Geomorphic Impact Assessment (Part 8 of 9)

Environmental Impact Assessment

Project Number: 52111-001 February 2020

Samoa: Alaoa Multi-purpose Dam Project

Volume 2: Geomorphic Impact Assessment (Part 8 of 9)

Prepared by Hydro-Electric Corporation for the Asian Development Bank.

This environmental impact assessment is a document of the borrower. The views expressed herein do not necessarily represent those of ADB's Board of Directors, Management, or staff, and may be preliminary in nature. Your attention is directed to the “terms of use” section on ADB’s website.

In preparing any country program or strategy, financing any project, or by making any designation of or reference to a particular territory or geographic area in this document, the Asian Development Bank does not intend to make any judgments as to the legal or other status of any territory or area.

ALAOA MULTIPURPOSE

DAM

Geomorphic Impact Assessment

26 June 2019

Prepared by Hydro -Electric Corporation ABN48 072 377 158 t/a Entura 89 Cambridge Park Drive, Cambridge TAS 7170 Australia

Entura inAustralia is certified to the latest version of ISO9001, ISO14001, and OHSAS18001.

Entura has prepared this document for the sole use of the client and for a specific purpose, as expressly stated in the document. Entura undertakes no duty nor accepts any responsibility to any third party not being the intended recipient of this document. The information contained in this document has been carefully compiled based on the client’s requirements and Entura’s experience, having regard to the assumptions that Entura can reasonably be expected to make in accordance with sound professional principles. Entura may also have relied on information provided by the client and/or other parties to prepare this document, some of which may not have been verified. Subject to the above conditions, Entura recommends this document should only be transmitted, reproduced or disseminated in its entirety.

- Revision No: 0 26 June 2019

Alaoa Multipurpose Dam Geomorphic Impact Assessment ConsultDM no.

1. Document information

Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019 2. Executive summary

This report used standard, up to date and appropriate data and methodologies to undertake the assessment. Topographic data were from a LiDAR (Light Detection and Ranging) survey. Terrain analysis, involving mapping of topography and slope, definition of sub-catchments and drainage lines, and calculation of reservoir capacity as a function of water level, was undertaken using GIS (Geographic Information System). Hydrology data were a modelled 48- year long hourly time-step discharge series at various locations throughout the catchment, and a water level series for the reservoir. The flow regimes of the modelled flow series were characterised using a range of standard hydrological statistics that covered the full range of flow components: minimum flows, low flows, baseflows, high flows and flood flows, and also characterised the seasonal distribution of flow components. A number of approaches were used to estimate suspended sediment load and bedload of the river, and rate of reservoir sedimentation. A model was developed for assessing reservoir shoreline erosion. This report took a standard 3-step approach to impact assessment, first characterising the existing environment, second, assessing the Project’s impact to the environment, and third, recommending mitigation measures and a monitoring program. 2.2 Existing environment

The topography, geology, soils, climate and hydrology of the Vaisigano River catchment were described in detail. Hydrological analysis revealed that the current regime is highly altered relative to the natural regime. The load of suspended sediment from the catchment of the proposed Alaoa Dam was estimated using a number of methods. The results covered a wide range, suggesting a high level of uncertainty in the estimate of sediment load. However, the results suggest that the Vaisigano River has a high specific suspended sediment yield by world standards. Bed material particle size was measured in the field, and these data were used in a model to predict the bedload transport rate. This established that bedload transport rates were relatively low. Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019

2.3 Key impacts of the proposed Alaoa Dam

The main potential impacts of the proposed dam and its operation on geomorphic processes and forms are:

• Trapping of sediment in the dam reservoir, reducing the load of sediment to the river downstream, and to Apia Bay;

• Scour of the shoreline of the reservoir rim due to focus of wind-generated waves at the normal full supply level (NFSL); i Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019 • Potential mortality of trees, shrubs and ground cover on hillslopes within the reservoir flood storage area during extended periods of inundation during high inflows, with subsequent risk of hillslope instability;

• The rate of reservoir water level drawdown on flood recessions exceeding the rate at which hillslope soils can drain, with subsequent risk of hillslope slumping; and

• Altered hydrology reducing the frequency of bed material mobilisation and reducing bed material load.

Each of these impacts were assessed using numerical models. 2.4 Mitigation and monitoring

The likelihood and consequence (risk) of each of the identified direct impacts of the proposed Alaoa Dam were assessed. The impacts were judged almost certain to occur, but the consequences were subjective and partly of an ecological nature. Reduced suspended sediment delivery to the river system downstream of the dam cannot be mitigated, but the geomorphic consequence of this is insignificant because nearly all of the sediment that would otherwise have passed through the river system would have entered Apia Bay and then flowed out to deeper water. Deposition of a large volume of fine sediment within the reservoir can only be mitigated by occasional flushing of the sediment by opening the low level offtake. This procedure was estimated to be required every 15 – 20 years, but the estimate is uncertain. The consequences of sediment flushing can be mitigated by arranging it to be done at a time when the catchment is experiencing a natural flood event, to assist transport of the turbid water to Apia Bay. Downstream of the Dam, the rate of bedload transport will be reduced due to reduced magnitude of flood peaks. However, it appears that the natural rate of bedload transport is low, so the consequences of a reduction are likely to be minor. With the dam operational, channel forming flows will operate at a lower magnitude, so the river will likely slowly adjust by contracting in width. The potential for mitigating this impact by environmental flows was investigated by Elvey and Gippel (2019). Their conclusion was that the most appropriate course of action was to allow the channel to contract and stabilise at a new dynamic equilibrium. Scour of the shoreline of the reservoir rim due to the action of wind waves is inevitable. Mitigation of this would be impractical, so it is recommended to allow the shoreline to erode to bare rock.

2.4.1 i i Submergence and waterlogging of trees, shrubs and ground cover on hillslopes within the reservoir flood storage area, as well as rapid drawdown, will create the risk of loss of soil and vegetation. It is recommended to seek additional expert advice on this matter. Mitigation could take the form of management of water levels and maintenance of good cover of vegetation that will tolerate the hydrologic and hydraulic conditions within the flood storage area. Monitoring of geomorphic form and process should focus on indicators that effectively characterise the main project impacts identified in this report, to determine if and to what extent the predicted impacts occur, and also the implemented mitigation measures, to determine the effectiveness or otherwise of the measures. The monitoring program would be a component of an adaptive approach to management of the dam and the Vaisigano River system.

ii Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019 Monitoring that will assist sediment and erosion management includes:

Annual sounding survey of the reservoir to determine sedimentosition dep rate

Surveillance of the reservoir flood storage area immediately following drawdown of raised water levels in the wet season. Initial survey could be done by AUV (aerial unmanned vehicle), with any areas of potential vegetation death or soil instability inspected on the ground.

iii

Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019 3. Contents

1. Introduction 1

2. Methodology and data 3 2.1 Morphology of Vaisigano River catchment and Apia Bay 3 2.2 Geology and soils 3 2.3 Climate 4 2.4 Hydrology 6 2.5 River sediment character, transport and deposition 8 2.5.1 Introduction 8 2.5.2 Suspended sediment load 9 2.5.3 Bedload 14 2.5.4 Hydraulic modelling 15 2.5.5 Bed material particle size distribution measurement 16 2.5.6 Reservoir sediment trap efficiency 17 2.5.7 Reservoir shoreline erosion 19 2.5.8 Stability of hillslopes within the reservoir 22

3. Description of the existing physical environment 26 3.1 Morphology of Vaisigano River catchment and Apia Bay 26 3.2 Geology and soils 33 3.3 Climate 38 3.4 Hydrology 40 3.5 River sediment character, transport and deposition 47 3.5.1 Suspended sediment load 47 3.5.2 Bed material particle size distribution 50 3.5.3 Bed material transport 52

4. Assessment of Project impact on the physical environment 55 4.1 Introduction to the impact assessment 55 4.2 Trapping of sediment in the reservoir 55 4.3 Scour of the shoreline at normal full supply level (NFSL) due to wind-generated waves 61 4.4 Duration of periods of inundation within the flood storage zone during high inflow events 62 4.5 Slumping of side slopes during reservoir drawdown 64 4.6 Altered hydrology and geomorphic response 66

5. Mitigation and monitoring 76 5.1 Mitigation 76 5.2 Monitoring 76

v Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019 6. References 80 Alaoa Multipurpose Dam ConsultDM no.

3.1 List of figures

Figure 1.1: Sub-catchments of the Vaisigano River catchment. 2

Figure 2.1: Location of key rainfall stations and a key hydrological model output node referred to in this report. 5

Figure 2.2: Relationships between sediment concentration and load and discharge indices for data from Fagaˈalu Stream, located southwest of Pago Pago Harbour, Tuitila Island, American Samoa. The plots are reproductions of data digitised from Figs 9 and 10 in Holst Rice et al. (2016). 11

Figure 2.3: Location of bed material sampling sites. 16

Figure 2.4: Three processes of geomorphic wave-shoreline interaction: a) undercutting; b) endstripping; c) over-wash. Source: Lorang and Stanford (1983). 21

Figure 3.1: Topography of Upolu, Samoa, showing location of Vaisigano River catchment. 27

Figure 3.2: Topography of Vaisigano River catchment. 28

Figure 3.3: Topography of the area in the vicinity of the proposed Alaoa Reservoir. 29

Figure 3.4: Slope of Vaisigano River catchment. 30

Figure 3.5: Slope of the area in the vicinity of the proposed Alaoa Reservoir. 31

Figure 3.6: Time series of selected historical aerial photographs, Apia Bay. Source: 1954, 1970 and 1987 extracted from Solomon (1994) and rectified; 2018 is World Imagery. Note the flood sediment plume in the Bay emerging from the Vaisigano River in the 2018 photograph. 32

Figure 3.7: Extract of the geological map of Upolu produced by New Zealand Geological Survey (1958) in association with the paper of Kear and Wood (1959). 34

Figure 3.8: Digitised version of part of the Provisional Soil Map of Upolu, Western Samoa (New Zealand Soil Bureau, 1956) covering the Vaisigano River catchment. 35

Figure 3.9: Digitised version of part of the Provisional Soil Map of Upolu, Western Samoa (New Zealand Soil Bureau, 1956) covering the area in the vicinity of the proposed Alaoa Reservoir. 36

Figure 3.10: Soil type key of the Provisional Soil Map of Upolu, Western Samoa (New Zealand Soil Bureau, vi Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019 1956). 37

Figure 3.11: Mean monthly rainfall distributions for three rainfall stations within or nearby to the Vaisigano River catchment. The stations Apia (2 m), Alaoa (260 m) and Afiamalu (798 m) are located at increasing elevations. 38

3.1.1 v i Figure 3.12: Annual rainfall distributions for five rainfall stations within or nearby to the Vaisigano River catchment. Tiavi and Mt Le Pue have each only two complete years of data. Lanafala data (from 1985) were simulated, not observed. 39

Figure 3.13: Distribution of wind speed by month (top) and by direction (bottom) for Lanafala (13.9°S 171.75°W 421 m asl). Based on hourly weather model simulations 1985 – 2019, data provided by meteoblue (www.meteoblue.com). 40

Figure 3.14: Mean daily flow calculated for each day of the year at 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios. Calculated from 48-year long hourly modelled time series. 42

Figure 3.15: Flow duration curve at 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios. Calculated from 48-year long hourly modelled time series. 42

Figure 3.16: Low flow indices (minimum hourly discharge, and flow exceeded for 95% of the time) and baseflow index (flow exceeded 50 percent of the time) for each month, at 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios. Calculated from 48-year long hourly modelled time series. 43

Figure 3.17: High flow indices (flow exceeded for 1% and 0.1% of the time) for each month, at 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios. Calculated from 48-year long hourly modelled time series. 44

Figure 3.18: Partial duration series flood frequency for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios, based on 4th order polynomial curves fitted to frequency data from 45 years of modelled peak daily discharge data (2013 – 2057). 45

Figure 3.19: Annual flow and annual peak hourly discharge for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios. Calculated from 48-year long hourly modelled time series. 46

vii Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019 Figure 3.20: Flow Health individual and combined flow deviation indicator scores for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled current scenario relative to the natural scenario. 46

Figure 3.21: Flow Health combined annual flow deviation indicator score for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled current series relative to the natural series. 47

Figure 3.22: Annual load of suspended sediment from the catchment of the proposed Alaoa Dam, estimated using three empirical equations from Holst Rice et al. (2016). 48

Figure 3.23: Particle size distributions of sampled stream bed material (excludes silt and sand material finer than 2 mm and exposed bedrock). 52

Figure 3.24: Estimated bedload transport rate across the full range of potential discharge over the 245 m long surveyed and modelled reach at Site 1b, 3.4 kilometres downstream Samasoni Weir. Bed elevation and water surface profile at 40 m3/s shown for reference. 53

Figure 3.25: Estimated annual bedload transport load and peak annual daily rate at Site 1b, 3.4 kilometres downstream Samasoni Weir for the natural and current scenarios. 54 Alaoa Multipurpose Dam ConsultDM no.

Figure 4.1: Proposed Alaoa Dam reservoir relationship between water level and volume. 56

Figure 4.2: Modelled daily time series of water level of proposed Alaoa Dam reservoir. 57 Figure 4.3: Mean water level for each day of the year (top) and water level duration curve (bottom) for the proposed Alaoa Dam reservoir, derived from the modelled 48-year hourly time series. 58

Figure 4.4: Daily Churchill trap efficiency (top), and annual Churchill trap efficiency compared with annual reservoir inflows (bottom). 59

Figure 4.5: Cumulative volume of sediment trapped in the reservoir, based on four estimates of annual sediment load. 60

Figure 4.6: Predicted extent and slope of reservoir wave erosion zone at normal full supply level. 61

Figure 4.7: Percent of years that water level was exceeded for the proposed Alaoa Dam reservoir, derived from the modelled 48-year hourly time series 62

Figure 4.8: Characteristics of spells of events exceeding a range of reservoir water levels. 63

Figure 4.9: Rates of fall exceeded 1% of the time (top) and 5% of the time (bottom) across the range of elevation of the flood storage zone to above the spillway crest at 174.6 m asl. 65

Figure 4.10: Rates of fall exceeded 25% of the time (top) and 50% of the time (bottom) across the range of elevation of the flood storage zone to above the spillway crest at 174.6 m asl. 66

viii Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019 Figure 4.11: Mean daily flow calculated for each day of the year at 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from 48-year long hourly modelled time series. 68

Figure 4.12: Flow duration curve at 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from 48-year long hourly modelled time series. 68

Figure 4.13: Low flow indices (minimum hourly discharge, and flow exceeded for 95% of the time) and baseflow index (flow exceeded 50 percent of the time) for each month, at 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from 48-year long hourly modelled time series. 69

Figure 4.14: High flow indices (flow exceeded for 1% and 0.1% of the time) for each month, at 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from 48-year long hourly modelled time series. 70

Figure 4.15: Partial duration series flood frequency for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios, based on 4th order polynomial curves fitted to frequency data from 45 years of modelled peak daily discharge data (2013 – 2057). 71

Figure 4.16: Annual flow and annual peak hourly discharge for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from 48- year long hourly modelled time series. 72

3.1.2 v iii Figure 4.17: Flow Health individual and combined flow deviation indicator scores for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled with Alaoa Dam scenario relative to the current scenario. 72

Figure 4.18: Flow Health combined annual flow deviation indicator score for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled with Alaoa Dam series relative to the current series, compared with the modelled current series relative to the natural series. 73

Figure 4.19: Estimated annual bedload transport load and peak annual daily rate at Site 1b, 3.4 kilometres downstream Samasoni Weir for the natural, current and with development scenarios. 74

ix Alaoa Multipurpose Dam - Geomorphic Impact Assessment Revision No: 0 ConsultDM no. 26 June 2019

Table 2.2: Hydraulic conductivity of volcanic Andisol soils reported in the literature.

Table 2.3: Hydraulic conductivity of volcanicOxisol (Latosol)soils reported in the literature.

Table 3.1: Estimated sediment load from catchment of proposed Alaoa Dam using Fournier equation, calculated for two different rainfall stations and twoassumptions for sediment specific gravity.

Table 3.2: Estimated sediment load from catchment of proposed Alaoa Dam and the entireVaisigano River catchment, using various equations.

Table 3.3: Characteristics of sampled stream bed material size distributions

Table 5.1: Potential geomorphic impacts and consequences of the proposed Alaoa Dam.

Table 5.2: Potential geomorphic impacts and suggested mitigations.

