected to exceed the other 50% of the time by 
random chance.  
	We concluded from this analysis that there was no single probability distribution function 
that characterized all the parameters we measured.  We therefore opted to use the non-parametric 
Kruskal-Wallis (KW) test (Sokal and Rohlf 1981) to determine whether there were significant 
differences in water quality between stations.  Based on the KW test, there was a highly 
significant difference in silicate concentrations between the 20 stations (p = 3x10-7).  The stations 
fell into three groups.  Group 1 consisted of all Kaneohe Stream stations, and Group 2 consisted 
of all Waimanalo Stream stations except station 10.  These two groups had geometric mean and 
median silicate concentrations of 238 and 313 ?M (Group 1) and 365 and 455 ?M (Group 2), 
respectively.  The KW test indicated that the silicate concentrations differed significantly 
between these two groups (p = 2x10-11).  Group 3 consisted of Waimanalo Stream station 10 
(Table 1).  Water at that station is brackish, and the silicate concentration is significantly reduced 
by the contribution of nearshore seawater, which has a silicate concentration less than 5 ?M 
(Laws et al. 1999). 
Based on the KW test, there was a highly significant difference (p < 10-8) in nitrate and 
TN concentrations between stations.  The stations logically fell into three groups.  The first 
group consisted of Waimanalo station 10 and all Kaneohe stations except station 5.  The second 
group consisted of all Waimanalo stations except station 10.  The third group consisted of a 
single station, Kaneohe station 5.  Geometric mean and median nitrate concentrations for these 
three groups are 6.7 and 9.8 ?M (Group 1), 429 and 456 ?M (Group 2), and 103 and 148 ?M 
(Group 3).  Corresponding TN concentrations are 22.1 and 22.0 ?M (Group 1), 572 and 642 ?M 
(Group 2), and 157 and 188 ?M (Group 3).  None of the geometric means satisfies HIDOH 
(2000) water quality criteria.
	TSS concentrations were significantly different between stations (p = 4x10-9).  In this 
case the stations logically separated into two groups, with the Waimanalo stations having the 
higher TSS concentrations (p = 4x10-8).  Geometric mean and median TSS concentrations at the 
ten Kaneohe stations were 4.2 and 4.1 mg L-1, respectively.  Corresponding values for the 
Waimanalo stations were 9.7 and 10.2 mg L-1, respectively.  Both geometric means satisfy 
HIDOH (2000) water quality criteria.
	There was a significant difference in TP concentrations between all stations (p = 8x10-4), 
and again between all Waimanalo stations and all Kaneohe stations considered as two groups (p 
= 0.0055).  Geometric mean and median TP concentrations at the ten Kaneohe stations were 0.48 
and 0.46 ?M, respectively.  Corresponding values for the Waimanalo stations were 0.72 and 0.75 
?M, respectively.  Both geometric means satisfy HIDOH (2000) water quality criteria.
	Figures 3-6 show box-and-whisker plots of temperature, pH, dissolved oxygen 
concentration (DO), and DO as a percent of saturation, respectively.  The boxes show lower 
quartile, median, and upper quartile values.  The whiskers show the range of values, except in the 
case of outliers, which are indicated by +'s.  Temperature in Luluku Stream (Kaneohe station 1) 
