Factors affecting long-term reduction in river flow in a large forested catchment in Aya, Miyazaki Prefecture, Japan

This study aimed to assess factors that determine changes in river water level in a relatively large forested catchment. Analysis of official river flow data for the Aya-minami River in Miyazaki, Japan, revealed a significant reduction for several years after 1970 and a prolonged reduction over 36 years in 95-day flow, 185-day flow, and 275-day flow; the climatic factors and changes in water use during this period had not contributed to this reduction. Forests cover approximately 90% of the 5,884 ha catchment area, and an analysis of tree ages showed that the clear-cut in the 1960s and 1970s most probably resulted in an elevated flow level around 1970, and that the tree growth thereafter have reduced discharges from the forest area. No statistically significant reduction trend was observed in the 355-day flow. Discharges from broad-leaved forests during periods of less precipitation tended to be higher than that of coniferous forests in small catchments, though we could not confirm that larger discharges from broad-leaved forests replenish the 355-day flow of the Aya-minami River.


INTRODUCTION
Paired catchment studies reveal that clear-cutting of forest trees generally increases water discharges (Bosch and Hewlett, 1982).These hydrological tests are exclusively performed in catchments smaller than 1,000 hectare (ha) (Stednick, 1996), and the results for these small catchments need to be validated in large catchments where many factors affect the runoff.In basins where forest cover regenerates after clear-cut, however, water yield has been shown to decline with increase in evapotranspiration from growing trees (Brown et al., 2005).Promoting extensive forestation therefore results in reduced water yields and has a negative effect on river water supply (Calder, 2005); studies involving larger catchment areas are required to confirm this observation.
During the post-war period in Japan, large areas within natural broad-leaved forests had been converted into coniferous plantations.It is quite likely that these growing coniferous trees are now affecting the river water levels throughout Japan, although this assumption has yet to be confirmed (Kuraji, 2010).Forest cover is referred to as "green dams" because of its function to retain water; the presence of forest cover is expected to decrease flood discharges and also to help increase water level during droughts (Kuraji and Hoyano, 2004).However, these contributions of forest cover also need to be quantitatively evaluated.The difference in the discharge levels from broadleaved forests and those from coniferous forests also need to be precisely compared (Murai, 1993).Soil water analysis in the Yoshino River catchment area suggests that water recharge and flood flow mitigation is enhanced in forests with abundant broad-leaved litter (Yoshinogawa Vision 21 Committee, 2004).
In the present study, we investigated the relation between long-term fluctuation of river flow and forest growth in a relatively large river basin for which official data on factors such as river flow, forest management records, and water use are available.We also observed multiple smaller catchments with differing vegetation cover, and studied whether tree type, i.e., evergreen broad-leaved or coniferous, affects discharge levels.

Geology, vegetation, and land use
We conducted our study in the basin of the Aya-minami River (officially Honjo River) in Miyazaki Prefecture, Japan.The Aya-minami and Aya-kita Rivers are tributaries of the Oyodo River, which flows east on the Kyushu Island into the Pacific.The geology of the study area is characterized by sedimental sandstones of the Hyuga Formation (Paleogene, Shimanto Formation) and the Miyazaki Formation.The average precipitation at the Aya-minami Dam for 1958-2008 is 2900 mm, generally showing higher amount in summer (June to September) and lower in winter (November to February).The potential vegetation for this region is warm-temperate evergreen broad-leaved forest.A vast area of natural forest had been replaced by Japanese cedar and Japanese cypress forests in the 1960s and 1970s; natural forests along the ridges and on steeper slopes escaped logging, and now evergreen broad-leaved forests and coniferous plantations are found in a mosaic pattern.The present study area involved a river reach and its catchment from the Aya-minami Bridge (N 31° 59' 37.0'', E 131° 14' 44.1'') up to the Aya-minami Dam (N 32° 3' 28.8'', E 131°7 ' 13.9''), which includes a tributary, the Takora River (Figure 1).The total catchment area is 5,884 ha, with forests covering 90% of the area (Table I).

