[Next Section] [Previous Section] [Table of Contents]

Discussion

The initial analysis of data from this study examined the behavior of the measured variables through time. A description of variables by basin follows the discussion of responses in individual variables of peak (Q_p), volume (Q_v), and antecedent precipitation indices (APIs).

It is difficult to hold variables constant when analyzing basin behavior. This is particularly important for precipitation, which is a critical factor in stream flow. Statistical tests indicated there was no significant change in mean precipitation between the early and late periods, suggesting that differences in peak flows between the two periods were not attributable to differences in precipitation.

Although total precipitation throughout the entire period may not have changed, precipitation did vary monthly and yearly. These shorter term variations may have influenced individual peak flow events. The Mediterranean climate of the area is characterized by high rainfall and streamflows in winter and low rainfall and streamflows in summer. This rainfall pattern is the major factor in producing peak flows, and explains why peak flows greater than a recurrence interval of 1 year happen only in the months of November through May. The wettest months are December and January, and in each of the four basins, the highest number of peak flow events were in the January-February subset. This is probably because soils are closest to saturation and less water is able to infiltrate after precipitation events in these winter months.

Another influence of climate is the existence of periods of drought. In periods of drought, less water is available in the hydrological system of a basin. There is less input of precipitation and less groundwater, both of which could lead to lower streamflows. In this study, several years in the late period included a regional drought. This could have the effect of masking potential changes to peak flows made by urbanization. Under these conditions, the effects of urbanization could be reduced until groundwater is recharged. The late period in this study included a moderate and extended drought (Figure 5), but also several years of above average rainfall. Approximately an equal number of above average and below average rainfall years occurred during the late period, so the effects of drought were probably negligible in the study results.

Antecedent precipitation for five time intervals was measured for each peak flow event, to assess the effect of precipitation. By comparing each specific peak flow to the precipitation that preceded it (Q_p/API), a relationship between peak and precipitation can be established for the basin. As Q_p increases relative to APIs, the basin is responding with higher peak events to the same amounts of precipitation. This eliminates changes in precipitation as a potential cause of changes in Q_p or changes in Q_v. If that ratio increases after urbanization, it strongly suggests the effects of urbanization as a causative factor.

In the reference basin, the Luckiamute, there were isolated increased APIs. In the January-February subset, the increase in 90 and 120 days APIs indicated a need for higher rainfall to cause peak flows, which was corroborated by a decrease in the ratio of Q_p to 60, 90, and 120 day APIs. However, also in the January-February subset, the 15 day API decreased, suggesting a flashier basin response to that input period of precipitation.

In the Johnson Creek basin, all APIs decreased in the January-February subset, and several APIs decreased in the other subsets. However, the number of decreased APIs in the Johnson Creek basin, reflecting a variety of time periods, are evidence of a change in basin dynamics. In the Johnson Creek basin, APIs decreased in the subsets of RI 1 – 1.9, November-December, and January-February. These decreases were generally 15 to 30 percent and paralleled several other variables that also suggested an increasingly flashy basin response.

As the reference basin, it was not expected the Luckiamute would have changes in Q_p and there were none. If increased peak flows are a by-product of urbanization, a logical expectation could be that the most densely urbanized basin would have the largest increases in peak flows, and that effect would diminish with less dense populations. That was not the case in the suite of basins in this study.

Through the entire period of record, the only basin with an increase in the Q_p variable was the Newaukum, the least densely populated of the three urbanizing basins. Because precipitation has been eliminated as a variable, the most likely reason for this increase is urbanization. The Newaukum was also the basin with the strongest correlation between peak flows and time.

The Newaukum is unique in this study in other respects. It is the smallest of the urbanizing basins. It is the only basin to have been significantly affected by glaciation, leaving poorly developed soils with uneven conductivity and drainage characteristics. At the lowest elevations, perched water tables and annual flooding were common even before European development of the valley.

After a stable population from 1930 to 1970, the Newaukum basin population increased 60% over the next 25 years. To accommodate this population increase, most of which was clustered near the river, more land was drained for development. This land must be kept dry enough permanently to prevent damage to structures during peak flow events. These drainage structures route water much more quickly to the stream channel, thus contributing to increasingly high peak flows.

Two basins had changes in Q_p values in at least one data subset. The Newaukum had increases in Q_p in two subsets, January-February and RI 2–10 years. Each of these increases was 20 - 29%, similar to the increase for the entire flow period.