was 4-7oC cooler than at the other Kaneohe stations.  While stream hardening and the absence of 
shade along the course of Kamooalii/Kaneohe Stream undoubtedly contributed to this 
differential, an important additional factor is the temperature of the overflow from Kamooalii 
Reservoir.  Sampling of the water immediately downstream of the spillway revealed an average 
temperature of 25.0 with a standard deviation (SD) of  1.6oC.  However, median temperatures 
below the Likelike Highway were even higher (27-28oC) with excusions above 30oC.  In 
Waimanalo Stream the lowest median temperature was again at the station furthest upstream.  
Upper quartile temperatures rose steadily from stations 1 to station 5 and exceeded 28oC below 
station 2.  The discharge from the storm sewer (station 6) was anomalously low, but upper 
quartile temperatures in the stream channel again rose steadily from station 7 to station 9, and 
peak temperatures reached 32oC.  HIDOH (2000) water quality criteria specify that stream 
temperature shall not deviate by more than one degree Celsius from ambient conditions.  This 
condition is clearly violated in the hardened sections of both Kaneohe and Waimanalo streams.
	The pattern in pH values also reflected the influence of hardening.  In Kaneohe Stream 
the pH values in the stream channel increased steadily with distance downstream.  The lowest 
median pH values of 6.6 and 7.1 were recorded in Luluku Stream and the spring at station 5, 
respectively.  Upper quartile values in the stream channel exceeded 8.0 at all stations within the 
hardened section of the stream, and upper quartile pH values exceeded 8.5 at stations 7-9.  In 
Waimanalo Stream the median pH values in the stream channel also increased steadily from 
station 1 to station 5 and exceeded 8.0 at stations 4 and 5.  However, median pH values dropped 
along the restored steam channel from stations 7 to 9.  The increase in pH between stations 9 and 
10 probably reflects the influence of seawater at station 10.  The pH of surface seawater in the 
vicinity of Hawaii lies in the range 8.0-8.1 (JGOFS 2002).
	Oxygen concentrations displayed somewhat different patterns in Kaneohe and 
Waimanalo streams.  By far the lowest O2 concentrations occurred in the spring that seeps into 
Kamooalii Stream at station 5.  Median concentrations there were about 3 ppm (Fig. 5), which 
corresponds to less than 40% of saturation (Fig. 6).  Median O2 concentrations at the other 
Kaneohe stations were all above 80% saturation.  The HIDOH (2000) water quality standard for 
oxygen states that the O2 concentration should not be less than 80% of saturation.  Lower 
quartile O2 concentrations at stations 7 and 10 were both less than 80% of saturation.  In 
Waimanalo Stream, the O2 concentrations at station 1 were 5-6 ppm, lower than any other 
Waimanalo station.  These corresponded to about 70% of saturation.  The O2 concentrations rose 
dramatically between stations 1 and 2 and continued to rise to stations 4 and 5, where they 
averaged about 11 ppm or 140% of saturation.  O2 concentrations were anomalously low (~6 
ppm) in the storm sewer discharge and declined steadily between stations 7 and 10, i.e., 
downstream of the hardened section of the stream.  Median concentrations at stations 9 and 10 
were below the HIDOH 80% criterion.