3.2 List of tables

Table 2.1: Locations and elevations of rainfall stations. 4

4. 1. Introduction

The Asian Development Bank (ADB) commissioned Entura and Fluvial Systems to carry out a geomorphological characterisation and impact assessment for the proposed Alaoa Multi- Purpose Dam Project, on the Vaisigano River, Samoa. The proposed dam site is located mid- catchment, capturing flow from two eastern branches of the headwaters, which together comprise 47.8 percent of the total catchment area (Figure 1.1). This report is intended to be incorporated into the overall social and environmental impact assessment (SEIA) for the project. The main purpose of the proposed Alaoa Dam is flood protection. The proposed dam crest elevation is 179.6 m asl (metres above sea level), the spillways crest is 174.6 m asl high, and the dam wall height is 59.6 m. To retain the capacity to store large inflows and prevent flooding downstream, the normal full supply level (NFSL) is 152.4 m asl, some 22 metres below the spillway level. The dam has been designed such that it will rarely spill. The dam is designed to release water to a new power station with maximum capacity of 1 m3/s, located about 350 m downstream of the dam, which is about 100 m downstream of the tailrace of the existing Alaoa Power Station; a water supply outlet to supply water to the existing water treatment plant during times when the power station is not operating; and, a mid-level outlet on the dam wall at NFSL is designed to release water at a maximum rate of 3 m3/s in order to prevent the reservoir rising into the flood storage during moderate river flows. The mid-level outlet will operate approximately 35 to 40 percent of the time annually. The dam has an emergency dewatering valve at the base of the dam with a 10 m3/s capacity that is not anticipated to be used during general operation. The project, as briefly described above, has implications for the geomorphology of the project area, the assessment of which is the topic of this report. In addition to the dam trapping the majority of inflowing sediment, the reservoir rim at the NFSL, and the hillslopes within the flood storage area, will be subject to the risk of erosion. Also, the dam will alter the hydrological and sedimentological regime of the Vaisigano River downstream of the dam to Apia Bay. Environmental flows to mitigate this impact were assessed and recommended by Elvey and Gippel (2019). Some of the information provided in this geomorphologic report was utilised in the environmental flow assessment. The objectives of this report are to describe the physical characteristics of the Vaisigano River catchment and Apia Bay under existing conditions, to assess the impacts of the proposed Alaoa Dam on those physical characteristics, and to recommend appropriate monitoring and mitigation measures. This report consists of the following sections:

• Section 1 - Introduction

• Section 2 – Methodology and data

• Section 3 – Description of the existing physical environment

• Section 4 – Assessment of Project impact on the physical environment

• Section 5 – Mitigation and monitoring

Figure 1 . 1 : Sub-catchments of the Vaisigano River catchment.

5. 2. Methodology and data 5.1 2.1 Morphology of Vaisigano River catchment and Apia Bay Characterising the morphology (topography) of the Vaisigano River catchment is important to this geomorphologic impact assessment as it determines the characteristics of the drainage network and conditions the source, transfer and depositional geomorphic zones of the catchment. The topography and vegetation cover influence the process of sediment delivery from the hillslopes to the stream system. Topographic and bathymetric Airborne Light Detection and Ranging (LiDAR) was flown for both the main islands of Samoa, Upolu and Savaii as a separate project by 2.5. The LiDAR survey data was collected by Fugro LADS Corporation Pty Ltd (FLC) and Fugro Geospatial Services (FGS) from flights conducted over the period 6 July to 9 August 2015. A total of 15

flights were flown for the bathymetric LiDAR survey and a total of 46 flights were flown for the topographic LiDAR survey. The vertical accuracy of the LiDAR survey is ±0.5 m (Entura, 2018). MNRE made available a 1 m grid Digital Elevation Model (DEM) derived from this LiDAR survey, covering the entire Vaisigano River catchment area.The model comprises a branched network of nodes and links, representing the subcatchments and stream reached respectively. The Vaisigano basin was subdivided into subcatchments of roughly equal area and within which hydrometeorological conditions are similar. These sub-areas are connected together in a mathematical network which simulates the actual stream network with excess rainfall as the input and streamflow as the output. The sub-catchments were derived using the GIS Digital Elevation Model from the available LiDAR survey data. Terrain analysis, involving mapping of topography and slope, definition of sub-catchments and drainage lines, and calculation of reservoir capacity as a function of water level, was undertaken using Global Mapper™ V20.1.0 Feb 25 2019 Build (Blue Marble Geographics) GIS (Geographic Information System). The drainage network was automatically generated at a resolution of 1 m by flow accumulation using the standard 8-direction pour point algorithm (D-8) (Jenson and Domingue, 1988). The reservoir capacity – water level relationship was calculated at 0.1 m increments. Understanding the geomorphology of Apia Bay is important to this geomorphologic impact assessment as the ecological and social values of Apia Bay, particularly in the harbour, are partly conditioned by the inflow of water and sediment from the Vaisigano River. The main resource used was Solomon (1994), who used air photo interpretation, literature review and ground survey to develop an understanding of the geomorphic processes influencing the coastal and nearshore sedimentary systems of Apia Bay. Estimates of sediment loads to Apia Bay by Richmond (1992) and Rubin (1984) were also referred to.

5.2 2.2 Geology and soils Understanding the geology of the Vaisigano River catchment is important to this environmental flow assessment as it determines the depth and erodibility of the soil on hillslopes and the character of the sediment transported by the river system. Sediment transport processes influence ecological processes. The geology of Samoa was described by Kear and Wood (1959), Kear (1967) and Keating (1992). A geological map of Upolu produced by New Zealand Geological Survey (1958) in association with the paper of Kear and Wood (1959). The soils of Upolu were mapped and described by the Provisional Soil Map of Upolu, Western Samoa (New Zealand Soil Bureau, 1956). 5.3 2.3 Climate The two aspects of climate referred to in this geomorphologic report were rainfall, and wind. Rainfall data were available for six locations (Table 2.1, Figure 2.1). Daily rainfall data, used in assessment of sediment load, were available from Alaoa from 1958 to 2016 and Apia from 1923 to 2017. Tiavi and Mt Le Pue stations had short patchy records. Mean monthly rainfall data, based on records from 2010 to 2017, were available for Afiamalu station (Elvey and Gippel, 2019). Also available was a simulated hourly rainfall time series over the period 1985 – 2019 obtained from meteoblue AG (www.meteoblue.com) for Lanafala (13.9°S 171.75°W 421 m asl), located 2 km upstream of the proposed Alaoa Dam, within the eastern branch catchment.

Figure 2 . 1 : Location of key rainfall stations and a key hydrological model output node referred to in this report.

Understanding the characteristics of the wind environment is important to characterisation of Project impacts as the reservoir created by the Alaoa Dam would be subjected to the effect -of wind generated waves impinging on the shoreline.

The two main wind seasons in Samoa are:

• June - October where trade winds prevail more than 50 % of the time (east and southeast winds).

• November - May where winds of wider directional distribution occur. The traditional Samoan names for wind provided by Samoa Meteorology Division (http://www.samet.gov.ws/index.php/climate-of-samoa) are:

• Winds from the North are called Toelau

• Winds from the South are called Toga

• Winds from the West are called La'i

• Winds from the East are called Mataupolu

• Winds from the North-West called La'i Toelau

• Winds from the North-East called Vaitoelau

• Winds from the South-West called La'i Toga

• Winds from the South-East called Tuaoloa

Solomon (1994) provided a wind rose for Western Samoa from Holden (1991), based on reports from ships in the region from 1951 - 1980. Solomon (1994) summarised the wind climate as follows. The regional wind climate is dominated by the easterly trade winds, with winds from the east, southeast and northeast comprising more than 70% of the total winds in the region. The average wind speeds from these three directions is about 20 km/h. Winds at Apia of less than 5 km/h prevail about 38% (possibly an error, with 3.8% more likely) of the time while winds between 5 and 23 km/h prevail about 50% of the time. Winds above 49 km/h occur less than 0.5% of the time (Solomon, 1994). More recent and detailed simulated hourly wind data over the period 1985 – 2019 were obtained from meteoblue AG (www.meteoblue.com) for Lanafala (13.9°S 171.75°W 421 m asl), located 2 km upstream of the proposed Alaoa Dam, within the eastern branch catchment.

5.4 2.4 Hydrology Understanding the hydrology of the Vaisigano River catchment is important to this geomorphologic impact assessment as the flow regime of the river is a determinant of the sediment transport process. In order to estimate design inflows and outflows from the Alaoa Dam as well as downstream flows along the Vaisigano River, an AWBM rainfall-runoff hydrological model was developed for the catchment by Entura (2018). The model used a 51-year data series of 1-hour rainfall values derived from seven years of observed hourly rainfall (2011 to 2018) and the remainder synthesised hourly data, as described by Entura (2018). The hydrological model was calibrated to the flow record at Alaoa East river gauge. The hydrological model simulated the operation of the current Alaoa Power Station as well as the proposed Alaoa Reservoir and Power Station and the inflow requirements for the currently operating Water Treatment Plant. The model network was also modified to take account of canal offtakes from the river network. To simulate the operational storage effects, the proposed Alaoa Reservoir storage was divided conceptually into an active flood storage zone up to the spillway crest at 174.6 m asl, an active hydro storage zone between 152.3 and 152.4 m asl, a water supply storage zone, and a dead storage zone for sediment accumulation of 0.5 × 106 m3 (Entura, 2018). The final runoff model output comprised a 48-year long hourly time-step discharge series at various locations throughout the catchment, and a water level series for the reservoir, beginning 1 September 2010. The water year was assumed to start on 1 September. A key model output node referred to in this report is at 2.5 km downstream Samasoni Weir (Figure 2.1). A separate hydrological model was developed by Entura (2018) to assess the natural flows of the Vaisigano River system without any of the currently existing hydropower or water supply infrastructure. For this model the basic operational river model was used but all water diversions were removed from the model. This model was then run using the same input

rainfall sequence to produce a simulated natural flow output file for each of the required data output locations (Entura, 2018). The flow regimes of the modelled flow series were characterised using a range of standard hydrological statistics that covered the full range of flow components: minimum flows, low flows, baseflows, high flows and flood flows, and also characterised the seasonal distribution of flow components. Also, an integrative flow alteration index score, known as Flow Health (Gippel et al., 2012), was calculated for the flow series. Flow Health calculates deviation of a monthly flow time series from a reference condition, which can be assigned as the natural, or some other defined benchmark, condition. Deviation is measured for each year of the test period with respect to the statistical characteristics of flows over a period of time when flows were not regulated, or over a period of modelled natural flows. Some statistical parameters in Flow Health can be varied by the user, but in this application the default settings described in Gippel et al. (2012) were used. Flow Health indicators are measured over a scale of 0 -1, where 0 = very distant from reference, and 1 = close to reference. There are 9 flow deviation indicators, but one of these (PH) is used as a multiplier on another (LF) to give 8 indicators. These can be divided into two groups: one representing High Flow Health, and one representing Low Flow Health. The combined index is called Flow Health. Low Flow Health (LFH):

• Low flow (LF)

• Persistently higher (PH)

• Lowest monthly (LM)

• Persistently lower (PL)

• Persistently very low (PVL)

High Flow Health (HFH):

• High flow (HF)

• Highest monthly (HM)

• Seasonality flow shift (SFS)

• Flood flow interval (FFI)

Flow Health (FH)

• [(LF * PH) + LM + PL + PVL + HF + HM + SFS + FFI]/8

Unlike most published indicators of flow deviation, which calculate an overall degree of deviation from natural for a period of record, Flow Health calculates a score for every year of record in the test series. In most rivers, the degree and character of hydrological alteration is not uniform through time. In some years flow regulation is more severe than in other years, and in some years flows are naturally quite different to the normal range of flows (e.g. in severe droughts). Flow Health makes no distinction between natural and unnatural flow deviations. This follows from the reality that a deviation of the same magnitude and character will have the same ecological impact whether it was a natural deviation or an unnatural deviation.

When using Flow Health to determine the overall hydrological health of a site over a test period, it is necessary to calculate a statistic that summarises the values for each year of the test period. Suitable statistics might be the 50th, 25th or 10th percentile values. The 10th percentile would be used if the main interest was the years with the largest flow deviation. The scale of deviation from natural is rated by index score classes: 1.0 -0.8 (very small), 0.8 – 0.6 (small), 0.6 – 0.4 (moderate), 0.4 – 0.2 (large) and 0.2 – 0.0 (very large). Because natural hydrological extremes will cause Flow Health scores to fall, the scores in a reference period are not always equal to 1. It would be rare that all 9 indicators scored 1 in a single year. Experience with application of Flow Health in many different rivers from Australia and Asia suggest that overall Flow Health scores are usually higher than 0.8 in a reference series. It is uncommon for scores to fall below 0.6 in a reference series, and extremely rare for scores to fall below 0.4. Thus, for a test series, if the 50th percentile of the annual scores over the test period is greater than 0.8, then hydrology of the site over the test period would be regarded as minimally impaired. In drought periods, Flow Health scores are naturally lower, typically by about 0.2. Hydrologically impaired flow regimes are easily distinguished from flow regimes with natural impairment due to occurrence of drought. Drought will cause the flow indicator values to fall sporadically and temporarily, while flow regulation causes persistently low indicator values.

5.5 2.5 River sediment character, transport and deposition

5.5.1 2.5.1 Introduction Sediment transport is of relevance to this environmental assessment because dams have the effect of interrupting the downstream transfer of sediment. The morphology of a river reflects the interaction of sediment supply (from hillslopes, bed and banks), flow hydraulics (which determines whether sediment will be mobilised, suspended or deposited), resistance by in- channel vegetation (stabilisation of particles by vegetative cover) and hydrological regime (temporal pattern of rainfall and resulting flow, which determines sediment supply and in- channel hydraulic conditions). If a dam reduces sediment supply but maintains a river’s hydraulic and hydrologic capacity to transport sediment, the river could widen and/or incise; conversely, if a dam reduces a river’s hydraulic and hydrologic capacity to transport sediment and check the growth of instream vegetation, the river could aggrade (Gippel, 2001). In both cases, the quality of instream hydraulic habitat would change. Riverine and estuarine ecosystems that rely on the supply of sediment and associated nutrients from upstream for normal functioning could be impacted by construction of a dam upstream. From an operational perspective, over time, trapping of sediment in the reservoir will reduce its water storage capacity and could limit the effective life of the dam. Sediment load comprises bedload, bed material load and suspended load. It is necessary to make a distinction between these components of the sediment transport process because different methodologies are used to estimate bedload, bed material and suspended sediment load. Bedload transport involves movement of particles by rolling, sliding or saltation, while suspended sediment transport involves movement of particles fully suspended within the water column. Sand-size particles can overlap between these process categories; under low flow conditions being transported on the bed, and under high flow conditions, in suspension. Bed material load comprises bed load plus sand-sized material that is transported as bedload. Suspended load is mostly material delivered from hillslopes, bed and banks, while bedload particles are principally sourced from the streambed. In this report the distinction between bed material and suspended material is based on a conventional threshold size of 2 mm, which is the upper size of sand (Gordon et al., 2004, p. 172), i.e. sand was lumped with suspended load.