The mostly densely populated basin, Johnson Creek, had increases in Q_p in two subsets. Urbanization is the most likely explanation for these increases, but the lack of increase for the period of record as a whole is somewhat surprising, given that this is the most densely populated basin, and previous researchers have found increases in peak flow (Clement, 1984; Woodward-Clyde Consultants, 1995). This study used a longer flow record than previous studies, and some of the most recent years were in a drought period. This may be an example of lack of water in the basin system masking some signs of changed basin response. Other variables, such as Q_p/Q_v discussed below, did show a large increase in peak flows.

The Tualatin basin had no changes in Q_p, even with the second highest population density in the study. Though it was the largest basin in the study, it had the shallowest average slope, at only 7% (Table 3). This slope reflects the shape of the basin, with a broad valley floor. Most of the land in and near the floodplain has been kept in agriculture or parks, even in the most highly urbanized parts of the basin. Although wetland destruction continues, it may be that water is still allowed to pond in fields that become temporarily unusable but still available for three-season activities in agriculture or recreation.

An increase in this ratio is one of the clearest indications of change in basin response. In comparing Q_p directly with five periods of antecedent precipitation, so increase indicates a higher peak flow per amount of precipitation. As expected, the Luckiamute basin, as the reference, had no changes in this variable.

The Newaukum had an increase in the ratio of Q_p to all five APIs (18 - 22%). These increases roughly paralleled the increases seen in Q_p without considering antecedent precipitation. This is a clear indication that the increase in peak flows is due to changes in basin characteristics and not an artifact of climate variation.

Although the Johnson Creek basin did not have an increase in Q_p, there was an increase in the ratio of Q_p to the 30, 60, 90, and 120 day APIs (32 - 54%). These findings are not necessarily conflicting. The Q_p variable measures absolute increases in peaks; during a drought, peak events still occur but may be lower on average than in periods with more water entering the hydrologic system. The Q_p/API measures the response of the basin to specific precipitation events or series. Even in a period with lower absolute peak values, a basin might react to specific amounts of precipitation with increasing peak flows.

Through the entire period of record, the Tualatin had only one change in any variable, a decrease in the Q_p/90 day API. The reason for this single decrease in unclear, though it is consistent with a few other variables in this basin decreasing in the late period.

Three basins had changes in Q_p/API ratios. In the Newaukum basin, the only increase in this ratio was in the RI 2-10 subset, where the Q_p/90 day API and Q_p/120 day API ratios increased. It is difficult to understand how peak flows could increase over the entire period of record, but not increase in any subsets. The smaller sizes of the data subsets, however, make it difficult to meet the threshold of statistical significance.

Increased Q_p/API ratios were also found in the three data subsets (RI 1-1.9 years, November-December, and January-February) of the Johnson Creek basin, with increases of 10-65%. This is strong evidence of an increasingly flashy basin response.

Increases in Q_v could be another indication of changed basin response due to urbanization. As with the Q_p variable discussed above, however, basin values for Q_v were not consistent. Also, as with the Q_p variable, the Luckiamute basin had no changes in Q_v.

Only the Newaukum basin had a change in the Q_v variable through the complete period of record, an increase of 10%. The increase in Q_v is about half the increase of Q_p and the correlation is about half as strong, with an r value of less than 0.2.

Volumes increased in the Newaukum basin in two subsets, about the same percentage as for the entire flow record. Less definitive changes in Q_v were seen in the Johnson Creek basin. There was no change in Q_v through the period of record, but Q_v decreased 17-27% in two subsets.

Since both the Newaukum and Johnson Creek basins had increases in peak flows, the differing Q_v values in the two basins require different explanations. Urbanization can disrupt the natural system by creating artificial ponding areas. For example, roads, structures, and parking lots can create large areas where rain cannot infiltrate into groundwater, as it would in an undisturbed system. The trapped water must pond on the impervious surface until it can flow off the surface, is removed by storm sewers, or evaporates.

If the temporarily ponded water eventually flows off the surface, volume will tend to increase, since rain that originally would have entered the subsurface goes directly into a stream channel. If the ponded water is diverted to storm sewers, the effect on volume depends on where the storm sewers reintroduce water into the basin (or whether they divert water into another basin). If water evaporates, volumes will tend to decrease. It is unlikely that evaporation from temporary surface depression storage could take place on a large enough scale to greatly influence flow rates in large basins.

At the lowest elevations of the Newaukum basin in its natural state, the water table during winter months is at or near the surface. Urban development means that this ground must be kept dry enough to prevent damage to structures and roads. This has been done by tiling and draining the valley floor. With development, more land is drained and channeled to quickly remove water into stream channels, most likely explaining the volume increases in the basin.