Water use
The Aya-minami Dam was constructed in 1958 (catchment area, 10,100 ha).As shown in Figure 1, the impounded water is led to a power plant (outside the study area [outside]) through an aqueduct constructed under Mount Ohmori (1180 m in height), and is discharged into the Ayakita River.The water is returned to the Aya-minami River 18 km downstream from the dam (within the study area [within]) through the Yotsue Water Recovery Channel, which was built in 1958.A small portion of the impounded water is released to the Aya-minami River through the Oguchi-dani outlet [within], which was built 1.2 km downstream from the dam in 1996.Water withdrawn from the Takora River [within] is sent to the Aya-kita River [outside], and water withdrawn at the Minami Power Plant [within] is all returned to the Aya-minami River [within].Agricultural water withdrawn from the Yotsue Weir [within] flows over agricultural fields and returns to the Aya-minami River [within], and that withdrawn at Miyanotani [within] returns to the Aya-minami River, but downstream from the Aya-minami Bridge [outside].

Climate data, flow data, and related information
We obtained air temperature data  for Kobayashi, located approximately 15 km west of the Aya-minami Dam, from the website of the Japan Meteorological Agency (http://www.data.jma.go.jp/obd/ stats/etrn/index.php).Daily precipitation and the amount of water released from the Aya-minami Dam were retrieved from the Monthly Management Report for the Aya-minami Dam (1958Dam ( -2008) ) (Miyazaki Prefecture Civil Engineering Office).Daily river flow at the Aya-minami Bridge (1969, 1971-1973, 1976-1977, 1979, 1981-2005) and the daily withdrawal at the Minami Power Station (2004Station ( -2008) ) were provided by the Kyushu Regional Development Bureau.Daily water influx at the Yotsue Water Recovery Channel (2004Channel ( and 2006Channel ( -2008) ) and daily withdrawal at the Takora Weir (2007)(2008) were provided by the Miyazaki Prefectural Enterprise Bureau.Monthly agricultural withdrawal data at Miyanotani, from May to October (2007)(2008), were provided by the Aya Land Improvement Office (Aya Town Office).No official flow data was available for agricultural withdrawal from the Yotsue Weir, which began in the Edo period (Aya Town History Compiling Committee, 1972); hence the flow amount examined on June 12, 2009.
We also studied 24 streams discharged into the Ayaminami and Aya-kita Rivers and their tributaries, to determine the discharge levels from catchments with different vegetation compositions.The discharge levels were measured during periods when there was less precipitation; November 23 and 24, 2006;November 19 and 20, 2007;and November 3, 2008.We presumed that flow measurements of multiple streams in concert, during a period of less precipitation, would reflect the effects of catchment vegetation to an extent.The 7-day rainfall before the measurement was 29 mm in 2006, 0 mm in 2007, and 13 mm in 2008.We measured the width and depth of the streams, together with the flow rate by using a propeller current meter (VR-201; Kenek).The catchment area of the streams ranged in size from 9 to 417 ha.

Forest cover data and analysis
We obtained data from the Forest Resource Database (Kyushu Regional Forest Office), which was compiled in 2007, for the tree species in, forest age of, and area of national forests, and also obtained similar data for prefectural, municipal, and private forests from the Forest Database (Miyazaki Prefecture), which was compiled also in 2007.The areas of forests of the same age were summed up to determine forest age composition in 2007.In addition, forest age proportion were calculated for 2007, 2000, 1990, 1980, 1970, and 1960; the proportion for 2000 was estimated by subtracting 7 years from the age of each forest section in 2007, and those for 1990, 1980, 1970, and 1960 were calculated by subtracting 10 years from the age of each forest section in the later decades.Ages of the forests that existed before the logging for plantations were not recorded in the databases, and in these cases, we tentatively classified the forest ages before logging as "over 100 years".
The forest type composition of small catchments was also calculated from the data obtained from these forest databases by summing up the areas of forests of the same type, i.e., evergreen broad-leaved or coniferous.

Statistical analysis
Linear regression analyses were performed to determine the extent of increase or decrease in discharge level, and long-term statistical trends were evaluated using the Mann-Kendall test.

Long-term reduction in river flow
Annual flow amount of the Aya-minami River (Figure 2) showed a significant reduction trend (p < 0.05) when we considered all the official data from 1969 to 2005.However, the data set without the first three high values, i.e. 1969, 1971, and 1972, showed no long-term trend (p > 0.2).The mean (standard error [SE]) annual flow was 232 (19.5) × 10 6 m 3 .
The yearly changes in the 95th-day flow, 185th-day flow, 275th-day flow, and 355th-day flow water levels are shown in Figure 3. Linear regression analysis indicated a reduction in all water levels, and this reduction trend was statistically significant with respect to the 95th-day (p < 0.05), 185thday (p < 0.01), and 275th-day flows (p < 0.01) (Figure 3a, b, c); no significant trend was detected in 355th-day flow (p > 0.1) (Figure 3d).