5.5.2 2.5.2 Suspended sediment load The most reliable method of estimating the suspended load of a river is to undertake field measurements of suspended sediment concentration and flow rate over a wide range of hydrological conditions. This would involve establishment of a long-term sediment and flow monitoring program, which has never been undertaken on the Vaisigano River system. An alternative is to scale, or appropriately adjust, monitoring data obtained from a similar nearby river system. Two sediment-related studies have been undertaken in the American Samoa and Western Samoa islands. Terry et al. (2006) used the radionuclide 137Cs to quantify the recent historical rate of sediment deposition on a lowland alluvial floodplain in the Falefa River basin, Upolu. The Falefa River basin is 58 km2, significantly larger than the 34.4 km2 area of the Vaisigano River at its mouth. The Vaisigano River floodplain was not suitable for investigation because it has been developed for urbanisation and has been largely alienated from the river by flood protection works, i.e. most of the sediment load is transported to Apia Bay. Annual rainfall in the Falefa River catchment ranges from 3000 mm at the coast to 5000 mm in the headwaters, similar to that of the Vaisigano River. The measured rate of vertical accretion over the last 40 years was 4.0 ±0.4 cm per year, which is a rate that exceeds most observations in humid environments. On the basis of this value, and other studies, Terry et al. (2006) concluded that fluvial sedimentation rates on tropical Pacific islands are some of the highest in the world. They attributed this to the rivers having a ‘near-catastrophic’ flood variability index (standard deviation of the base 10 logarithm of the annual maximum flood series exceeds 0.6), reflecting the influence of cyclonic storms. This study established that the rate of sedimentation in Alaoa Dam can be expected to be higher than rates observed in most areas of the world. A suspended sediment monitoring program was undertaken over the period 2012 – 2014 on Fagaˈalu Stream, located southwest of Pago Pago Harbour, Tuitila Island, American Samoa, by the National Oceanic and Atmospheric Administration (NOAA) (Holst Rice et al., 2016). Fagaˈalu Stream drains a mostly forested, steep coastal catchment 1.86 km2 in area, ranging in elevation from 0 to 653 m, and with mean annual rainfall ranging from 3,800 mm on the coast to 6,350 mm on the headwaters. Apart from the much smaller size of the Fagaˈalu Stream catchment, the physiographic character of the forested headwater half of the catchment (about 1 km2) is similar to that of the Vaisigano River. Leta et al. (2017) used suspended sediment data from the Fagaˈalu Stream study to estimate a specific yield of 30 t/km2/yr from the forested headwater part of the catchment. On a global scale this would be considered a relatively low sediment yield for such a small, steep catchment. Suspended sediment concentration, sediment event load, instantaneous discharge and event peak discharge data from the Fagaˈalu Stream study were accurately digitised from plots in Holst Rice et al. (2016, their Figs 9 and 10) (Figure 2.2). Relationships between the variables were then established (all correlation coefficients significant at ɑ < 0.001) (Figure 2.2). These relationships were applied to the 48-year long Vaisigano River modelled hourly flow series (Entura, 2018) to make estimates of annual sediment load. The linear relationship between suspended sediment concentration and discharge established for the Fagaˈalu Stream (Figure 2.2) was based on data from a relatively small catchment, and if applied to the larger Alaoa Dam catchment would predict much higher sediment concentrations than measured at Fagaˈalu Stream (487 mg/L). It was necessary to place a realistic upper limit on predicted sediment concentration for the Vaisigano River in the Alaoa Dam catchment. This was done on the basis of observations reported from other similar rivers. Hyper-concentrations of suspended sediment can be expected in areas draining loess, with values exceeding 1 x 106 mg/L reported (Liu et al.,2013). On other geologies much smaller concentration would be expected. Syvitski et al. (2014) plotted observed sediment concentration data for 35 tropical river basins, ranging from ~30 to ~7000 mg/L. Clark et al. (2017) collected suspended sediment data from two study catchments are in the El Yunque National Forest in northeastern Puerto Rico. The Mameyes River at Puente Roto is 16.6 km long with a catchment area of 17.8 km2 and an elevation range of 83 – 1050 m asl. The Rio

Icacos is 2.0 km long with a catchment area of 3.26 km2 and an elevation range of 620 – 832 m asl. Mean annual rainfall in the headwaters is approximately 4,400 mm. In terms of catchment area, steep topography, forested vegetation cover, volcanic geology and tropical climate, Mameyes River catchment is similar to the Alaoa Dam catchment. The study of Clark et al. (2017) collected weekly water samples, including storm events, since 1991, providing about 4400 values of suspended sediment concentration. When plotted against rainfall intensity, the concentration values for the Mameyes River ranged from <1 mg/L up to about 9,000 mg/L for rainfall intensities 10 – 70 mm/hr. Suspended sediment concentration values in the smaller Rio Icacos reached a maximum of about 3,000 mg/L. On this basis an upper limit of 9,000 mg/L was assumed for sediment concentration for the Vaisigano River in the Alaoa Dam catchment. Application of the Holst Rice et al. (2016) relationships (Figure 2.2) to the Vaisigano River was uncertain for four main reasons: first, the data were not from the catchment of interest; second, the Fagaˈalu Stream catchment was much smaller than the catchment of interest; third, the Fagaˈalu Stream data were collected over a relatively short period; and finally, this ‘rating curve’ approach to prediction of suspended sediment load is known to under-predict high, and over-predict low sediment concentrations, but errors can be reduced to an acceptable level (at least ±20%) by applying the method over an annual time-scale (Horowitz, 2003). Given this uncertainty, some alternative methods of sediment load estimation were applied for comparison. In the absence of local suspended sediment data, sediment load can be estimated using general predictive equations from the literature. These equations, based on global data sets, relate sediment erosion from hillslopes, or sediment load in rivers, to physiographic characteristics of the catchments. Delivery of sediment to rivers is a complex process, and prediction of sediment load in ungauged and unmonitored catchments using general equations is a highly uncertain procedure (Walling, 1983; Walling, 1999; de Vente et al., 2007). The Revised Universal Soil Loss Equation (RUSLE) is an empirical equation that describes the relationship between annual soil loss from hillslopes related to parameters rainfall erosivity, soil erodibility, topography, vegetation cover and conservation practices. This widely used approach is based on a large number of observations on plots under natural and simulated rainfall conditions (Renard et al., 1991).

The RUSLE equation is: 퐴 = 푅 ∙ 퐾 ∙ 퐿푆 ∙ 퐶 ∙ 푃 ( 1 ) where, 퐴 = average annual soil loss (tonnes /hectare) 푅 = rainfall erosivity index 퐾 = soil erodibility factor 퐿푆 = topographic relationship factor (퐿 = slope length and 푆 = slope angle) 퐶 = cropping or coverage factor (e.g. vegetation, gravel or rock) 푃 = conservation or management practice factor Entura (2018) applied the RUSLE over 32 sub-catchments within the catchment of the proposed Alaoa Dam to give an estimate of total soil loss of 61,100 t/yr, or 47,000 m3/yr assuming a sediment specific gravity of 1.3 (appropriate to sediment accumulation in the base of the dam). The load of sediment delivered to rivers is usually less than the mass of soil eroded from hillslopes due to deposition of coarser grains during downslope transport. This process is quantified in a simple way by the sediment delivery ratio. The equation of Renfo (1975) based on catchment relief and length, and the equation of Williams and Berndt (1977) based on river slope, predict a sediment delivery of 1 for the catchment of Alaoa Dam. In a study of a high gradient tropical catchment in the Philippines with an area of 4,123 km2, Atkinson (1995) found that sediment deposition had no significant effects on the overall sediment yield, so a sediment delivery ratio of 1 was appropriate in that case. Given the steep nature of the catchment of Alaoa Dam and absence of major floodplain and other sediment storage areas, a sediment delivery ratio of 1 was assumed. The implication of this is that the sediment load of the river is equivalent to the RUSLE prediction of mean annual hillslope soil loss, i.e. specific yield of 3,715 t/km2/yr. Fournier (1960) was the first to estimate the sediment yields of world rivers and to establish the primary controls on the global pattern of denudation. Fournier (1960) relied on data from only 96 rivers, mostly from North America, with no data available for Africa, South America or Australasia, and limited data from Asia. Despite progress made since that time, the databases of global sediment loads remain limited, with some areas poorly represented (Walling and Webb, 1996). Fournier’s (1960) estimate of total suspended sediment to the oceans remains the highest of 14 estimates listed by Walling and Webb (1996). Asquith et al. (1994) used the empirical equation of Fournier (1960) to estimate the sediment loads transported to the oceans by rivers in the South Pacific, including Samoa. The Fournier equation for high relief, humid climates is:

2 푆푦 = 52.49 푝 ⁄푃 − 513.21 ( 2 ) where, 3 2 푆푦 = sediment yield (m /km /yr) 푝 = maximum mean monthly rainfall (mm) 푃 = mean long-term annual rainfall (mm) Asquith et al. (1994) applied the Fournier equation to the effective surface drainage areas of Upolu and Savaiˈi. The calculation for Upolu was for a drainage area of 708 km2 and based on rainfall data from 8 stations. The volumetric yield was converted to mass yield by applying a specific gravity of 2.1 and 2.6, to give two estimates of specific sediment yield. The two values calculated for Upolu were 6,087 and 7,536 t/km2/yr. These values were later quoted in a desktop study by Kjerfve et al. (2002). On a global scale, these yields would be considered high, but not among the highest, values of specific sediment yield. The maximum reported values of mean annual specific suspended sediment yield from world rivers listed by Walling

and Webb (1996) ranged from 12,736 to 53,500 t/km2/yr. Dedkov and Gusarov (2006) estimated global sediment yield on the basis of data from 4,140 basins across the world. The highest specific yields were from Asian rivers, but second ranked were the islands of the west and southwest parts of the Pacific, with Taiwan, Papua New Guinea and New Zealand achieving average yields of 14,200, 1,200 and 800 t/km2/yr respectively. Syvitski et al. (2014) also ranked tropical rivers highly in terms of suspended sediment yield. Chakrapani (2005) provided a table of the specific sediment yields of 15 major world rivers, including the Ganges, Huang He, Mekong and Amazon rivers, which have a reputation for high sediment loads, ranging from 4 to 1,400 t/km2/yr (excluding the Nile which was assigned zero sediment yield). But these were large basins, and specific sediment yields decline as catchment area increases. Converting the two yield estimates made by Asquith et al. (1994) for Upolu to sediment load from the Alaoa Dam catchment (16.447 km2) gives 100,114 and 123,946 t/yr, which is an order of magnitude higher than that of the RUSLE (see above). In this report, the Fournier equation was applied to the Alaoa Dam catchment using mean monthly rainfall calculated from Alaoa daily data from 1958 to 2016, and also using mean monthly rainfall data from Afiamalu. Alaoa and Afiamalu rainfall data were also used in the Fournier equation to estimate sediment load from the upland middle branch, western branch, and far western branch sub-catchments of the Vaisigano River, while Apia rainfall data from 1923 to 2017 were used to estimate sediment load from downstream sub-catchments. Together, this provided a revised estimate of suspended sediment load from the Vaisigano River to Apia Bay using the Fournier equation. Syvitski and Milliman (2007) proposed the BQART formula for predicting annual sediment load, which has an average uncertainty of 38% compared to the measured sediment loads of 488 global rivers that drain 63% of the global land surface:

0.31 0.5 푄푠 = 600 ∙ (1 − 푇퐸) ∙ 퐸퐻 ∙ 퐼 ∙ 퐿 ∙ 푄 ∙ 퐴 ∙ 푅 ∙ 푇 (푓표푟 푇 ≥ 2℃) ( 3 ) where,

푄푠 = long-term (>30 years) sediment yield (t/yr) 푇퐸 = trap efficiency of lakes, reservoirs and floodplains 퐸퐻 = human-influenced soil erosion factor 퐼 = glacier erosion factor 퐿 = lithology factor 푄 = catchment discharge of water (km3/yr) 퐴 = upstream catchment area (km2) 푅 = upstream catchment relief (km) 푇 = upstream catchment-averaged temperature (℃) In this report, the BQART formula was applied to the Alaoa Dam catchment. Appropriate values for the variables 푇퐸 = 0, 퐸퐻 = 1, 퐼 = 1, 퐿 = 1, 푄 = 0.0349 (for the Alaoa Dam catchment), 퐴 = 16.447 (for the Alaoa Dam catchment), 푅 = 0.4839, and 푇 = 24.2℃. Discharge was mean annual modelled value over 48 years, catchment area was measured from GIS, based on LiDAR-defined area, relief was mean catchment perimeter elevation minus catchment elevation at the outlet, and mean temperature was calculated from hourly weather model simulations over the period 1985 – 2019 for Lanafala (13.9°S 171.75°W 421 m asl), located 2 km upstream of the proposed Alaoa Dam, within the eastern branch catchment. These data were provided by meteoblue (www.meteoblue.com).

5.5.3 2.5.3 Bedload Movement of bedload in a river is sporadic, with the majority transported during high flows when the hydraulic conditions are such that the river has the capacity to mobilise and maintain coarse particles in motion (Ferguson, 2005). In addition, load is a function of discharge and sediment concentration, so high discharges tend to produce high loads. There is a wide body

of literature concerning theory and practice of bedload transport prediction, including model selection for practical applications, with a few examples being Parker (1990), Bennett (1995), Kleinhans and van Rijn (2002) and Barry et al. (2004). This report does not attempt to review the literature on this topic, but draws on conventional and accepted theory and practice of bedload transport prediction. In bed material transport modelling, competence refers to the diameter of largest size of sediment particle (usually expressed in mm) that the flow is capable of moving and is limited by hydraulics; capacity is the maximum amount of sediment of a given size that a stream can transport, expressed as mass per unit of time (e.g. kg/s or t/d); sediment supply refers to the amount and size of sediment available for sediment transport (Hicken, 1995). The capacity transport for a given grain size is only achieved if the supply of sediment of that calibre is not limiting. Thus, there are two potential constraints on bed material transport, hydraulics and sediment supply, so a distinction is often made between supply-limited and capacity-limited transport. Most rivers probably function in a sediment-supply limited condition (Hicken, 1995, p. 71), and this is almost certainly the case for the Vaisigano River. Bedload transport rates can be modelled using capacity limiting sediment transport equations. Ideally, such equations should be calibrated against local measurements of bedload transport. Direct or indirect measurements of bedload are not available for the Vaisigano River, so the estimates made here are uncalibrated and assume unlimited sediment supply. As the river is likely to be supply limited, the estimates are conservative (i.e. the actual bedload is likely to be lower than that estimated here). Bedload transport in a gravel and cobble bed river such as the Vaisigano River is usually conceptualised as a threshold process; there are two approaches for defining the threshold; either mean shear stress is used or, following Bagnold (1977; 1980), unit stream power (e.g. Ferguson, 2005). There is considerable debate in the literature regarding which bed load equation should be applied to particular circumstances, and much variation in the results of different equations. In this investigation the commonly used equation of Meyer-Peter and Müller (1948) based on excess shear stress. This equation was chosen for its suitability to gravel and cobble bed rivers and based on the advice of Gray and Simões (2008) and Fleißner and Dorfmann (2013). Additionally, Yang (1996) recommended that in the absence of measured sediment loads for comparison, the first choice equation was that of Meyer-Peter and Müller’s (1948) when the bed material is coarser than 5 mm. Application of a bedload transport equation required knowledge of river hydraulics and bed material size distribution. These data were available for the hydraulic survey and HEC-RAS modelling reach on the Vaisigano River 3.4 kilometres downstream of Samasoni Weir (see following sections of this report for descriptions of methodologies). The standard equation of Meyer-Peter and Müller (1948) was implemented in Bedload Analyzer (Fleißner and Dorfmann, 2013). This equation assumes a value of 0.047 for dimensionless critical Shields stress, which determines competence. The approach was to analyse 9 of the 29 surveyed and modelled cross-sections, selected to be representative of the reach morphology and hydraulics. The hydraulic conditions of the HEC-RAS model were replicated in the Bedload Analyzer model. At each section, bedload transport rate was modelled over the full range of discharge experienced at the site, to produce rating curves of bedload transport. The bedload transport conditions were variable within the reach, so the cross-section that represented median conditions of the reach was selected to represent bedload transport of the river. The rating curve for the representative cross-section was applied to the hourly modelled discharge time series (for natural, current and with development scenarios) (Entura, 2018), and the resulting hourly bedload transport time series was used to produce a time series of annual bedload transport in tonnes per year.

5.5.4 2.5.4 Hydraulic modelling A one-dimensional hydraulic model was developed for two reaches within the Study Area:

• 245 metre long reach was surveyed in the main branch of the river with the upstream of the survey reach starting 3.4 kilometres downstream Samasoni Weir; and

• 145 metre long reach was surveyed in the middle branch of the river starting immediately upstream and downstream of the offtake weir for the canal diversion to the Alaoa Headpond with the downstream extent finishing immediately below the first of a series of waterfalls between the canal offtake and where this branch enters the western branch.

HEC-RAS 5.0.3 was used to model the one-dimensional hydraulics of the reaches in accordance with the user’s manual (U.S. Army Corps of Engineers, 2016). The models were based on field transect surveys. Input variables relating to reach lengths and bank station heights were calculated using ArcGIS software and the transect survey data. Manning’s n roughness coefficients were initially calculated using the photo catalogue and known roughness relationships with various surface types including sediment size and vegetation. A mixed flow regime steady flow model was used to run the hydraulic models due to the mix of super-critical and sub-critical flows present within the representative reaches. Model calibration was achieved using the observed water levels and a combination of transect interpolations and manipulating Manning’s n values within natural and reasonable bounds, until the observed water level was in close agreement with the measured flows during the field survey. Further details of the hydraulic modelling can be found in Entura (2019). In this report, only the results of the hydraulic modelling of the reach 3.4 kilometres downstream Samasoni Weir were used. These results were used to support estimation of bedload transport using the Meyer-Peter and Müller’s (1948) equation. 5.5.5 2.5.5 Bed material particle size distribution measurement The bed material in the streams within the Study Area was sampled at 7 locations (Figure 2.3) using the Wolman Pebble Count method (Wolman, 1954; Gordon et al., 2004, p. 105), measuring the size of at least 100 stones randomly sampled from the stream bed on riffle crests. The particle size data were converted to distributions of percent finer by weight (in phi classes), corrected for bias in sampling using the method of Leopold (1970). The smallest size material that could be measured in this way was 2 mm (upper end of sand size class). Where continuous areas of sand or bedrock were encountered in the sampling, its presence was recorded, allowing an estimate of the proportion of the total bed area covered by sand or bedrock to be made. Degree of sorting was characterised by the Folk and Ward (1957) sorting coefficient. Other standard indices were also used to characterise the particle size distributions.

Figure 2 . 3 : Location of bed material sampling sites.