Evaluation

The objectives of the project were attained.  Alterations to habitat as a result of stream hardening 
and land use in the Kaneohe and Waimanalo watersheds are evident in a variety of ways.  The 
temperature regime in the streams has clearly been altered by the absence of shade and wide, flat 
concrete concrete channels.  Stream temperatures upstream of the hardening were 22-24oC.  
Within the hardened sections, temperatures rose to 26-28oC (Fig. 3).  This is clearly a violation 
of HIDOH (2000) water quality criteria.  The wide, flat bottom of the hardened stream bottoms is 
certainly part of the problem.  At Waimanalo station 1 (natural channel bottom) the depth at the 
thalweg averaged 20  8 cm.  At stations 2-5 the corresponding depth was only 3  1 cm.  
Restoration efforts have created a low-flow channel below the Kalanianaole Highway culvert.  
The average water depth at stations 7-10 was 12  2 cm.  Nevertheless, the absence of shade 
resulted in water temperatures of 26-27oC at these latter stations.  A qualitatively similar physical 
situation exists along the hardened portion of Kamooalii/Kaneohe Stream.  The depth at the 
thalweg at Kaneohe station 1 averaged 24  10 cm.  At the hardened stations (2-4 and 6-10) the 
water column was only about half as deep.  The depth averaged 13  4 cm.
	Headwater streams are generally heterotrophic systems (Rankin et al. 1999, p. 14).  
Hawaiian streams are by default all low-order streams because of the short distance from their 
headwaters to their mouths.  One might therefore expect to find that Hawaiian streams consume 
oxygen and organic matter, and this appears to be the case along the section of Waimanalo 
Stream where there is a natural streambed, i.e., between stations 7 and 10 (Fig. 5).    The 
sediments are an obvious place for heterotrophic activity to occur, and the absence of sediments 
in a stream with a flat concrete bottom tends to shift the metabolism of the stream toward 
autotrophy.  In the case of Kaneohe and Waimanalo streams, this tendency is further stimulated 
by the absence of shade.
	Given the low O2 concentrations within stream sediments, the consumption of organic 
matter that occurs in the sediments is often accompanied by denitrification.  The importance of 
this process in the case of Waimanalo Stream is dramatically illustrated by examining the 
relationship between silicate and nitrate concentrations in hardened and natural sections of the 
stream (Fig. 7).  The straight line was drawn through median data from the hardened sections of 
the stream and the oceanic endpoint, which occurs at silicate and nitrate concentrations of about 
1.0 ?M and 0.01 ?M, respectively (JGOFS 2002, Laws et al. 1999).  If nitrate concentrations in 
the stream were reduced merely by dilution with seawater, all data would be expected to lie near 
the regression line.  In fact all data from the estuarine region (stations 10-12) lie well below the 
regression line, as do the median values from stations 7-9.  This pattern is very likely due to 
biological uptake of nitrate, with most of the uptake evidently occurring between stations 9 and 
10.  Although photosynthetic uptake of nitrate may well occur in the channel below Waimanalo 
station 7, the fact that O2 concentrations decline between stations 7 and 10 (Fig. 5) indicates that 
the system is heterotrophic, not autotrophic.  Denitrification is an obvious mechanism for 
heterotrophic metabolism to reduce nitrate concentrations, but denitrification can occur only in 
the virtual absence of O2.  Since the water column itself contains in excess of 4 ppm O2 (Fig. 5), 
the denitrification is very likely occurring in the sediments.  Because the hardened sections of the 
channels contain virtually no sediments, removal of nitrate by this mechanism is precluded.  
	The high nitrate concentrations in the hardened channels of Waimanalo Stream and to a 
much lesser extent Kaneohe Stream cannot be attributed entirely to the absence of stream 
sediments and associated metabolic processes.  Both streams experience remarkably high inputs 
of nitrate from allochthonous sources.  The storm sewer at station 6 in Waimanalo contains in 
excess of 600 ?M nitrate, and the spring that seeps into Kaneohe Stream at station 5 contains 
more than 100 ?M nitrate.  The ability of both streams to accommodate these inputs, however, is 
compromised by the absence of a natural streambed.
	The results of this project have been incorporated into a manuscript that has been 
submitted for publication to the journal Pacific Science.



		___________________________________   	____________________
				Edward Laws					Date



References

American Public Health Association. 1998. Standard Methods for the Examination of Water and 
Wastewater, 20th ed. Water Environment Federation, Alexandria, VA.
Hawaii Department of Health. 2000. Hawaii Administrative Rules. Title 11. Chapter 54. Water 
Quality Criteria.  http://www.state.hi.us/health/rules/11-54.
Laws, E. A., and L. Ferentinos. 2002. Human impacts on the fluxes of nutrients and sediment in 
Waimanalo Stream, Oahu, Hawaiian Islands. Pac. Sci. (in press).
Rankin, E., R. Miltner, C. Yoder, and D. Mishne. 1999. Association Between Nutrients, Habitat, 
and the Aquatic Biota in Ohio Rivers and Streams. Ohio EPA Tech. Bull. MAS/1999-1-
1, Ohio Environmental Protection Agency, Columbus, OH, 70 pp.





