In the Johnson Creek basin, storm sewers did not become widespread until the 1980s. This system diverted some of the runoff that would previously have gone into Johnson Creek and routed it into Portland’s combined storm sewer system that empties into the Willamette River. Hundreds of dry wells were introduced in the Johnson Creek basin in the 1980s, allowing a possible diversion from the surface to the subsurface. These two processes might explain why volumes around peak flows decreased. Another explanation might be a more general hydrologic equation. With a given amount of precipitation, if peak flows increase, there must be a corresponding decrease somewhere in the system. If the impervious surfaces in the Johnson Creek basin are moving water more quickly through the basin, some of the water that used to be part of a several day increase in volume is leaving as part of the increased instantaneous peak flow.

The Tualatin basin had no significant changes in Q_v through either the entire period of record or any subset. The correlation between Q_v and time was about the same in this basin as in the Newaukum basin (<0.20), which did have a change in Q_v values.

As with the Q_p/API ratio, the Q_v/API ratio compares a flow variable directly to five periods of antecedent precipitation. None of the four basins had a change in these ratios for the entire period of record.

All four basins had at least one change in Q_v in a data subset, though in three of the basins the changes were seen in only a few examples. The reference basin, the Luckiamute, had a lower Q_v/90 day API ratio in the RI 1-1.9 subset, and a lower Q_v/15 day API ratio in the November-December subset. The Tualatin had only one change, a decrease in the Q_v/15 day API in the January-February subset. The Newaukum basin had an increase in the ratios of Q_v to 90 and 120 day APIs in the January-February subset. It is difficult to explain such seemingly random changes, although they are each consistent with other variables that show similar changes in each basin.

In the Johnson Creek basin, each of the five Q_v/API ratios increased (32 to 70%) in the January-February subset. This is a completely different result than the decreases in Q_v seen in two subsets. Since this is the wettest part of the year, precipitation runoff may be forced to temporarily pond because of limitations of the stream channel, thus overriding the effect of storm sewers or increased overland flow.

This ratio allowed comparison of peak flows to event volume, whether a basin experienced changes in Q_v, Q_p, or both.

Johnson Creek was the only basin to have a change in the Q_p/Q_v ratio over the entire flow period, an increase of 20%. This finding reinforces findings by other researchers of increases in peak flows (Clement, 1984; Woodward-Clyde Consultants, 1995). The Q_p/Q_v ratio increased, even though neither Q_p nor Q_v changed significantly over this period. The r value for correlation of this ratio with the passage of time was 0.03, essentially uncorrelated. Correlation of flow variables with time may not be a useful exercise for most flow variables, since the process of urbanization is gradual, and the flow record is influenced by other factors such as precipitation.

It is noteworthy that the Newaukum basin, which had the largest changes in Q_p, did not have a significant change in the Q_p/Q_v ratio for the period of record. Because the Newaukum volumes increased as peaks did, the ratio remained stable. This underscores the importance of looking at basin variables in several different ways.

Although the Luckiamute basin had no changes in the Q_p/Q_v ratio in any subset, the other basins did.

The Q_p/Q_v ratio increased in two Johnson Creek subsets (32 - 44%). Considering that in these subsets, Johnson Creek peaks generally increased while volumes decreased, these increases in the Q_p/Q_v ratio are not surprising. Q_p/Q_v increases in Johnson Creek subsets bolster the argument that the increase in the Q_p/Q_v ratio during the entire record is a robust change in basin response.

There was a 23% increase Q_p/Q_v ratio in the Newaukum basin during January-February. This is the one period where Newaukum volumes did not increase; since peak flows did increase, an increase in this variable is logical.

The Tualatin basin had a 7% decrease in the January-February subset, when both the peak and volume variables decreased. It is unclear why this ratio should decrease in this basin, given that even the reference basin had no change, and the other urbanizing basins had some increases.

As with the individual variables, comparing the is a method of eliminating variations in precipitation from variations in peak flow. As the Q_p/Q_v/API ratio increases, it indicates that the basin is responding more quickly to equal amounts of precipitation.

The only basin where this ratio changed over the entire flow record was the Johnson Creek basin. Each of the five ratios increased in the later period (32-54%). The percentage increases in the (Q_p/Q_v)/API ratios are larger than the Q_p/Q_v values. This is an indication that the basin response in Q_p alone may have been masked by several years of below normal precipitation.

Three basins had at least one change in the (Q_p/Q_v)/API ratio.