5.5.6 2.5.6 Reservoir sediment trap efficiency A commonly applied method for estimating the trap efficiency (푇퐸) of reservoirs is that of Brune

(1953), who produced three curves of 푇퐸 as a function of reservoir capacity/annual inflow ratio (푉⁄푄) based on data from 44 reservoirs. The upper and lower curves represented the envelope of the data for normally ponded reservoirs, and the middle curve represented the median for normally ponded reservoirs. Normally ponded reservoirs were distinguished from those that are periodically drawn down to expose sediments, for desilting or flood control, for example. At a later time, the USDA-SCS (1983) interpreted the three curves to apply to sand and gravel (upper curve), mixture of sediment textures (middle curve) and clay-silt material (lower curve), but this appears to be arbitrary (Verstraeten and Poesen, 2000). Equations for

휏 푇퐸 = 100 ( ) 퐵푟 0 . 012 +1 . 2 휏

푇퐸 = trap efficiency (0– 100%)

휏 = 푉 ⁄ 푄

푉 = reservoir volume at capacity (m3)

푄 = mean annual inflow (m3/yr)

The majority of the suspended sediment in the Vaisigano River is likely to be fine grained, so the lower Brune curve is most applicable. The formula for the lower curve,1 >for 푉 ⁄푄 > 0. 02 ( which applies to the Alaoa Dam catchment), is:

1 . 92 푇퐸퐵푟 = 94 − (3 . 38 |푙푛 (푉 ⁄푄 ) | )

An alternative method of estimating trap efficiency, known as the Sediment Index Method, was proposed by Churchill (1948), and wasdescribed in U.S. Army Corps of Engineers (1989). The method was developed using Tennessee Valley Authority Reservoir data. Lewis et al. (2013) provided the formula for the Churchill (1948) curve, converted to metric units:

−0.2 9 . 61 × 10 6 휏 푇퐸 = 112 − 800 ( ) 퐶 ℎ 푢

3 . 17 × 10 8 푄 푢 = ( ) 퐴 the three curves were provided by Verstraeten and Poesen (2000). The basic Brune curve formula for the median curve was given by Heinemann (1981) as:

( 4 ) where,

( 5 )

( 6 ) where,

퐴 = cross-sectional area of reservoir (m2) The value of 퐴 is usually calculated by the volume of the reservoir (푉) divided by the length of the reservoir water surface from the point of maximum upstream extent to the dam wall (퐿). Lewis et al. (2013) also modified the Brune and Churchill equations so that daily trap efficiencies could be calculated to account for shorter reservoir residence times expected in tropical rivers with relatively high intra-annual flow variability, as would apply to the Vaisigano River system. Lewis et al. (2013) pointed out that the Brune curve is not suitable for calculating trap efficiency at the event time-step (Verstraeten and Poesen, 2000), while the Churchill equation was developed using quarterly data and was applicable over periods as short as 5 days. Lewis et al. (2013) proposed that the modified daily Churchill method was suitable for broad application. For the Churchill equation modified for estimation of daily trap efficiency, the daily residence time is calculated as:

= 휏∗ 푄푉𝑖⁄365 ( 7 ) where, 3 푄𝑖 = the reservoir inflow on day 𝑖 (m )

Lewis et al. (2013) then calculated daily 푇퐸퐶ℎ,𝑖 by substituting Eq’n 7 for 휏 in Eq’n 6. They argued that the daily 푇퐸퐶ℎ,𝑖 should be weighted based on daily inflow volumes, as the majority of river sediment is transported during higher inflow periods. The equation for daily calculation of the Churchill method is then:

∑푛 = 푇퐸푛 퐶 ℎ𝑖,𝑖푄𝑖 푇퐸퐶ℎ∗ = 𝑖 1∑ 𝑖 = 1 푄 ( 8 )

Lewis et al. (2013) applied the daily 푇퐸퐶ℎ,𝑖 method over the water year, where 푛 is the number of days in the water year. The Alaoa Dam is operated as a flood control dam, with potentially highly variable water levels, and thus volume (푉), during times of flood inflows, when the majority of sediment is transported into the reservoir. For application to this reservoir, the Lewis et al. (2013) equation for daily residence time was refined to:

= 휏∗ 푄푉𝑖𝑖⁄365 ( 9 )

where, 3 푉𝑖 = the reservoir capacity volume on day 𝑖 (m )

On days of very high inflow, the formula can return negative values of 푇퐸퐶ℎ, which in this report were assigned zero trap efficiency. As reservoir volume varies from day to day, so too does the reservoir length. For application to the Alaoa Dam, we made a refinement to the Lewis et al. (2013) equation by introducing daily variable cross-sectional area of the reservoir. An empirical relationship was developed from bathymetric data to describe reservoir length as a function of reservoir water surface elevation:

퐿𝑖 = 10.27 퐻𝑖 − 698.2 ( 10 )

where, 퐿𝑖 = the reservoir length on day 𝑖 (m)

퐻𝑖 = the reservoir water surface elevation on day 𝑖 (m asl) In this report, two methods of estimating the trap efficiency of the Alaoa Dam reservoir were applied. The lower Brune curve was applied to estimate the long-term trap efficiency, and the daily Churchill formula was applied, as descried by Lewis et al. (2013), but with the refinements described above, using the 48-year modelled daily flow series, and daily calculated reservoir length, volume and surface area based on bathymetric relationships determined from LiDAR data using GIS tools. The annual water year trap efficiencies, beginning in September, were then applied to the modelled annual sediment load of the river to estimate the annual time series of sediment deposited in the reservoir. It would be expected that the daily calculation method of 푇퐸 would produce a more accurate estimate of the long term mass of sediment trapped in the reservoir than applying a single Brune estimate of 푇퐸 to estimated mean annual sediment load.

5.5.7 2.5.7 Reservoir shoreline erosion Shoreline erosion will occur in the reservoir of the Alaoa Dam. Sediments from eroded soils could increase turbidity and potentially reduce water storage capability. Nutrients from eroded soils could increase risk of algal blooms. The dam will be operated to be usually within a narrow range between the NMOL (Normal Minimum Operating Level) ay 152.3 m asl and NFSL (Normal Full Supply Level) at 152.4 m asl (Entura, 2018), with the water level rising under storm event inflow conditions, very occasionally reaching the spillway crest level at 174.6 m asl. This mode of operation will concentrate erosion around the rim of the reservoir at the NFSL due to focused attack by wind generated waves. It is also possible that the upper levels of the reservoir could be subject to slumping of saturated banks during the period when the reservoir draws down following flood events. Operation of the dam would result in existing vegetation within the reservoir under the NFSL to die. This will leave the face of the shoreline exposed to wave attack. If left unprotected, with time, as the shoreline retreated from erosion, it would leave a wave-cut bench around the perimeter. The rate of erosion would decline to a very low rate when the wedge of easily eroded surface soil was removed. The height of the waves generated by winds depends primarily on three factors:

• wind speed,

• wind duration, and

• fetch over which the wind blows

Prediction of the impact of wind-generated waves on shoreline erosion rate is a complex problem. Ekebom et al (2002) derived a method for measuring fetch length, fetch direction and wave exposure in coastal areas by applying GIS, averaged wind data, wave height forecasting curves and a linear wave model. The second part of the problem, which is predicting the geomorphological response of the shoreline to the prevailing wave climate, is problematic, requiring a number of simplifying assumptions. Elçi & Work (2003) developed a method for predicting wave induced shoreline erosion on a reservoir by relating erosion rates to wind wave forces (via shear stress) and assuming a simplified representation of the shape of the beach profile. Elçi & Work (2003) and Elçi et al (2007) quantified shoreline erosion rates on Harwell Lake, South Carolina and Georgia, based on measured fetch, parameterized beach profile shape, and measured wind vectors. The methodology for the prediction of shoreline erosion was calibrated and validated using digital aerial photos of the reservoir taken in different years, and indicated approximately one metre per year of shoreline retreat for several locations. It would be difficult to apply this method to the proposed Alaoa Dam due to lack of calibration data. Lorang and Stanford (1983) observed that since regulation of Flathead Lake, Montana, wave erosion occurred on the entire lake shoreline, but erosion was greatest along the north shore because of the prevailing winds and more easily erodible soils. This lake was around 10 km wide and 35 km long, so it was much larger than the proposed reservoir associated with the proposed Alaoa Dam. Lorang and Stanford (1983) identified three main modes of shoreline erosion: undercutting, endstripping, and over-wash (Figure 2.4). Each resulted in different rates of shoreline retreat and produced spatial variations in measured shoreline change. The steep valley slopes within the area to be impounded by the Alaoa Dam means that undercutting would be the dominant mode of shoreline erosion. Lorang and Stanford (1983) found that the process of undercutting was relatively constant in response to wave energy (high and low wave energy), but sporadic slumping occurred when the vegetation could no longer support the bank. On occasions, high winds associated with storm events blew trees down because erosion beneath the root wads had weakened anchorage (Lorang and Stanford 1983). On Lake Hawea in New Zealand, Kirk et al (2000) reported extensive cliff erosion during an extreme event when a period of very strong winds coincided with high lake levels. On Lake Thunderbird, Oklahoma, Allen (2001) found that shoreline geometry and bathymetry played a major role in determining the degree of erosion at a particular shoreline site. Sites with straight shorelines or headlands that were exposed to long wind fetches from prevailing wind directions were particularly vulnerable to more frequent and higher waves. Conversely, sites that were within coves or that were behind peninsulas or islands that blocked the wind were more protected from waves. In these areas vegetation was often present and erosion was less severe or even minimal. Lake bathymetry also influenced wave action. The shallower was the nearshore, and the wider was an underwater bench, the higher was the drag or resistance to waves. Waves were subsequently smaller in such areas, in contrast to areas where the water deepened abruptly and there was less resistance or bottom roughness to influence the waves.

Figure 2 . 4 : Three processes of geomorphic wave -shoreline interaction: a) undercutting; b) endstripping; c)over - wash. Source: Lorang and Stanford (1983).

The potential erosion of the proposed shoreline of the proposedAlaoa Dam reservoir was examined

152.4 m asl . It was also assumed thatthe effects of waterlogging and attack from wind generated waves would cause vegetation on the shoreline to die, exposing the bank to erosion. Erosion would by assuming that for the majority of the time, the dam would be kept between 152.3 and strip away the soil to the underlying weathered parent material. The eroded soil would deposit within the storage. The volume of material eroded from the shoreline would be a function of soil depth and design wave height. The majority of the Vaisigano River catchment is Salani Volcanics (S) (Late Pleistocene, around 1 Ma). These volcanics have developed a soil thickness of about 300 mm (Kear and Wood, 1959; Keating, 1992). Entura (2018) evaluated wind speed data from Nafanua (Apia) from 2011 to 2017, with the maximum speed observed being 55 km/h. These data were comparable with hourly weather model simulation data over the period 1985 – 2019 for Lanafala (13.9°S 171.75°W 421 m asl)

from 1985 – 2019, provided by meteoblue (www.meteoblue.com). These data indicated a maximum wind speed of 97 km/h and wind speed exceeded 0.01 percent of the time of 52.5 km/h. In the absence of longterm wind data, for the probable maximum event, Entura (2018) adopted the Samoa 500-year return period cyclonic wind speed standard of 66 m/s (238 km/h) for Apia (Siqueira et al., p. 11). For the 100-year return period event Entura (2018) adopted a value of 44 m/s (159 km/h). These could be underestimates, as the 500-year return period cyclonic wind speed estimated for Apia by Siqueira et al. (2014, p. 11), using the Tropical Cyclone Risk Model, was 79.9 m/s, and 66 m/s was between the 25-year and 50-year return period modelled values. Entura (2018) calculated the effective fetch to be 334.8 m and the maximum fetch to be 817 m. They used these data to predict the 1 in 100 year recurrence interval design wave height near the dam wall to be 0.7 m and the probable maximum event design wave height was 0.9 m. Over most of the reservoir perimeter, the fetch would be lower than that estimated by Entura (2018) for the dam wall. Also, wind speed is significantly reduced under forest compared to open land due to shielding and surface roughness (e.g. Silversides, 1978; Tahir and Yousif, 2013; Moon et al., 2013). This forest effect from the surrounding land cover, combined with the effect of shielding from ridge topography surrounding the reservoir, would be expected to temper maximum wind speeds on the reservoir surface compared with those experienced at Apia. Considering these factors, for modelling shoreline erosion, this report adopted a value of 0.7 m for maximum wave height. The shoreline erosion model assumed that all the soil (assumed depth 300 mm) would be stripped from the area around the reservoir shoreline from 300 mm below the NMOL (Normal Minimum Operating Level) at 152.3 m asl (due to wave turbulence) to 300 mm above the NFSL (Normal Full Supply Level) at 152.4 m asl plus the design wave height of 0.7 m. The 300 mm allowance above the design wave height was to allow for collapse of undercut soil. Thus, the total height of the scour zone would be 1.4 m, over the elevation range 152.0 – 153.4 m asl. The volume of soil that would be eroded over this zone was calculated using GIS as the surface area multiplied by the soil depth of 300 mm.

5.5.8 2.5.8 Stability of hillslopes within the reservoir During storm events, the reservoir will function as a buffer to reduce downstream flood impacts. Inflows to the reservoir will result in the water levels rising, very occasionally up to and over the spillway. Two processes could impact the stability of the hillslopes within the reservoir:

1. Potential for mortality of the terrestrial vegetation within the reservoir area above NFSL due to waterlogging caused by submergence during periods of raised water levels, with subsequent higher risk of erosion due to rainfall events on exposed hillslopes with weakened vegetative cover.

2. Drawdown of reservoir water levels during flood recession at a rate that exceeds the rate at which the saturated soils can drain, creating a risk of slumping, especially given the steep slopes within the reservoir.

The first risk above was investigated by characterising the duration of periods of submergence at a range of elevations from the 48-year modelled hourly reservoir level time series (Entura, 2018). Investigation of the capacity of the vegetation within the Alaoa Dam reservoir to withstand submergence was not within the scope of this report, but it is noted that losses of biomass within reservoirs as a result of poor seed germination, arrested plant growth, and accelerated mortality of trees is a recognised problem (Kozlowski, 2002). The relative capacity of plant species within flood control reservoirs to survive prolonged inundation has received attention within the literature, particularly for the case of the reservoir of the Three Gorges Dam, China (Liao et al., 2010; Yang et al., 2012; Wang et al., 2014; Wang et al., 2016).

The second risk above was investigated by characterising the drawdown rate from the 48- year modelled hourly reservoir level time series (Entura, 2018). The draw down rates were compared with the expected hydraulic conductivity of the soils. In the area of the proposed Alaoa Dam reservoir, soils are very steep Latosolic soils from basalt, andesite and basic volcanic ash or calcareous basic tuff, and clay hill soils (New Zealand Soil Bureau, 1956). Latosols are found in tropical forest areas and are typically classified as Oxisols (USDA soil taxonomy) or Ferralsols (World Reference Base for Soil Resources). Weathering of volcanic materials such as ash, tuff, and pumice typically forms Andisols. A review of literature citing hydraulic conductivity of Andisols (Table 2.2) revealed that this soil type commonly has high hydraulic conductivities, up to and exceeding 100 mm/hr, although values lower than 10 mm/hr have been recorded. The literature suggests that Latosols (Oxisols), which are likely the major soil type in the reservoir area, tend to have low hydraulic conductivities (Table 2.3), although values exceeding 100 mm/hr have been recorded for some soils from Brazil. Eswaran and Reich (2005) described the texture of Oxisols as sometimes loamy or even coarser, with many extremely clayey, but with the clay being aggregated in a strong grade of fine and very fine granular structure. They noted that this strong granular structure causes most Oxisols to have a much more rapid permeability than would be predicted by the particle-size distribution class. Therefore, free drainage can take place soon after rain without puddling (Eswaran and Reich, 2005). Similarly, College of Tropical Agriculture and Human Resources (2019) described highly weathered tropical Oxisols of Maui, Hawaiˈi, as freely draining and resistant to compaction. On the basis of this literature review, there is a high level of uncertainty regarding the rate at which soils within the reservoir would drain during drawdown.

Table 2.2: Hydraulic conductivity of volcanic Andisol soils reported in the literature.

Table 2.3: Hydraulic conductivity of volcanic Oxisol (Latosol) soils reported in the literature.