Long-term changes in climatic factors
Although an increase in air temperature promotes evaporation from the ground surface and may thus reduce water yields, the mean annual temperature at Kobayashi showed only a slight elevation between 1976 and 2008 (0.04°C per 30 years), and this trend was not statistically significant (p > 0.1).
Precipitation at the Aya-minami Dam increased slightly between1965 and 2008 (Figure 4a), although the long-term trend of precipitation was not statistically significant (p > 0.1).
The mean annual total (SE) amount of water released from the Aya-minami Dam was 43.4 (5.2) × 10 6 m 3 , and comprised approximately 19% of the average annual river flow.It is often thought that the amount of water released from a dam constructed in the upper reaches of rivers governs the downstream flow levels (Kuraji, 2009).However, the impounded water of the Aya-minami Dam is normally led to the Aya-kita River for power generation, and direct release into the Aya-minami River occurs only in flood events; the release frequency vary from 0 to 51 days a year.Changes  in the yearly total release (Figure 4b) shows a slight increase as in the precipitation changes but with no significant longterm trend (p > 0.1).
These indicate that climatic factors are not responsible for the reduction in the annual river flow.

Water use by humans
The average (SE) annual river maintenance discharge from the Yotsue Water Restoration Channel was 83.3 (3.4) × 10 6 m 3 .This discharge comprised approximately one third of the mean annual river flow.We could not investigate for the presence of long-term trends because no annual discharge data before 2004 were available.However, the maximum flow rate of the channel is 5 m 3 s −1 , and even if the channel maintained its maximum flow throughout the year, the annual flow would only be 158 × 10 6 m 3 , which indicates that it is unlikely that the flow from the channel has greatly decreased to a level that can cause a large reduction in the river flow.Inflows from the Restoration Channel, however, showed a seasonal fluctuation; the average (SE) of the 355day flow was 0.7 (± 0.2) m 3 s −1 .On the other hand, the average (SE) annual inflow from the Oguchi-dani outlet was 5.1 (± 0.2) × 10 6 m 3 and comprised only 2% of the river flow.It did not show any long-term trends between 1996 and 2008 (p > 0.1).
The mean annual withdrawal at the Takora Weir was 10.9 × 10 6 m 3 , which comprised only 5% of the river flow.The mean (SD) annual withdrawal at the Minami Power Station was 27.1 (1.6) × 10 6 m 3 and comprised 12% of the river flow, though all of this water was returned to the Ayaminami River after power generation, and thus the withdrawal at this power station does not contribute to the reduction in the river flow.
The annual withdrawal from the Yotsue Weir, calculated from the measurements in June, a high season for agricultural use, was 17 × 10 6 m 3 , and the amount of water returned to the river was 5.6 × 10 6 m 3 .The water amount lost through evapotranspiration in the fields corresponded only to 4% of the river flow.The average annual withdrawal at Miyanotani was 1.3 × 10 6 m 3 and comprised only 0.6% of the river flow.
Thus the inflows and withdrawals due to human activities did not contribute to the reduction trend of the river flow.

Extensive forestation and clear-cutting
The age compositions of broad-leaved and coniferous forests in 2007 are shown in Figure 5.A large portion of the broad-leaved forests (99%) consisted of natural broadleaved forests that escaped logging, and of areas where broad-leaved forests have regenerated naturally after logging (Figure 5a).In contrast, a large portion of the coniferous forests (86%) consisted of cedar and cypress plantations established in the 1960s and 1970s (Figure 5b); small areas of pine and hemlock plantations and natural fir stands comprised the remaining portion.
The forestry procedures adopted in the 1960s and 1970s involved selective harvest of large trees, followed by clearcutting the remaining smaller trees, and then planting coniferous species (Teruha Forest Association, 2009).The areas where this procedure was used must have lost their forest cover for some time until the size of the planted trees considerably increased.We estimated the composition of   the forest ages between 1960 and 2007 (Figure 6), and notably, it was found that forest stands that were younger than 10 years occupied 25% of the study area in 1970.