In the Johnson Creek basin, this ratio increased for all five APIs in the RI 1-1.9 subset and the November-December subset, and several ratios in the January-February subset. Because the increases are seen in each of the subsets as well as the complete period of record, this is a strong indication of major changes in streamflow response in the basin.

The Luckiamute basin had a decrease of 8% in the (Q_p/Q_v)/90 day API ratio, and the Newaukum had a 34% increase in the (Q_p/Q_v)/120 day API ratio. It is difficult to understand how important these particular API period results are, but they are consistent with other variables showing decreases in the Luckiamute basin and increases in the Newaukum basin.

The Luckiamute is a typical basin of the western part of the Willamette Valley. It includes uplands of basalt and marine volcanics and sediments. The soil is generally well drained. The upland areas are still largely covered with forests, although logging continues, and lower elevations are largely covered by farms.

The low population density (less than 50 people/km²) would suggest that urbanization should not have affected the flows of this stream. Statistical tests show that over the period this basin has been gaged, there have not been significant changes in peak flows and volumes. There were, however, some decreases in the data subsets. In the RI 1-1.9 subset, there were decreases of 6-8% in the ratios of 90 day antecedent precipitation compared to Q_p, Q_v, and Q_p/Q_v. There was also a decrease in the Q_p/15 day API (8%) and Q_v/15 day API (16%) in the November-December subset. In the January-February subset, several variables decreased when compared to 90 and 120 day APIs. Although the reasons for the decreases are unknown, they are important, and may be worth further study.

The null hypotheses for this basin were that there have been no changes in peak flows (Q_p), partial event volume (Q_v), ratio of peak flow to partial volume (Q_p/Q_v), or ratio of these values to an antecedent precipitation index (Q_p/API, Q_v/API, (Q_p/Q_v)/API). None of the null hypotheses were rejected.

The Tualatin is the largest basin in the study and has the largest population, though not the highest population density. Urban development is concentrated in the eastern half of the basin, and population density will continue to increase in this area as planning goals are implemented. From 1930 to 1996, basin population increased 1000%; from just 1990 to 1996, basin population increased 8%. Although the basin as a whole has not reached the Census Bureau definition of urbanized, most of the development has occurred in the lower half of the basin. This substantially increases the effective population density in this part of the basin.

A moderate correlation (0.75) between Q_p and Q_v was seen in the Tualatin basin. Only a few variables, mostly in data subsets, indicate any change in variables between the early and late periods, and each changed variable indicates a decrease in peak flows. This pattern of a few decreases in stream measurements parallels that of the reference basin in this study.

Given the projected development in this basin, it is critical to determine whether the years of drought in the study has masked the effects of an increasingly dense population, or whether the soil and geographic characteristics and size of the basin will continue to be able to absorb changes to the natural hydrologic system. To better assess the effect of longer term precipitation trends, an additional API of 730 days was calculated. There were no significant differences in the 730 day API between the early and late period, and no difference between the two periods in the Q_p/730 API, Q_v/730 API, and (Q_p/Q_v)/730 API ratios. The lack of effect of 730 days of precipitation suggest that the drought played no substantial role in dampening peak flows in the post-urbanization period.

There is a perception among the public that floods in the Tualatin basin have increased (Pulaski, 1997). If flooding is defined as streamflows that affect human development, it may be reasonable to say that peak flows have not increased, but flooding has, due to development in the flood plain. There are more people in the basin to see water rising above its banks and there has been flooding in very visible areas, such as downtown Tualatin. This small city had a population of 193 in 1930, but increased to 20,040 in 1996 (US Census Bureau, 1998). Roads and parking lots that were covered by floodwaters in 1996 were in the news for several days. The same areas might well have been covered by water for an equal amount of time in previous years with few people noticing.

The null hypotheses for the Tualatin basin were that there have been no changes in peak flows (Q_p), partial event volume (Q_v), ratio of peak flow to partial volume (Q_p/Q_v), or ratio of these values to an antecedent precipitation index (Q_p/API, Q_v/API, (Q_p/Q_v)/API). None of the null hypotheses were rejected.

Soils of the Newaukum basin are different from that of the other basins in the study. Lying on the east side of the Puget Sound-Willamette Valley lowland, soils and structure are mostly determined by the composition and history of the North Cascade Range. The volcanic soils are well drained in the steep slopes of the upland, but not well drained on the valley floor. Glacial activity left a broad, flat floodplain for the river, but also left perched aquifers and impermeable soils. The water table along the floodplain is often above ground level during the winter. For any human development, whether agricultural, urban, or road building, this land must be kept dry throughout the year. For this reason, extensive tiling has been done in the Newaukum basin.