6. 3. Description of the existing physical environment 6.1 3.1 Morphology of Vaisigano River catchment and Apia Bay The island of Upolu, Samoa is mountainous, with an elevation range from sea level up to 1158 m (Figure 3.1). The Vaisigano River headwaters are on the central ridge of the island, and the river flows northwards, approximately in the middle of the island (Figure 3.2). It is the largest

river on Upolu, with a catchment area of 34.4 km2 and mainstream length of approximately 14.1 km. The lower Vaisigano River flows through central Apia and it is of key importance in providing both water supply and power generation for the urban area. The Vaisigano River catchment is roughly wedge shaped with the upper catchment approximately 7 km wide (Figure 3.2). The Mulivai and Vaivase catchments are situated to the west and east respectively. The river rises on the main divide of Upolu at 1158 m asl (metres above sea level) at Mt. Fito and comprises four main tributaries, the western, middle and middle-eastern and eastern branches which merge to a single channel at Alaoa, about 5 km above its estuary in Apia Bay (Figure 1.1). These tributary names were used in this report for convenience. Local stream names were indicated on a map in Freeman (1944). The eastern branch was known as Soaga Stream, which had upper tributaries named Maualuga Stream and Vai o Le Fe'e. Soaga Stream is also named as such on the geological map of Upolu produced by New Zealand Geological Survey (1958). The middleeastern branch was known as Puale'ile'i River. Soaga Stream and Puale'ile'i River joined to form the Vaisigano River. The western branch was named Puao River on the map of Freeman (1944), but was named Vaisigano River on the geological map of the New Zealand Geological Survey (1958). The catchment of the Vaisigano River is deeply dissected (Figure 3.2). The drainage lines emerge from amphitheatere-headed canyons, with deep poorly-graded valleys (Kear and Wood, 1959). The relief of the valleys in the vicinity of the proposed Alaoa Dam in the order of 200 m (Figure 3.3). The lower 0.5 km of the Vaisigano River is tidal. The natural lowland floodplain of the Vaisigano River is now mostly alienated from the river due to the urban development of Apia, with associated flood protection works. The Vaisigano catchment has a high proportion of very steep hillslopes (Figure 3.4), which in combination with the steep stream gradients renders it conducive to rapidly-rising floods following heavy rain, as well as potentially high sedimentation loads. The upper 2.0 km portion of the Vaisigano catchment descends rapidly at a slope of about 33%, which then changes to 7% for the next 5.6 km (Entura, 2018) (Figure 3.4). The valleys in which the proposed Alaoa Dam reservoir would be set are very steeply sloping (Figure 3.5). The slopes are in the range that would pose a risk of slope instability. Rainforest is the predominant vegetation in the upper part of the catchment. Sixty eight percent of the Vaisigano catchment is forestry. The lower flanks of the catchment are a mixture of plantations, scrub, grazing lands and settlement. The river basin is bounded to the south by the central chain of volcanic cones that form the Upolu ridge.

Figure 3 . 1 : Topography of Upolu,Samoa, showing location of Vaisigano River catchment.

Figure 3 . 2 : Topography of Vaisigano River catchment.

Figure 3 . 3 : Topography of the areain the vicinity of the proposed Alaoa Reservoir.

Figure 3 . 4 : Slope of Vaisigano River catchment.

Figure 3 . 5 : Slope of the area in the vicinity of the proposed Alaoa Reservoir.

Solomon (1994) used air photo interpretation, literature review and ground survey to develop an understanding of the geomorphic processes influencing the coastal and nearshore sedimentary systems of Apia Bay. The focus of the study was the Apia Harbour area and Mulinu'u Peninsula immediately to the west. Solomon (1994) reported that small scale mining of beaches has been practised in Samoa for a considerable time. Large scale dredging adjacent to Mulinu'u Point commenced sometime after 1970. Solomon (1994) also reported that the harbour had been dredged in order to perform land reclamation and to maintain navigable depths, and sediment deposited at the mouth of the Vaisigano River had been removed regularly by heavy equipment. Major dredging and construction began in Apia Harbour in 1964 (Gauss, 1981, reported by Solomon, 1994) when the main wharf was constructed and the land area to the west of the Inner Harbour was reclaimed. Historical aerial photography indicates the extent of enclosure of the harbour (Figure 3.6). Apia Bay, although

open to the sea, was partially protected by fringing reefs, but the reclamation, mostly completed by 1970, further enclosed the harbour. Within Apia Harbour, sediments range from fine to medium sand to silt and mud (Gauss, 1981, reported by Solomon, 1994), while sand predominates in the outer harbour (Solomon, 1994). Richmond (1992) reported the presence of a shallow sub/intertidal delta at the mouth of the Vaisigano River. The river appeared to deposit its coarse bedload (sand/gravel and cobble) within 150 m of the mouth, beyond which silt and clay was deposited. Waves generate a net longshore current which moves a mixture of carbonate and volcanic sand east to west. The source of this sand is the reef flat (Solomon, 1994).

6.2 3.2 Geology and soils The geology of Samoa was described by Kear and Wood (1959), Kear (1967) and Keating (1992). A geological map of Upolu produced by New Zealand Geological Survey (1958) in association with the paper of Kear and Wood (1959) (Figure 3.7). The majority of the Vaisigano River catchment is Salani Volcanics (S) (Late Pleistocene, around 1 Ma). The lithology of Salani Volcanics is critic basalt and olivine chlorite and basalts that typically grade upward from porphyritic basalt through vesicular basalt to rubbly a’a (Richmond, 1992; Keating, 1992). The Salani Volcanics overly the older (1.5 – 2.8 Ma) Fagaloa Volcanics (F). Most of the drainage system flows through Salani Volcanics, but the geological map indicates the presence of a strip of Pu'apu'a Volcanics (P) on the Eastern Branch river channel from

Most of the coastal area around Apia area is underlain by material of Holocene age. The lowland floodplain of the Vaisigano River is Alluvial swamp, and the Holocene unit closer to the coast is known as Tafagamanu SandFigure ( 3 . 7 ) , described by Kear and Wood (1959) and Richmond (1992) as a raised beach deposit of coral sand with some coral and basalt gravel.hmond Ric (1992) classified the coastline east of Apia as Type III, described as fringing reefs and narrow coastal strip or beaches, barrier spits, coastal swamps associated with streams. West of Apia the coastline was Type I, described as wide fringing reefransitional t to shallow barrier reef.

downstream of the dam site to 'O le Fale o le Fe'e. This later young volcanic formation (around 300 BP) consists of lava flows and very little weathering. Fukuyama Shoji Company Limited (2013) reported that this material was very hard to drill.

Figure 3 . 7 : Extract of thegeological map of Upolu produced by New Zealand Geological Survey (1958) in association with the paper of Kear and Wood (1959).

The soils of Upolu were mapped and described by the Provisional Soil Map of Upolu, Western Samoa (New Zealand Soil Bureau, 1956) (Figure 3.8, Figure 3.9 and Figure 3.10). The majority of the headwaters and mid-catchment areas of the Vaisigano River catchment are in steep Latosolic (tropical) soils from basalt, andesite and basic volcanic ash (Figure 3.8 and Figure 3.10). In the area of the proposed Alaoa Dam reservoir, soils are very steep Latosolic soils from basalt, andesite and basic volcanic ash or calcareous basic tuff, and clay hill soils (Figure 3.9 and Figure 3.10). The strip of Pu'apu'a Volcanics (P) on the Eastern Branch river channel is associated with Recent soils from basaltic alluvium, clay. Additional description of the soils of Upolu was provided by Keating (1992). Early studies by Hamilton and Grange (1938) and Seelye et al. (1938) showed that soils were relatively shallow, heterogeneous, with frequent stones and boulders. On Salani Volcanics, the surfaces of lava flows are often deeply weathered, with soil often more than 100 mm thick. On the distal

flows (i.e. distant from the source crater), over 300 mm of soil has formed (Keating, 1992, p. 143). Kersch et al. (2013) undertook a geological site investigation along the bed and slopes of the Vaisigano River in the area proposed for the Alaoa Dam. The survey began at the intake to the eastern canal and extended to the dam site. It was found that much of the valley slopes in the catchment were unstable, evidenced by landslips. The tuffs and other loose volcanic material had weathered to loamy materials that were easily eroded from hillslopes, and undercutting of weaker material underlying resistant basalts resulted in rock fall.

Figure 3.8: Digitised version of part of the Provisional Soil Map of Upolu, Western Samoa (New Zealand Soil Bureau, 1956) covering the Vaisigano River catchment.

Figure 3 . 9 : Digitised version of part of theProvisional Soil Map of Upolu, Western Samoa (New Zealand Soil Bureau, 1956) covering the area in the vicinity of the proposed Alaoa Reservoir.

Figure 3.10: Soil type key of the Provisional Soil Map of Upolu, Western Samoa (New Zealand Soil Bureau, 1956).

6.3 3.3 Climate Samoa has an equatorial/monsoonal climate with all months above 18°C, and an average annual temperature of 26.5°C, classified as Af on the Koeppen-Geiger classification. The wet season is from

October/November to April/May. Typhoons can sometimes occur over the period December to March/April. Trade winds temper the climate in the dry season between April/May and October/November. Samoa’s mountains have a significant effect on rainfall distribution. Wetter areas are located in the south-east and relatively sheltered, drier areas in the north- west. Samoa’s climate varies considerably from year to year due to the El Niño- Southern Oscillation (Elvey and Gippel, 2019). The available rainfall data for Upolu was reviewed and summarised by Entura (2018). Relatively long records of daily observed rainfall were available for Apia and Alaoa stations, while Tiavi and Mt Le Pue stations had short patchy records. For Afiamalu station a mean monthly rainfall distribution was available. In addition, a simulated hourly rainfall time series was available (www.meteoblue.com) for Lanafala. These stations were located across arrange of locations (Figure 2.1) and elevations (Table 2.1) within or nearby to the Vaisigano River catchment. Rainfall increased markedly with elevation (Figure 3.11), with mean annual totals for Apia (2 m), Alaoa (260 m) and Afiamalu (798 m) stations being 2,993 mm, 4,020 mm and 4,849 mm respectively. The simulated rainfall from Lanafala were lower than all other stations (Figure 3.12), even though located mid-catchment, suggesting the simulation systematically underestimated rainfall. From 2007 onwards, Alaoa rainfall data indicated a positive step change that was not present in Lanafala or Apia data (Figure 3.12). The small amount of data from Tiavi and Mt Le Pue suggested that they had similar rainfall, and high rainfall compared to other lower-elevation stations (Figure 3.12).

Figure 3 . 12 : Annual rainfall distributions for five rainfall stations within or nearbyto the Vaisigano River catchment. Tiavi and Mt Le Pue have each only two complete years of data. Lanafala data (from 1985) were simulated, not observed.

Statistics describing the simulated wind data from LanafalaFigure ( 3 . 13 ) were comparable to those of Holden (1991). For example, in the simulated data, winds from the east, southeastortheast and n comprised 78.5% of the total winds, andhe t average wind speeds from these three directionswas 24.2 km/h . Winds at Lanafala of less than 5 km/h prevailed 3% of the time while winds between 5 and 23 km/h prevailed50 % of the time. Winds above 49m/h k occurred less than 0.2% of the time.

Figure 3 . 13 : Distribution of wind speed by month ( top) and by direction (bottom) for Lanafala(13.9 °S 171.75 °W 421 m asl) . Based on hourly weather modelsimulations 1985 – 2019 , data provided by meteoblue ( www.meteoblue.com ) .

6.4 3.4 Hydrology The modelled hourly discharge series from the node at 2.5 km downstream Samasoni Weir (Figure 2.1) was selected to represent the existing hydrological regime of the lower Vaisigano River, and to characterise the degree of flow alteration that has occurred relative to the natural condition, with no hydropower offtakes. This location is between Samasoni Weir and Samasoni Power Station, so the river is impacted by diversion of water at the weir to a pipe that returns the water to the river downstream at the power station. Samasoni hydropower scheme was commissioned in 1982 and has a design discharge of 1.29 m3/s (JICA, 2003, p. 4-3). Other locations on the river system would be impacted by the current operation of hydropower plants to varying degrees, depending on the proportion of water diverted from

upstream. Analysis of hydrological time series was undertaken for other locations for the environmental flow assessment (Elvey and Gippel, 2019). The mean daily flow calculated for each day of the year demonstrated that although the current regime retained the natural seasonal distribution of flows, the mean flows for each day of the year were lower by about 1 – 2 m3/s, which represents a reduction of about 80 percent in the low flow period, and about 30 in the high flow period (Figure 3.14). The flow duration curve illustrates the same impact, suggesting that the impact diminishes only at high flows (Figure 3.15). Low flow indices (minimum hourly discharge, and flow exceeded for 95% of the time) and baseflow index (flow exceeded 50 percent of the time) for each month indicate that under the current regime, flow is precariously low relative to the natural regime in the months July to December, and low flows are dramatically reduced from natural in the high flow season (Figure 3.16). High flow indices (flow exceeded for 1% and 0.1% of the time) for each month indicate that moderate flood flows are not greatly altered under the current regime relative to the natural regime, although small flow events (freshes) have been noticeably reduced in every month (Figure 3.17). Flood frequency analysis suggests that for any recurrence interval, peak flood magnitude has only been slightly reduced under the current regime relative to the natural regime (Figure 3.18). The annual time series of annual flow indicates that the total flow each year under the current regime is on average 54 percent lower than under the natural regime, with the maximum reduction in any year being 65 percent (Figure 3.19). The annual peak hourly discharge under the current regime was on average 5.2% lower relative to the natural regime, with the maximum reduction in any year being 15.7 percent (Figure 3.19). The Flow Health indicators were calculated using default statistical parameters, with the exception that the high flow season was manually specified as extending from November to May inclusive, consistent with the definition of the wet season by Elvey and Gippel (2019). As would be expected from the results of the other statistical comparisons made above, the integrative index classified the Vaisigano River 2.5 km downstream Samasoni Weir highly modified from natural, with scores for 5 of the nine indicators persistently in the very large deviation class (0.0 – 0.2) (Figure 3.20). Less impacted aspects of the flow regime were Persistently Higher (an indicator of unseasonal regulated irrigation flows), Seasonality Flow Shift, and Persistently Very Low (an indicator of unnatural cease to flow). Flood Interval indicator suggested a long periodic impact covering over half of the series (Figure 3.20). The median annual Flow Health score over the entire modelled series was 0.29 (Figure 3.21), which is within the large deviation class. Overall, the hydrological analysis, including application of Flow Health, suggested that the Vaisigano River at 2.5 km downstream of Samasoni Weir is highly impaired relative to natural, and the impairment is persistent from year to year. The data strongly suggest that the river would also have experienced impaired ecological health due to the existing degree of flow regulation. Some reaches of the river system would not be impacted to the same degree, but others might be impacted to a greater degree.

Figure 3 . 14 : Mean daily flow calculated for each day of the year at 2.5 km downstream ofsoni Sama Weir for the modelled natural and current scenarios. Calculated from-year 48 long hourly modelled

time series.

Figure 3.15: Flow duration curve at 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios. Calculated from 48-year long hourly modelled time series.

Figure 3 . 17 : High flow indices (flow exceeded for 1% and 0.1% of the time) for each month, at 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios. Calculated from- 48 year long hourly modelled time series.

Figure 3 . 18 : Partial duration series flood frequency for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios, based on 4th order polynomial curves fitted to frequency data from 45 years of modelled peak daily discharge (2013 data – 2057) .

Figure 3 . 19 : Annual flow and annual peak hourly discharge for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled natural and current scenarios. Calculated from- 48 year long hourly modelled time series.

Figure 3.20: Flow Health individual and combined flow deviation indicator scores for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled current scenario relative to the natural scenario.

Figure 3 . 21 : Flow Health combined annual flow deviation indicator score for Vaisigano the River 2.5 km downstream of Samasoni Weir for the modelled current series relative to the natural series.

3.5 River sediment character, transport and deposition

3.5.1 Suspended sediment load

The existing suspended sediment load of the proposed Alaoa catchmentDam was estimated using the three different empirical relationships of Holst Rice et al. (2016), one based on suspended sediment concentration as a function of instantaneous discharge, one on event suspended sediment load as a function of event peakdischarge, and the other on event suspended sediment load as a function of event total discharge.

The 48-year modelled hourly time series of dam inflows was converted to suspended sediment concentration using the relationship of Holst Rice et al. (2016),capping maximum suspended sediment concentration at 9,000 mg/L (required on only 10 instances over- years) 48 and assigning zero concentration to discharges less than m 0.13 /s (0.54% of the time over 48-years). The hourly sediment load was then calculateds adischarge multiplied by concentration. The estimates of sediment load were then summed to annual loads for water years, with a mean value of 2 8,867 tonne/yr (range 1,636– 95,120 t/yr), or a specific yield of 539 t/km /yr (Figure 3 . 22 ).