Catchment vegetation and forest discharge
Discharges per unit catchment area from multiple small catchments are shown in Figure 7.Although this value varied according to the different proportion of evergreen broadleaved forest cover, the catchment that had a lesser amount of broad-leaved forests, i.e., catchments with higher proportion of coniferous forests, did not yield a large amount of water discharge.

Reduction in river water level caused by forest growth
It is known that annual discharges increase in catchments where more than 20% of the forest area is clear-logged (Stednick, 1996).Our present study showed that the establishment of extensive coniferous plantations in the 1960s resulted in young (1-10 years old) forest areas, which comprised over 25% of the total study area in 1970 (Figure 6).This suggested that extensive logging in the 1960s must have increased the annual flow of the Aya-minami River around 1970.
Increased annual flow due to logging is known to decline as young forest stands in the catchment grow and evapotranspiration from the forest area increases (Tsukamoto, 1992).The increased annual flow in the present study indeed declined several years after the extensive logging, though showed no significant trend thereafter.This may indicate that elevated discharges after logging terminates within several years.On the other hand, the 95thday, 185th-day, and 275th-day flows showed a long-term reduction trend even after 1973 (Figures 3b, c).We therefore suggest that prolonged decline in forest discharges, due to forest tree growth, is found more prominent in these water levels, particularly in 185th-day and 257th-day flows, rather than the annual flow.
Changes in discharges from forest areas are exclusively monitored in manageable experimental fields (Brown et al., 2005), and those in relatively large catchment area have only been confirmed in 1 area in the eastern US (Eschner and Satterlund, 1966).In the present study, we demonstrated that forest growth of a relatively large catchment of the Ayaminami River is responsible for the long-term reduction in river flow.
The drought water levels are affected by various factors; for example, increases in timber volume of the catchment (Ishii, 1997) and declines in precipitation during drought seasons (Kuraji and Shibano, 2002).Our concurrent observation of multiple small catchments, during the period of less precipitation, revealed that no large discharges are observed in streams from coniferous forests (Figure 7).It is reported that coniferous forests show higher evapotranspiration and discharge less water than do broad-leaved forests in eastern US (Swank and Douglass, 1974), while no evapotranspiration difference is confirmed between the 2 types of forests in Japan (Komatsu et al., 2007).In our study, the long-term reduction trend of the 355-day flow in the large catchment was not statistically significant (Figure 3d).Because broad-leaved forests occupy approximately half of the catchment area (Table I), we first assumed that larger discharges from broad-leaved forests during the drought period replenish river water level which had been reduced by vigorous coniferous growth.However, the 355day water level ± SE of the river (0.54 ± 0.07 m 3 /s) seems to be managed mostly by the inflows from the Yotsue Water Recovery Channel (0.73 ± 0.15 m 3 /s) and from the Oguchidani Outlet (0.14 ± 0.01 m 3 /s); no withdrawal was observed for 355-day flow at the Takora Weir and Miyanotani, and the withdrawal at the Minami Power Station was all returned to the river, hence neither affects the river flow.It was thus impossible from our study to detect the replenishing function of broad-leaved forests during the drought period.

CONCLUSION
Our present study revealed a significant water level reduction in the annual flow for several years after 1970, and long-term reductions in 95-day flow, 185-day flow, and 275-day flow of the Aya-minami River.We demonstrated that this reduction was not caused by changes in the climate or fluctuation in water uses by humans, but by the growth of conifer plantations in a catchment as large as 5,884 ha.Although discharges from small catchments during the drought season tended to be larger in streams with larger proportion of broad-leaved forests, we could not confirm this tendency in the large basin of the Aya-minami River.

Figure 1 .
Figure 1.River flow, influx channels, and withdrawals of the study area along the Aya-minami River.

Figure 2 .
Figure 2. Annual flow changes in the Aya-minami River from 1969 to 2005.

Figure 4 .
Figure 4. (a) Annual precipitation changes at the Aya-minami Dam, and (b) annual changes in total release from the dam to the Aya-minami River.

Figure 5 .
Figure 5. Forest-age distribution in 2007 based on tree age class: (a) broad-leaved forests; (b) coniferous plantations.One age class includes 5 consecutive tree ages.

Figure 7 .
Figure 7. Discharges per unit catchment area from multiple small catchments with different proportion of evergreen broad-leaved forest cover.

Table I .
Area and proportion of forest types * Residence areas, agricultural fields, roads, and riverbeds.