After a stable population in the basin for 40 years, population growth began in the 1970s. From 1970 to 1996, the population of the basin increased 59%. Most of the population lives in the lower half of the basin, increasing the effective density in that part of the basin.

R values show an expected correlation (0.70) between Q_p and Q_v. There is a weak correlation between time and Q_p (0.39), time and Q_v (0.17), or time and Q_p/Q_v (0.20). In each of these categories, values were significantly higher in the later period. Because no significant differences in precipitation were seen, these increases through time were the result of some other change in the basin.

Q_p values generally increased about 20%, with smaller increases in volume. These changes were seen in both the entire flow record and in data subsets. Q_p can be expected to continue to increase as the basin is further developed.

The Newaukum was the only basin without enough events in the RI 1-1.9 year to be tested, but was the only basin with an adequate number of events in the RI 2-10 year to be tested. It was also the only basin with consistently increased Q_v values. It is not clear how, or even if, these three statistics are related; this is an area for future study.

The null hypotheses for this basin were that there have been no changes in peak flows (Q_p), partial event volume (Q_v), ratio of peak flow to partial volume (Q_p/Q_v), or ratio of these values to an antecedent precipitation index (Q_p/API, Q_v/API, (Q_p/Q_v)/API). Several of the null hypotheses were rejected: Q_p increased 21%; Q_v increased 10%; and Q_p/all APIs increased about 20%.

Soils of the Johnson Creek basin are largely derived from the volcanic uplands around the basin, and from deposits left by the Columbia River. The volcanic uplands to the south are an area of low infiltration and faster runoff, providing most of the increase in stream flow after storm events in the natural hydrology of the basin. As the northern portion of the basin has been increasingly developed, however, the natural drainage systems have been disrupted by impervious surfaces replacing areas of former infiltration.

Johnson Creek is the smallest basin in the study, and has the densest population. At nearly 700 people/km², it is the only basin that meets the Census Bureau definition of urban. Because antecedent precipitation indices were lower in the later period, all increases in Q_p and ratios including Q_p must be related to some other factor, most likely population and concomitant development. R values show a moderate correlation (0.73) between Q_p and Q_v. When analyzing basin behavior through time, there is a weak correlation between time and Q_p (0.22), time and Q_v (0.20), or time and a Q_p/Q_v ratio (0.32). In each of these categories, values were higher in the later period.

Many variables increased in the later period. Among those variables are Q_p/Q_v; Q_p/API ratio for 30, 60, 90, and 120 day APIs; and all Q_p/Q_v/API ratios. Each of these results indicates a more flashy basin response. This result was confirmed by the behavior of the basin in the seasonal and event size subsets.

Findings of an increased flashy response are consistent with early studies showing a change in Johnson Creek’s behavior (Clement, 1984, Woodward-Clyde Consultants, 1995). Because almost the entire basin in within an area designated for medium to high density development, it is likely that peak flows will be an increasing problem unless significant measures are undertaken to mitigate the effects of development.

The null hypotheses for this study were that there have been no changes in peak flows (Q_p), partial event volume (Q_v), ratio of peak flow to partial volume (Q_p/Q_v), or ratio of these values to an antecedent precipitation index (Q_p/API, Q_v/API, (Q_p/Q_v)/API). Several of the null hypotheses were rejected: Q_p/Q_v increased 20%; Q_p/API for periods of 30, 60, 90 and 120 days increased 21% - 34%; (Q_p/Q_v)/API for all periods increased 32% - 54.

As the largest basin in the study, the Tualatin may be an example of the ability of large basins to better absorb changes to basin use than can small basins. Its population density is higher than that of the Newaukum, and it has had a significant reduction in its storage capacity because of wetlands reduction, yet has no increase in peak flows.

The Newaukum basin is about 15% the size of the Tualatin basin and five times larger than the Johnson Creek basin. The significant increase in peak flows and volumes in the Newaukum is surprising. It is the only basin in the study where these variables increased through the period of record. Because the geomorphology of this drainage contrasts with others in the study, it is difficult to determine if the effects of urbanization were so pronounced because of basin scale dynamics or a more vulnerable drainage system.

In the smallest basin, Johnson Creek, the ratio of peaks to volumes increased, even though neither peak nor volumes increased as individual variables, probably reflecting both a drought period and storm drainage modifications.

If the lack of response to urbanization in the Tualatin is due to basin scale dynamics, stream response may change with further urbanization that might lead to greater synchronicity among tributary streams.

[Next Section] [Previous Section] [Table of Contents]