The 48-year modelled hourly time series of dam inflows was analysed using spells analysis to isolate flow events with discharge exceeding 1 m3/s, and event independence defined by an interval of 7days between peaks. This produced 271 independent events. The total discharge of each event was calculated and then converted to sediment load using the relationship of Holst Rice et al. (2016). The estimates of sediment load were then summed to annual loads for water years, with a mean value of 1,269 tonne/ yr (range 463 – 2,160 t/yr), or a specific yield of 77.2 t/km2/yr (Figure 3.22). The estimate based on event peaks was made comparable with that based on total event discharge by extracting the largest 271 independent event peaks from the 48-year modelled hourly time series of dam inflows. The peak discharge of each event was then converted to event sediment load using the relationship of Holst Rice et al. (2016). The estimates of

sediment load were then summed to annual loads for water years, with a mean value of 108 tonne/yr (range 3.4 – 1,863 t/yr), or a specific yield of 6.6 t/km2/yr (Figure 3.22). Of the three empirical Holst Rice et al. (2016) relationships used to calculate annual sediment load, only the one based on instantaneous sediment concentration produced an expected result. The other two equations appeared to underestimate annual sediment load, likely because these two empirical equations were not transferrable to the larger catchment of the Alaoa Dam.

Figure 3 . 22 : Annual load of suspended sediment from the catchment of the proposed Alaoa Dam, estimated using three empirical equations from Holst Rice et al. (2016).

Application of the Fournier (1960) equation to estimate thean me annual load of suspended sediment of the catchment of the proposed Alaoa Dam using rainfall data from Alaoa and Afiamalu gave valuesTable ( 3 . 1 ) that were significantly higher than those suggested by application of the Fournier equation by Asquith et al. (1994) (100,114 and 123,946 t/yr). This is explained by the different (lower) rainfall data used by Asquith et al. (1994). Note also that the Fournier equation estimated much higher mean annual sediment loads than the empirical relationships of Holst Rice et al. (2016) (see above).

Table 3.1: Estimated sediment load from catchment of proposed Alaoa Dam using Fournier equation, calculated for two different rainfall stations and two assumptions for sediment specific gravity. ퟐ Rainfall 풑 ⁄푷 푺풚 Sediment yield mass Catchment sediment station Sediment yield (t/km^2/yr) load (t/yr) volume Specific Specific Specific Specific (m3/km2/yr) gravity 2.1 gravity 2.6 gravity 2.1 gravity 2.6 Alaoa 84.459 3,920 8,232 10,192 135,395 167,632 Afiamalu 117.771 5,669 11,904 14,738 195,788 242,402

Application of the BQART (Syvitski and Milliman, 2007) equation to estimate the mean annual load of suspended sediment of the catchment of the proposed Alaoa Dam gave a value of 100,702 t/yr (6,123 t/km2/yr). The BQART equation estimated somewhat lower sediment load than the Fournier equation (Table 3.1), but still much higher than the estimates made the empirical relationships of Holst Rice et al. (2016) (see above). Application of the RUSLE by Entura (2018) over 32 sub-catchments within the catchment of the proposed Alaoa Dam to give an estimate of mean annual total soil loss of 61,100 t/yr (3,715 t/km2/yr). Assuming a sediment delivery ratio of 1, the mean annual sediment load of the catchment would be the same. This estimate is lower than those of the Fournier and BQART equations, but still much higher than the estimates made the empirical relationships of Holst Rice et al. (2016) (see above). A summary of these results illustrates the range of estimates of mean annual suspended sediment load given by the methods applied (Table 3.2).

Table 3.2: Estimated sediment load from catchment of proposed Alaoa Dam and the entire Vaisigano River catchment, using various equations. 6.4.1 Catchment sediment load (t/yr) Method Alaoa Dam Vaisigano River catchment catchment Sediment concentration relationship (Holst Rice 8,867 - et al., 2016) Event total discharge relationship (Holst Rice et 1,269 - al., 2016) Event peak discharge relationship (Holst Rice et 108 - al., 2016) Fournier (1960) equation (this report) 135,395 - 242,402 373,501 – 462,430 BQART equation (Syvitski and Milliman, 2007) 100,702 221,901 RUSLE (Entura, 2018) 61,100 -

The Fournier (1960) equation was applied to the entire catchment of the proposed Alaoa Dam assuming rainfall data from Afiamalu applied to the Eastern branch, Middle eastern branch, Middle branch and Western branch, rainfall data from Alaoa applied to the Far western branch and area downstream of proposed Alaoa Dam, and rainfall data from Apia applied to the Lower catchment. The estimates of mean annual load of suspended sediment were 373,501 t/yr (10,855 t/km2/yr) assuming specific gravity of 2.1, and 462,430 t/yr (13,440 t/km2/yr) assuming specific gravity of 2.6 (Table 3.2). Application of the BQART (Syvitski and Milliman, 2007) equation to estimate the mean annual load of suspended sediment of the entire Vaisigano River catchment gave a value of 221,901 t/yr (6,651 t/km2/yr) (Table 3.2). The BQART equation estimated sediment load about half that of the Fournier equation.

3.5.2 Bed material particle size distribution

The median size of bed material found in the river system was large cobble class at the two headwater sites upstream of the proposed Alaoa Damite, sand small cobble class at the other sites (Figure 2 . 3 ) . However, the range of particle size at each site was large, covering sand to boulder size classes Table( 3 . 3 , Figure 3 . 23 ) . The poorly to moderately sorted nature of the bed material, characterised by the Folk and Ward (1957) sorting coefficient, suggests that the source of bed material was local, and the distance the material had been transported in the river system was relatively short. Exposed bedrock was prominent at Site 3 (upstream Samasoni Weir) and Site 6 (upstream of theinundation zone East branch).

Table 3.3: Characteristics of sampled stream bed material size distributions

Figure 3 . 23 : Particle size distributions of sampled stream bed material (excludes silt and sand material finer than 2 mm and exposed bedrock).

3.5.3 Bed material transport

The Meyer-Peter and Müller (1948) equation was applied to 245the m reach3.4 kilometres downstream Samasoni Weir that was surveyed and modelled using the one-dimensional HEC-RAS hydraulic model (Site 1b).Nine of 29 surveyed cross-sections were selected to represent the bedload transport process.The results indicated that bedload transport rate was highly variable within the reach, as dictated by the longitudinally variable hydraulic capacity of the Figureflow ( 3 . 24 ) . The central section of the reach had very low sediment transport capacity, while the downstream section of the reach, which had higher slope, would transport bed material at lower discharges, and would transport bed material at a higher rateFigure ( 3 . 24 ) .

The median bed material transport conditions of the modelled reach were represented by crosssection 26, at 235 m chainage (Figure 3.24). A rating curve was established for this cross-section to predict bed material transport rate as a continuous function of discharge, although transport did not commence until a flow rate of 40 m3/s was exceeded (i.e. the threshold flow to initiate sediment transport). This rating curve was applied to the 48-year hourly modelled discharge series for the node at 2.5 km downstream Samasoni Weir. The results indicated that annual bed material transport load (Figure 3.25) was 4 to 5 orders of magnitude lower than the suspended sediment transport load (Figure 3.22). Also, the current rate of bed material transport was similar to that under natural conditions (Figure 3.22).

Figure 3 . 24 : Estimated bedload transport rate across the full range of potential discharge over the 245 m long surveyed and modelled reach at Site 3.41b, kilometres downstream Samasoni Weir. Bed elevation and watersurface profile at 40 m 3 /s shown for reference.

Figure 3 . 25 : Estimated annual bedload transport load and peak annual daily rate at Site 1b, 3.4 kilometres downstream Samasoni Weir for the natural and current scenarios.

7. 4. Assessment of Project impact on the physical environment 7.1 4.1 Introduction to the impact assessment The proposed Alaoa Dam would not impact all of the aspects of the physical environment described in the previous section of this report. The main potential impacts of the proposed dam and its operation on geomorphic processes and forms are:

• Trapping of sediment in the dam reservoir, reducing the load of sediment to the river downstream, and to Apia Bay;

• Scour of the shoreline of the reservoir rim due to focus of wind-generated waves at the normal full supply level (NFSL);

• Potential mortality of trees, shrubs and ground cover on hillslopes within the reservoir flood storage area during extended periods of inundation during high inflows, with subsequent risk of hillslope instability;

• The rate of reservoir water level drawdown on flood recessions exceeding the rate at which hillslope soils can drain, with subsequent risk of hillslope slumping; and

• Altered hydrology reducing the frequency of bed material mobilisation and reducing bed material load.

7.2 4.2 Trapping of sediment in the reservoir A relationship between reservoir level and volume (capacity) was established (Figure 4.1) to facilitate application of sediment trapping efficiency equations. The Brune curve was based on normally ponded reservoirs, not flood control reservoirs like the proposed Alaoa Dam. The difficulty is that the Brune equation relies on a single value for reservoir capacity, when the proposed reservoir has variable capacity as the water level rises under flood event inflows and draws down during event recessions (Figure 4.2). Thus, at times of high inflows, when the majority of the annual sediment load is delivered to the reservoir, the water level is higher than the NFSL. The mean water level for each day of the year indicates that the water level would typically be elevated in December to April, particularly in February and March, but the water level would usually be well below the spillway (Figure 4.3). The water level 154.95 m asl, exceeded 3 percent of the time (Figure 4.3), was considered a reasonable level to use in application of the Brune curve. The volume at this level was 1.3445 × 106 m3 (Figure 4.1). The lower Brune curve for fine-grained sediment returned a value of 61.4% long-term sediment trap efficiency.

Figure 4 . 1 : Proposed Alaoa Dam reservoir relationship between water level and volume.

Figure 4 . 2 : Modelled daily time series of water level of proposed Alaoa Dam reservoir.

Figure 4 . 3 : Mean water level for each day of the year (top) and water level duration curve (bottom) for the proposed Alaoa Dam reservoir, derived from the modelled-year 48 hourly time series.

applied, as descried by Lewis et al. (2013), with some refinements, using- theyear 48 modelled daily

To overcome deficiencies of the Brune curve, the daily Churchill equation for trap efficiency was flow series, and daily calculated reservoir length, volume and cross-sectional surface area. The daily trap efficiencies (Figure 4.4) were used to estimate an annual time series of trap efficiency (Figure 4.4) by weighting daily trap efficiency by inflow volume, as descried by Lewis et al. (2013). The mean annual Churchill trap efficiency was 52.7 percent, significantly lower than the long-term trap efficiency calculated by the Brune curve.

Figure 4 . 4 : Daily Churchill trap efficiency (top), and annual Churchill trap efficiency compared with annual reservoir inflows (bottom).

The annual time series of Churchill trap efficiency was applied to the various estimates of suspended sediment load to the reservoir made in this report. The annual trap efficiencies were applied directly to the annual sediment loads estimated by the sediment concentration relationship of Holst Rice et al. (2016). The annual trap efficiencies were also applied to annual series of RUSLE, BQART and Fournier (upper estimate) long-term mean sediment loads (Table 3.2), created by weighting the estimated long-term mean sediment loads by annual dam inflow volume. The mass of sediment trapped was converted to volume by assuming a specific gravity of 1.3 for material deposited on the bed of the reservoir. The Alaoa Dam was designed with a 500,000 m3 dead storage volume (Entura, 2018). On the basis of catchment soil erosion estimated by the RUSLE and application of the Brune curve trap efficiency using a superseded estimate of mean annual dam inflows, Entura (2018) estimated that flushing of deposited sediment would be required at approximately 15-year

intervals. This would be achieved by opening the low level outlet, which was designed with a discharge passing capacity of 53.4 m3/s (Entura, 2018, p. 236). The cumulative volume of sediment trapped in the reservoir, calculated in this report using four different methods of estimating suspended sediment load to the reservoir, suggest that the frequency of sediment flushing required would vary from 5 to more than 50 years (Figure 4.5), depending on the estimate of sediment load. The RUSLE estimate was based on data from the Alaoa Dam catchment, so could be considered more reliable than the others. Application of the daily Churchill trap efficiency to the RUSLE sediment load estimate gave a mean annual volume of 24,368 m3/yr of sediment deposited, which is less than the estimate of 31,00 m3/yr made by Entura (2018) assuming a trap efficiency of 66%, as predicted by their application of the Brune curve. This lower sedimentation rate would increase the interval between sediment flushing to about 20 years

Figure 4 . 5 : Cumulative volume of sediment trapped in the reservoir, based on four estimates of annual sediment load.

(Figure 4.5). Richmond (1992) used data from North America to make an order of magnitude estimate of the load of sediment from the Vaisigano River to the Bay. The scale of the sediment load suggested to Richmond (1992) that most of the sediment was transported offshore to deeper waters. This estimate agreed with that of Rubin (1984), who estimated the annual rate of infilling of Apia Harbour (1,500 m3/year) was an order of magnitude lower than the volume of sediment that the Vaisigano River could supply in one day (~10,000 m3) under flood conditions. The results presented in this report support these assumptions. Thus, trapping of suspended sediment in Alaoa Dam reservoir would have minimal impact on the sediment dynamics of Apia Bay, because the volume of suspended sediment reaching the Bay would still vastly exceed the

volume that is deposited there annually. The trapping of sediment in Alaoa Dam reservoir would have minimal impact on the morphology of the Vaisigano River between the Alaoa Dam and Apia Bay, because there are few depositional zones (floodplains and wetlands) on the river where sediment deposition might be important, i.e. the vast majority of suspended sediment is flushed through the river system to Apia Bay, and then out to deeper water.

4.3 Scour of the shoreline at normal full supply level (NFSL) due to wind-generated

7.3 waves A narrow and mostly steep wave erosion zone will form around the rim of the reservoir at NFSL (Figure 4.6). The zone will be wider and less steep in the Eastern branch. The surface area of the erosion zone was 7,090 m2. Assuming a 300 mm depth of soil, the volume of sediment eroded from this zone would total 2,127 m3. This represents an insignificantly small volume of

sediment that would enter the reservoir relative to the annual volume of suspended sediment delivered to the reservoir by the river system (Table 3.2).

Figure 4.6: Predicted extent and slope of reservoir wave erosion zone at normal full supply level.

4.4 Duration of periods of inundation within the flood storage zone during high inflow

7.4 events Although the water level of the reservoir would be above NFSL for a relatively small percentage of the total time, the frequency of the water level exceeding NFSL will be quite high. For example, although a water level of 154.95 m (2.55 m higher than NFSL) would be exceeded for only 3 percent of the total time (Figure 4.3), this level would be exceeded in 88 percent of years (Figure 4.7). The water level would reach 168 m (15.6 m higher than NFSL) with an average frequency of approximately once every 10 years (Figure 4.7).

The duration of events higher than NFSL was highly variable, but the mean duration was 5 days or longer for levels between NFSL and 168m (Figure 4 . 8 ) , which is ample time for soils to become saturated. The estimated maximum duration of events higher than NFSL was– 40 20 days over much of the range of the flood storage zoneFigure ( 4 . 8 ) . The capacity of the vegetation within the Alaoa Dam reservoir to withstand submergence for periods of this length was beyond the scope this report, and would require assessment by a plant ecologist. However, a level of vegetation mortalityuld wo be expected, and this would almost certainly increase the risk of hillslope instability.

Figure 4 . 7 : Percent of years that water level was exceeded for the proposed Alaoa Dam reservoir, derived from the modelled 48-year hourly time series

Figure 4.8: Characteristics of spells of events exceeding a range of reservoir water levels.

7.5 4.5 Slumping of side slopes during reservoir drawdown The hourly rate of fall in reservoir water level was calculated from the 48-year modelled time series of water level. The statistics describing the rate of fall were calculated only for periods when the water level fell (i.e. times when water level was stable or rising were excluded). Extreme rates of fall, exceeding 1 metre over 1 hour, were associated with recession from above the spillway crest back to the spillway crest, during rare, extreme floods (Figure 4.9). Over the flood storage zone, extreme rates of fall were in the order of 70 – 140 mm/h. For more than half of the time that water level fell, the rates of fall were in the range 15 – 65 mm/h, varying over the flood storage zone (Figure 4.10). Given the large range of hydraulic conductivities of the soil type likely to be found in the reservoir area ( Table 2 . 2 and Table 2 . 3 ) , without site- specific soil data,it is not possible to make a definitive conclusionregarding the risk of soil slumping under rapid drawdown of reservoir water levels on flood recessions.However, the data suggest the potential for significant risk that the rates of fall in water level could often exceed the rate at which the soils canturally na drain.

The above analysis of factors impacting slope stability within the reservoirflood storage zone was undertaken to inform potential risk. The implications of this information for slope stability should be considered by a geotechnical engineer.

Figure 4 . 9 : Rates of fall exceeded 1% of the time (top) and 5% of the time (bottom) across the range of elevation of the flood storage zone to above the spillway crest at 174.6 m asl.

Figure 4 . 10 : Rates of fall exceeded 25% of the time (top) and 50% of the time (bottom) across the range of elevation of the flood storage zone to above the spillway crest at 174.6m asl.