Table 1.  Geometric mean (GM) and median (MN) concentrations of TN, nitrate, TP, TSS, and 
silicate at Waimanalo sampling stations.

TN (?M)
Nitrate(?M)
TP (?M)
TSS (mg/L)
Silicate (?M)
Station
GM
MN
GM
MN
GM
MN
GM
MN
GM
MN
1
442
406
364
375
0.81
0.67
13.5
15.2
318
435
2
552
621
422
434
0.43
0.36
7.2
8.3
381
467
3
538
530
406
428
0.38
0.39
7.7
6.1
336
425
4
644
704
472
489
0.39
0.35
7.2
7.4
296
417
5
713
738
484
460
0.83
0.97
13.6
8.9
307
450
6
762
759
602
616
0.45
0.23
0.98
1.02
427
611
7
573
646
443
400
1.22
1.20
20
18
371
455
8
557
670
422
444
1.68
2.10
27
30
379
455
9
498
519
404
364
1.10
1.10
11
20
488
483
10
26.6
27.4
5
9
0.91
0.68
20
21
63
144






Table 2.  Geometric mean (GM) and median (MN) concentrations of TN, nitrate, TP, TSS, and 
silicate at Kaneohe sampling stations.

TN (?M)
Nitrate(?M)
TP (?M)
TSS (mg/L)
Silicate (?M)
Station
GM
MN
GM
MN
GM
MN
GM
MN
GM
MN
1
15.3
15.5
5.6
6.8
0.63
1.18
2.2
2.3
224
407
2
20.9
21.3
3.6
7.1
0.56
0.76
4.5
6.4
290
330
3
26.3
23.4
5.3
10.5
0.44
0.49
3.1
2.6
199
311
4
28.8
29.0
10.4
13.6
0.41
0.40
4.4
3.4
199
310
5
157
188
103
148
1.31
1.65
7.6
11.7
296
366
6
24.3
22.1
12.9
16.4
0.30
0.29
2.8
2.6
205
305
7
23.4
22.0
10.3
11.0
0.30
0.32
4.0
4.2
199
258
8
23.1
24.4
7.7
8.8
0.48
0.41
6.5
7.0
273
331
9
21.1
21.5
10.3
14.3
0.41
0.44
4.2
4.1
303
346
10
16.3
15.1
2.8
7.1
0.49
0.49
5.4
4.2
190
260









 







Figure 1.  Waimanalo watershed and sampling stations.  Data collected from station 13 were not 
included in this analysis.







 




Figure 2.  Kaneohe watershed and sampling stations.  Data collected from stations 2a and 2b 
were not included in the analysis.






 





Figure 3.  Box and whisker plot of stream temperatures at Kaneohe and Waimanalo stations.  
Boxes indicate lower quartile, median, and upper quartile values.  Whiskers extend to lowest and 
highest values.  +'s are outliers. 










 


Figure 4.  Box and whisker plot of stream pH values at Kaneohe and Waimanalo stations.  Boxes 
indicate lower quartile, median, and upper quartile values.  Whiskers extend to lowest and 
highest values.  +'s are outliers. 








 


Figure 5.  Box and whisker plot of O2 concentrations at Kaneohe and Waimanalo stations.  
Boxes indicate lower quartile, median, and upper quartile values.  Whiskers extend to lowest and 
highest values.  +'s are outliers. 











 



Figure 6.  Box and whisker plot of stream oxygen percent saturation values at Kaneohe and 
Waimanalo stations.  Boxes indicate lower quartile, median, and upper quartile values.  Whiskers 
extend to lowest and highest values.  +'s are outliers. 






 




Figure 7. Silicate versus nitrate concentrations in Waimanalo Stream.  The straight line is a 
model II geometric mean regression line fit to the data from the hardened channel and the 
oceanic endpoint near the lower left-hand corner of the plot.  Data from the hardened stations 
and stations 7-9 are median values.  Data from stations 10-12 (estuary) are individual data points.	