Altered hydrology and geomorphic response

7.6 4.6 The modelled hourly discharge series from the node at 2.5 km downstream Samasoni Weir (Figure 2.1) was selected to represent the hydrological regime of the lower Vaisigano River with the Alaoa dam operational, and to characterise the degree of flow alteration relative to the natural and current conditions. This location is between Samasoni Weir and Samasoni Power Station, so the river is impacted by diversion of water at the weir to a pipe that returns the water to the river downstream at the power station. Other locations on the river system would be impacted by the current operation of hydropower plants, and future operation of Alaoa Dam, to varying degrees. Analysis of hydrological time series was undertaken for other locations for the environmental flow assessment (Elvey and Gippel, 2019).

The mean daily flow calculated for each day of the year suggests insignificant impact of the Alaoa Dam relative to the current regime (Figure 4.11). The flow duration curve illustrates the same lack of significant impact (Figure 4.12). Low flow indices (minimum hourly discharge, and flow exceeded for 95% of the time) and baseflow index (flow exceeded 50 percent of the time) for each month suggests insignificant impact of the Alaoa Dam relative to the current regime (Figure 4.13). High flow indices (flow exceeded for 1% and 0.1% of the time) for each month indicate that moderate flood flows and small flow events (freshes) would be significantly reduced in magnitude with Alaoa Dam operational relative to the current (and natural) regime (Figure 4.14). The impact, relative to current, was not significant in the months July to October/November. Flood frequency analysis suggests that with Alaoa Dam operational, for any recurrence interval, peak flood magnitude would be significantly reduced relative to the current (and natural) regime (Figure 4.15). The annual time series of annual flow indicates that with Alaoa Dam operational, the total flow each year would on average be 0.3 percent lower than under the natural regime, with the maximum reduction in any year being 1.1 percent (Figure 4.16). These are inconsequential differences. With Alaoa Dam operational, the annual peak hourly discharge would on average be 34 percent lower relative to the current regime (Figure 4.16). The Flow Health indicators were calculated using default statistical parameters, with the exception that the high flow season was manually specified as extending from November to May inclusive, consistent with the definition of the wet season by Elvey and Gippel (2019). As would be expected from the results of the other statistical comparisons made above, with Alaoa Dam operational, the integrative index classified the Vaisigano River 2.5 km downstream Samasoni Weir largely unmodified from current (Figure 4.17). The occurrence of sporadic scores outside the very small deviation class would be expected in unimpacted flow regimes. Flood Interval indicator suggested two periodic impacts in the time series (Figure 4.17). The Flow Health system is based on monthly flows, so is not directly sensitive to impacts on flood event peak magnitudes, however, the Flood Interval indicator did correctly identify an impact to flood flows. The median annual Flow Health score over the entire modelled series was 0.84 (Figure 4.18), which is within the very small deviation (unimpacted) class. Overall, the hydrological analysis, including application of Flow Health, suggested that the impact of Alaoa Dam on the hydrological regime of the Vaisigano River at 2.5 km downstream of Samasoni Weir would be very small compared to the degree of alteration imposed by the currently regime relative to natural. The main additional impact on the flow regime imposed by operation of Alaoa Dam would be on small to high flood event peak magnitudes, which would be significantly reduced. Some reaches of the river system would not be impacted to the same degree, but others might be impacted to a greater degree.

Figure 4 . 11 : Mean daily flow calculated for each day of the year at 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from-year 48 long hourly modelled time series.

Figure 4.12: Flow duration curve at 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from 48-year long hourly modelled time series.

Figure 4 . 14 : High flow indices (flow exceeded for 1% and 0.1% of the time) for each month, at 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from 48- year long hourly modelled time series.

Figure 4 . 15 : Partial duration series flood frequency for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled natural, current and ithw Alaoa Dam scenarios, based on 4th order polynomial curves fitted to frequency data from 45 years of modelled peak daily discharge data (2013 – 2057) .

Figure 4 . 16 : Annual flow and annual peak hourly discharge for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled natural, current and with Alaoa Dam scenarios. Calculated from 48-year long hourly modelled time series.

Figure 4.17: Flow Health individual and combined flow deviation indicator scores for the Vaisigano River 2.5 km downstream of Samasoni Weir for the modelled with Alaoa Dam scenario relative to the current scenario.

Given that the main hydrological impact of operation of Alaoa Dam on the regime of the Vaisigano River 2.5 km downstream of Samasoni Weir was to flood flows, the only expected geomorphic impact would be to channel maintenance (bankfull) flows that mobilise bed material and re-shape the banks. For the Vaisigano River 2.5 km downstream of Samasoni Weir, the daily peak flow of 1 year average recurrence interval (ARI) on the partial series under the current flow regime (Figure 4.15) was used as an index of the natural bankfull flow (for additional details, refer to Elvey and Gippel( 2019). The rationale for this is that in highly seasonal rivers with reasonably reliable rainfall from year to year (as in Samoa), the dynamic equilibrium (i.e. long-term average condition) channel form would be adjusted to events that occur almost every year. It was assumed that the river morphology was adjusted to the current flow regime. The flow level corresponding to the 1 year ARI event was found to be roughly equivalent to morphological bankfull on those surveyed cross-sections which displayed evidence of a channel-floodplain interface. Flows exceeding this level would create connectivity with floodplain environments. On the Vaisigano River 2.5 km downstream of Samasoni Weir, the 1 year ARI event was virtually the same under the natural and current conditions, but Alaoa Dam would reduce the 1 year ARI event peak magnitude from 74.3 m3/s to 48.5 m3/s (Figure 4.15). This was expected, as the primary function of Alaoa Dam is to reduce flood frequency in Apia. Application of the bed material transport predictive equation to the 48-year hourly modelled discharge series with Alaoa Dam operational indicated that annual bed material transport load would be markedly reduced relative to current conditions (Figure 4.19). With Alaoa Dam operational, for years experiencing sediment transport, the annual bedload transport would be on average 69 percent lower relative to the current regime. The impact would be to reduce the current average annual bedload load from 20.4 tonne to 4.8 tonne. Although this represents a large relative impact, the small absolute mass of annual bedload transport suggests that this is not a prominent geomorphic process of the Vaisigano River.

Despite the presence of the Alaoa Dam, floods would continue to occur on the Vaisigano River 2.5 km downstream of Samasoni Weir due to unregulated flows from other parts of the catchment. Thus, channel shaping processes would continue on the Vaisigano River 2.5 km downstream of Samasoni Weir but the river channel would likely slowly adjust to the altered flow regime by reducing its size.

Figure 4 . 19 : Estimated annual bedload transport load and peak annual daily rate at Site 1b, 3.4 kilometres downstream Samasoni Weir for the natural, current and with development scenarios.

The reach of river that would be most hydrologically impaired by the proposed Alaoa Dam is the 1.4 km-long reach immediately downstream of the dam wall to Samasoni Weir. The first 250 m section down to the existing Alaoa 1 power station outlet will only receive flow from normal dam operation when the mid-level outlet is releasing water, which is 3 m3/s on average 35 percent of the time annually (Elvey and Gippel, 2019). Alaoa 1 power station has a design discharge of 0.958 m3/s (JICA, 2003, p. 4-3). The proposed new power station, with design discharge of 1.33 m3/s (Entura, 2018, p. 234), would then release additional flow to the river a short distance downstream. Just above Samasoni Weir a small additional flow of 0.67 m3/s (JICA, 2003, p. 4-3) is contributed to the East branch by Fale o le Fe'e power station, and a major inflow is contributed by the West branch. While baseflows will be provided to this reach of river from Alaoa 1 power station downstream, and the mid-level dam outlet will release flows up to 3 m3/s for 35 percent of the time, it will not receive flood flows other than sediment flushing releases from the low level outlet, anticipated to be required every 15 to 20 years. The low level outlet was designed with a discharge passing capacity of 53.4 m3/s (Entura, 2018, p. 236).

It is not recommended to release an environmental flow from the base of the dam to minimise the potential for aquatic migratory species to congregate below the impassable barrier of the dam wall (Elvey and Gippel 2019). A 250 m long reach of river channel immediately downstream the dam to the outfall for the Alaoa power station will be dry most of the time and

The release of water from the low level outlet for the purpose of flushing sediment will involve release oflarge quantities ofsuspended sediment entrained from thematerial that will accumulate at the bottom of the dam. If not carefully managed, this procedure could cause deposition of significant quantities of sediment on the channel downstream.

the channel will by colonised by terrestrial vegetation. However, the flow regime and aquatic habitat in the rest of the 1.4 km reach will be maintained by the outflow from the existing Alaoa Power Station and the new power station (Elvey and Gippel 2019). For the 1.2 km reach from Alaoa Power Station to Samasoni Weir there would be no bed material transport for nearly all of the time, and little suspended sediment transport except when sediment flushing releases are made from the low level outlet.

8. 5. Mitigation and monitoring 8.1 5.1 Mitigation A number of potential geomorphic impacts and consequences were identified in this report (Table 5.1). The impacts were judged almost certain to occur, but the consequences were more difficult to rate, as the consequences were subjective and partly of an ecological nature (Table 5.1). Reduced suspended sediment delivery to the river system downstream of the dam cannot be mitigated, but the geomorphic consequence of this is insignificant because nearly all of the sediment that would otherwise have passed through the river system would have entered Apia Bay and then flowed out to deeper water (Table 5.2). Deposition of a large volume of fine sediment within the reservoir can only be mitigated by occasional flushing of the sediment by opening the low level offtake. This procedure was estimated to be required every 15 – 20 years, but the estimate is uncertain. The consequences of sediment flushing can be mitigated by arranging it to be done at a time when the catchment is experiencing a natural flood event, to assist transport of the turbid water to Apia Bay (Table 5.2). Downstream of the Dam, the rate of bedload transport will be reduced due to reduced magnitude of flood peaks. However, it appears that the natural rate of bedload transport is low, so the consequences of a reduction are likely to be minor (Table 5.2). With the dam operational, channel forming flows will operate at a lower magnitude, so the river will likely slowly adjust by contracting in width. The potential for mitigating this impact by environmental flows was investigated by Elvey and Gippel (2019). Their conclusion was that the most appropriate course of action was to allow the channel to contract and stabilise at a new dynamic equilibrium (Table 5.2). Scour of the shoreline of the reservoir rim due to the action of wind waves is inevitable. Mitigation of this would be impractical, so it is recommended to allow the shoreline to erode to bare rock (Table 5.2). Submergence and waterlogging of trees, shrubs and ground cover on hillslopes within the reservoir flood storage area, as well as rapid drawdown, will create the risk of loss of soil and vegetation. It is recommended to seek additional expert advice on this matter. Mitigation could take the form of management of water levels and maintenance of good cover of vegetation that will tolerate the hydrologic and hydraulic conditions within the flood storage area (Table 5.2).

8.2 5.2 Monitoring Monitoring of geomorphic form and process should focus on indicators that effectively characterise the main project impacts identified in this report, to determine if and to what extent the predicted impacts occur, and also the implemented mitigation measures, to determine the effectiveness or otherwise of the measures. The monitoring program would be a component of an adaptive approach to management of the dam and the Vaisigano River system. Monitoring that will assist sediment and erosion management includes:

• Annual sounding survey of the reservoir to determine sediment deposition rate

• Surveillance of the reservoir flood storage area immediately following drawdown of raised water levels in the wet season. Initial survey could be done by AUV (aerial unmanned vehicle), with any areas of potential vegetation death or soil instability inspected on the ground.

Table 5.1: Potential geomorphic impacts and consequences of the proposed Alaoa Dam. 8.2.1 Direct impact Likelihood Cause of impact Risk Consequence Reduced suspended ALMOST Trapping within Ecological impacts on INSIGNIFICANT sediment delivery to CERTAIN the reservoir organisms and the river system habitats that rely on downstream of the very high suspended dam sediment delivery Deposition of a large ALMOST Trapping within Exposure of MODERATE volume of fine CERTAIN the reservoir downstream sediment within the organisms and reservoir habitats to a high turbidity event during flushing operations, every 15 – 20 years Reduced rate of ALMOST Reduced flood Impaired bed habitat MINOR bedload transport CERTAIN magnitudes as a quality result of absorption of the majority of flood flows by operation of the reservoir

Channel forming ALMOST Reduced Contraction of MODERATE flows operating at a CERTAIN magnitude of flows channel area and lower magnitude of bankfull reduction of available frequency habitat area Scour of the ALMOST Focus of wind- Soil and vegetation MINOR shoreline of the CERTAIN generated waves loss over a narrow reservoir rim at the normal full bank through scour, supply level (NFSL) and contribution of sediment to the reservoir Submergence and ALMOST Extended periods Widespread mortality MAJOR waterlogging of CERTAIN of inundation of vegetation and trees, shrubs and during high inflows consequent soil loss ground cover on through hillslope hillslopes within the instability, and reservoir flood contribution of storage area sediment to the reservoir Saturated hillslope ALMOST Rate of drawdown Localised soil and MODERATE soils within the CERTAIN on flood recessions vegetation loss reservoir flood exceeding the rate through hillslope storage area at which hillslope slumping, and subjected to rapid soils can drain contribution of drawdown sediment to the reservoir

Table 5.2: Potential geomorphic impacts and suggested mitigations.

Direct impact Mitigation Reduced suspended sediment delivery to the No mitigation river system downstream of the dam Deposition of a large volume of fine sediment Manage flushing events to coincide with within the reservoir natural flood events Reduced rate of bedload transport No mitigation Channel forming flows operating at a lower Allow channel to contract and stabilise at a magnitude new dynamic equilibrium Scour of the shoreline of the reservoir rim No mitigation. Allow rim of reservoir to scour to bare rock Submergence and waterlogging of trees, shrubs Arrange for a plant physiologist to assess the and ground cover on hillslopes within the consequence. reservoir flood storage area Manage dam operation to minimise impact and manage hillslope vegetation to maximise resistance to erosion Subjecting saturated hillslope soils within the Arrange for a geotechnical engineer to assess reservoir flood storage area to rapidawdown dr the consequence. Manage dam operation to minimise impact and manage hillslope vegetation to maximise resistance to erosion

9. 6. References

Andrews, E.D. 1980. Effective and bankfull discharges of streams in the Yampa River Basin, Colorado and Wyoming. J. Hydrol. 46: 311–330. Alavi, G. and Tomer, M.D. 2001. Estimation of soil hydraulic parameters to simulate water flux in volcanic soils. New Zealand Journal of Forestry Science 31(1): 51-65. Allen, H. 2001. Shoreline erosion control plan, Lake Thunderbird, Cleveland County, Oklahoma. AllEnVironment Consulting, Vicksburg. Oklahoma Water Resources Board, Water Quality Programs Division, Oklahoma City. Andrews, E.D. 1980. Effective and bankfull discharges of streams in the Yampa River Basin, Colorado and Wyoming. J. Hydrol. 46: 311–330. Andrews, E.D. 1983. Entrainment of gravel from naturally sorted riverbed material. Bulletin of the Geological Society of America 94: 1225–1231.

Asquith, M., Kooge, F. and Morrison, R.J. 1994. Transporting sediments via rivers to the ocean, and the role of sediments as pollutants in the South Pacific. SPREP reports and studies series no. 72. Western Samoa South Pacific Regional Environment Programme, Apia, March. Atkinson, E. 1995. Methods for assessing sediment delivery in river systems. Technical Note. Hydrological Sciences -Journal- des Sciences Hydrologiques 40(2): 273-280. B Kjerfve, WJ Wiebe, HH Kremer, W Salomons and JI Marshall Crossland (Caribbean); N Morcom, N Harvey and JI Marshall Crossland (Oceania), 2002, Caribbean Basins: LOICZ Global Change Assessment and Synthesis of River Catchment/Island-Coastal Sea Interaction and Human Dimensions; with a desktop study of Oceania Basins. LOICZ Report & Studies No. 27, ii+174 pages, LOICZ, Texel, The Netherlands. Bagnold, R.A. 1977. Bedload transport by natural rivers. Water Resources Research 13: 303-312. Bagnold, R.A. 1980. An empirical correlation of bedload transport rates in flumes and natural rivers. Proceeding of the Royal Society London A 372: 453-473. Barry, J.J., Buffington, J.M. and King, J.G. 204. A general power equation for predicting bedload transport rates in gravel bed rivers. Water Resources Research 40(10), W10401, doi:10.1029/2004WR003190. Bennett, J.P. 1995, Algorithm for resistance to flow and transport in sand-bed channels. Journal of Hydraulic Engineering, ASCE 121(8), 578-590. Brizga, S. 1998. Methods addressing flow requirements for geomorphological purposes. In Arthington, A.H. and Zalucki, J.M. (eds) Comparative Evaluation of Environmental Flow Assessment Techniques: Review of Methods. Occasional Paper No. 27/98. Land and Water Resources Research & Development Corporation, Canberra, Australian Capital Territory, pp. 8–46. Brookes, A. 1995. The importance of high flows for riverine environments. In: Harper, D.M. and Ferguson, A.J.D. (eds), The Ecological Basis for River Management. John Wiley & Sons, Chichester, pp. 33-49. Brune, G.M. 1953. Trap efficiency of reservoirs. Trans. Am. Geophys. Union 34: 407–418. Buffington, J.M., and Montgomery, D.R. 1997. A systematic analysis of eight decades of incipient motion studies, with special reference to gravel-bedded rivers. Water Resources Research 33: 1993– 2029. Camacho-Tamayo1, J.H., Rubiano Sanabria, Y. and Santana, L.M. 2013. Management units based on the physical properties of an Oxisol. Journal of Soil Science and Plant Nutrition 13(4): 767-785. Cardoso, D.L. and Kaminski, T.B. 2008. Determination of the hydraulic conductivity of the soil in triaxial tests. In Central theme, technology for all: sharing the knowledge for development. Proceedings of the International Conference of Agricultural Engineering, XXXVII Brazilian Congress of Agricultural Engineering, International Livestock Environment Symposium - ILES VIII, Iguassu Falls City, Brazil, 31st August to 4th September, 2008. Carvalho, L.A. and Libardi, P.L. Condutividade hidráulica de um Latossolo Vermelho Amarelo, não- marelo, nãosaturado, utilizando- saturado, utilizando-se sonda de nêutrons se sonda de nêutrons se sonda de nêutro. Acta Scientiarum. Agronomy 32(1): 153-159. Chakrapani, G.J. 2005. Factors controlling variations in river sediment loads. Current Science 88(4): 569-575. Churchill, M.A. 1948. Discussion of paper by L. C. Gottschalk ‘‘Analy-ses and use of reservoir sedimentation data,’’ in Federal Inter-agency Sedimentation Conference Proceedings, U.S. Geol. Surv.,Denver, Colorado, pp. 139–140. Clark, K.E., Shanley, J.B., Scholl, M.A., Perdrial, N., Perdrial, J.N., Plante, A.F. and McDowell, W.H. 2017. Tropical river suspended sediment and solute dynamics in storms during an extreme drought. Water Resources Research 5(5): 3695-3712.

College of Tropical Agriculture and Human Resources 2019. Highly weathered tropical soils (Oxisol), Soils of Maui. Soil Nutrient Management for Maui County. University of HawaiˈI at Mānoa URL: https://www.ctahr.hawaii.edu/mauisoil/b_oxisol.aspx (accessed 15 Sep 2019). Cuevas, J., Horn, R., Seguel, O. and Dörner, J. 2013. Hydraulic conductivity variation in Chilean volcanic soils due to wheeling and management. Journal of Soil Science and Plant Nutrition 13(3): 756-766. de Vente, J., Poesen, J., Arabkhedri, M. and Verstraeten, G. 2007. The sediment delivery problem revisited. Progress in Physical Geography 31(2): 155–178. Dedkov, A.P. and Gusarov, A.V. 2006. Suspended sediment yield from continents into the World Ocean: spatial and temporal changeability. In Sediment Dynamics and the Hydromorphology of Fluvial Systems. Proceedings of a symposium held in Dundee, UK, July 2006. IAHS Publ. 306, pp. 3-11. Ekebom, J., Laihonen, P. and Suominen, T. 2002. Measuring fetch and estimating wave exposure in coastal areas. Littoral 2002, The Changing Coast. EUROCOAST / EUCC, Porto, Portugal, pp. 155-160. Elçi, S. and Work, P.A. 2003. Prediction of reservoir shoreline erosion. Proceedings of the 2003 Georgia Water Resources Conference, 23-24 April 2003, University of Georgia. K.J. Hatcher (ed.), Institute of Ecology, The University of Georgia, Athens, Georgia. Elçi, S., Work, P.A. and Hayter, E.J. 2007. Influence of stratification and shoreline erosion on reservoir sedimentation patterns. Journal of Hydraulic Engineering 133(3): 255-266. Entura 2018. Alaoa Multi-Purpose Dam Project, Draft Final Technical Feasibility Study Report. Prepared by Hydro-Electric Corporation, Entura, Cambridge, October. Eswaran, H. and Reich, P.F. 2005. World soil map. In Hillel, D. (Ed.) Encyclopedia of Soils in the Environment, Reference Module in Earth Systems and Environmental Sciences, Elsevier, pp. 352-365. Ferguson, R.I. 2005. Estimating critical stream power for bedload transport calculations in gravel-bed rivers. Geomorphology 70: 33-41. Ferreira, M.M., Fernandes, B. and Curi, N. 1999. Influência da mineralogia da fração argila nas propriedades físicas de latossolos da região sudeste do Brasil. Revista Brasileira de Ciência do Solo 23: 515–524. Fleißner, R and Dorfmann, C. 2013. Bed Load Analyzer. Software zur Berechnung von hydraulischen und sedimentologischen (Software to compute hydraulic and sedimentological values for irregularly shaped cross sections). Parametern in gegliederten Querschnitten. Referenzhandbuch. Version 2.0. Sustainicum, University of Natural Resources and Life Sciences Vienna, the University of Graz and the Graz University of Technology. Federal Ministry of Science and Research. February. Folk, R.L. and Ward, W.C. 1957. Brazos River Bar: a study in the significance of grain size parameters. Journal of Sedimentary Petrology 27(1): 3-26. Fournier, F. 1960. Climat et Erosion. PUF, Paris. Freeman, J.D. 1944. 'O le Fale o le Fe'e. The Journal of the Polynesian Society 53(4): 121-144. URL: http://www.jps.auckland.ac.nz/document//Volume_53_1944/Volume_53%2C_No._4/O_le_Fa le_o_l e_Fe%26%2339%3Be%2C_by_J._D._Freeman%2C_p_121-144/p1 (accessed 1/09/2019).

Fukuyama Shoji Company Limited 2013. Geotechnical Drilling Report Of Samoa Water Authority Alaoa Water Treatment Improvement Project. In Association with Isikuki Punivalu & Associates Limited. Meteorology Division, Ministry of Natural Resources and Environment, December. Gauss, G.A. 1981. Apia Harbour Survey, Samoa, 19 January - 6 February 1981, 20-31 March, 1981 Cruise WS 81(1). CCOP/SOPAC Cruise Report 55: 10 pages. South Pacific Applied Geoscience Commission.

Gippel, C.J. 2001. Geomorphic issues associated with environmental flow assessment in alluvial nontidal rivers. Australian Journal of Water Resources 5(1):3-19. Gippel, C.J., Marsh, N. and Grice, T. 2012. Flow Health - software to assess the deviation of river flows from reference and to design a monthly environmental flow regime. Technical Manual and User Guide, Version 2.0. ACEDP Australia-China Environment Development Partnership, River Health and Environmental Flow in China. International WaterCentre, Brisbane, Fluvial Systems Pty Ltd, Stockton, and Yorb Pty Ltd, Brisbane, September. URL: www.watercentre.org/. Gordon, N.D., McMahon, T.A., Finlayson, B.L., Gippel, C.J. and Nathan, R.J. 2004. Stream Hydrology: An Introduction for Ecologists. Second Edition, John Wiley & Sons, Chichester. Gray, J.R. and Simões, J.M. 2008. Estimating sediment discharge. Appendix D. In Garcia, M. (Ed.) Sedimentation engineering: processes, measurements, modeling, and practice. American Society of Civil Engineers. pp. 1067-1088. Hamilton, W.M. and Grange, L.I. 1938. The soils and agriculture of Western Samoa. New Zealand DSIR Bulletin 61, Wellington. Hicken, E.J. 1995. River Geomorphology. International Association of Geomorphologists Publication. Wiley, Chichester. Holden, B. 1991. Shoreline erosion and related problems, Western Samoa. SOPAC Preliminary Report 28. South Pacific Applied Geoscience Commission. Holst Rice, S., Messina, A.T., Biggs, T., Vargas-Angel, B. and Whitall. D.R. 2016. Baseline assessment of Faga’alu Watershed: a ridge to reef assessment in support of sediment reduction activities and future evaluation of their success. NOAA Technical Memorandum CRCP 23. Silver Spring, MD. 44 pp. doi:10.7289/V5BK19C3. URL: https://coastalscience.noaa.gov/data_reports/baseline-assessment-offagaalu-watershed-a- ridge-to-reef-assessment-in-support-of-sediment-reduction-activities-andfuture-evaluation-of- their-success/ (accessed 1 Sep 2019). Horowitz, A.J. 2003. An evaluation of sediment rating curves for estimating suspended sediment concentrations for subsequent flux calculations. Hydrol. Process. 17: 3387–3409. Jenson, S.K. and Domingue, J.O. 1988. Extracting topographic structure from digital elevation model data for geographic information system analysis. Photogrammetric Engineering & Remote Sensing 54(11): 1593–1600. JICA 2003. Hydropower generating plants and transmission/distribution system. Chapter 4 in Study on Electric Power Demand & Supply in Samoa. Japan International Cooperation Agency, Apia, March. Jowett, I.G. and Duncan, M.J. 1990. Flow variability in New Zealand rivers and its relationship to instream habitat and biota. N.Z. J. Marine and Freshwater Res. 24: 305-317. Kear, D. 1967. Geological notes on Western Samoa. New Zealand Journal of Geology and Geophysics 10(6): 1446-1451. Kear, D. and Wood, B.L. 1959. The geology and hydrology of Western Samoa. New Zealand Geological Survey Bulletin 63: 1-90. Keating, B.H. 1992. The Geology of the Samoan Islands. In Keating, B.H. and Bolton, B.R. (eds) Geology and offshore mineral resources of the central Pacific basin, Circum-Pacific Council for Energy and Mineral Resources Earth Sciences Series, V.14, Springer-Verlag. pp. 127-178. Kersch, P. Weixelberger, G. and Walk, J. 2013. Geology and Geotechnics, Technical Chapter – Annex. TA-8308 SAM: Renewable Energy Project – 1 Small Power Plant Scheme and Rehabilitation Plan (46044-001). Posh & Partners and Sustainable Technology Resources Pty Ltd. Electric Power Corporation (EPC), Asian Development Bank, December. Kirk, R.M., Komar, J.C. and Stephenson, W.J. 2000. Shoreline erosion on Lake Hawea, New Zealand, caused by high lake levels and storm-wave runup. Journal of Coastal Research 16(2): 346-356.

Kjerfve, B., Wiebe, W.J., Kremer, H.H., Salomons, W., Marshall Crossland, J.I., Morcom, N., and Harvey, N. 2002. Caribbean Basins: LOICZ Global Change Assessment and Synthesis of River Catchment/Island-Coastal Sea Interactions and Human Dimensions; with a desktop study of Oceania Basins. LOICZ Reports & Studies No. 27, ii + 174 pages, LOICZ IPO, Texel, The Netherlands. Kleinhans, M. and van Rijn, L. 2002. Stochastic prediction of sediment transport in sand-gravel bed rivers. J of Hydraulic Engineering, ASCE 128, Special Issue: Stochastic Hydraulics and Sediment Transport, 412-425. Komar, P.D. 1987. Selective gravel entrainment and the empirical evaluation of flow competence. Sedimentology 34:1165–1176. Kozlowski, T.T. 2002. Physiological-ecological impacts of flooding on riparian forest ecosystems. Wetlands 22(3): 550-561. Leopold, L 1970. An improved method for size distribution of stream bed gravel. Water Resources Research 6(5): 1357-1366. Leta, O.T., Dulaiova, H., Kadi-El, A., Messina, A. and Biggs, T. 2017. Assessing sediment yield and the effect of best management practices on sediment yield reduction for Tutuila island, American Samoa. Poster, AGU Fall Meeting 2017, New Orleans, American Geophysical Union. Lewis, S.E., Bainbridge, Z.T., Kuhnert, P.M., Sherman, B.S., Henderson, B., Dougall, C., Cooper, M. and Brodie, J.E. 2013. Calculating sediment trapping efficiencies for reservoirs in tropical settings: A case study from the Burdekin Falls Dam, NE Australia. Water Resources Research 49: 1017-1029. Liao, J.X., Mingxi, J. and Li, L.F. 2010. Effects of simulated submergence on survival and recovery growth of three species in water fluctuation zone of the Three Gorges Reservoir. Acta Ecologica Sinica 30(4): 216-220. Liu, Q.J., Shi, Z.H., Fang, N.F., Zhu, H.D. and Ai, L. 2016. Modeling the daily suspended sediment concentration in a hyperconcentrated river on the Loess Plateau, China, using the Wavelet–ANN approach. Geomorphology 186: 181-190. Lorang, M.S. and Hauer, F.R. 2003. Flow competence and streambed stability: an evaluation of technique and application. Journal of the North American Benthological Society 22(4):475–491. Lorang, M.S. and Stanford, J.A. 1983. Variability of shoreline erosion and accretion within a beach compartment of Flathead Lake, Montana. Limnol. Oceanogr. 38(8): 1783-I 795. Meyer-Peter, E. and Müller, R. 1948. Formulas for bed-load transport. Proceedings, 2nd Congress, International Association of Hydraulic Research, Stockholm, pp. 39-64. Moon, K., Duff, T.J. and Tolhurst, K.G. 2013. Characterising forest wind profiles for utilisation in fire spread models. In Proceedings 20th International Congress on Modelling and Simulation, Adelaide, Australia, 1–6 December 2013, pp. 214-220. New Zealand Geological Survey 1958. Geological Map No 2, Department of Scientific and Industrial Research, New Zealand Geological Survey Bulletin 63. New Zealand Soil Bureau 1956. Provisional Soil Map of Apia, Upolu, Western Samoa. Sheet 4. Soils by Wright, A.C.S. Department of Scientific and Industrial Research, New Zealand. Parker, G. 1990. Surface-based bedload transport relation for gravel rivers, J. Hydraul. Res. 28: 417 – 435. Reatto-Braga, A., Bruand, A., Silva, E.M., de Souza Martins, E. and Brossard, M. 2007. Hydraulic properties of the diagnostic horizon of Latosols of a regional toposequence across the Brazilian Central Plateau. Geoderma 139: 51-59. Regalado, C.M.and Muñoz-Carpena, R. 2004. Estimating the saturated hydraulic conductivity in a spatially variable soil with different permeameters: a stochastic Kozeny–Carman relation. Soil & Tillage Research 77: 189–202. Renard, K.G., Foster, G.R., Weesies, G.A. and Porter, J.P. 1991. RUSLE - Revised universal soil loss equation. Journal of Soil and Water Conservation. Jan.-Feb. 1991: 30-33

Richmond, B.M. 1992. Holocene geomorphology and reef history of islands in the south and central Pacific. Doctoral Thesis, University of California, Santa Cruz, March. Rubin D.M. 1984. Landfill materials and harbour surveys at Apia Harbour, Mulinu’u Point, Faleolo Airport, and Asau Harbour, Western Samoa, 15 May - 4 June 1984. CCOP/SOPAC Cruise Report 98: 8 pages. South Pacific Applied Geoscience Commission. Seelye, F.T., Grange, L.I. and Davies, L.H. 1938. The laterites of Western Samoa. Soil Science 46: 2331. Shields, A. 1936. Anwendung der Aehnlichkeitsmechanik und der Turbulenzforschung auf die Geschiebebewegung. Mitteilungen der Preuiβischen Versuchsanstalt fur Wasserbau und Schiffbau 26: 26 p. Silversides, R.H. 1978. Forest and airport wind speeds. Atmosphere-Ocean 16(3): 293-299. Siqueira, A.A., Arthur, C. and Woolf, M. 2014. Evaluation of severe wind hazard from tropical cyclones - current and future climate simulations. Pacific-Australia Climate Change Science and Adaptation Planning Program. Record 2014/47. Geoscience Australia, Canberra. Sobotkova, M., Snehota, M. Dohnal, M. and Ray, C. 2011. Determination of hydraulic properties of a tropical soil of Hawaii using column experiments and inverse model. R. Bras. Ci. Solo 35: 1229-1239. Solomon, S.M. 1994. A review of coastal processes and analysis of historical coastal change in the vicinity of Apia, Western Samoa. SOPAC Technical Report 208. Government of Canada, South Pacific Applied Geoscience Commission, June. Syvitski, J.P.M. and Milliman, J.D. 2007. Geology, geography, and humans battle for dominance over the delivery of fluvial sediment to the coastal ocean. J. Geol. 115: 1–19. Syvitski, J.P.M., Cohen, S., Kettner, A.J. and Brackenridge, G.R. 2014. How important and different are tropical rivers? — An overview. Geomorphology 227: 5-17. Tahir, H.M.M. and Yousif, T.A. 2013. Effect of urban trees on wind speed in Khartoum State. Journal of Natural Resources and Environmental Studies 1(2): 1-3. Terry, J.P., Kostaschuk, R.A. and Garimella, S. 2006. Sediment deposition rate in the Falefa Rive basin, Upolu Island, Samoa. Journal of Environmental Radioactivity 86: 45-63. U.S. Army Corps of Engineers 1989. Trap efficiency of reservoirs, Appendix F, Engineering design, sedimentation investigations of rivers and reservoirs. EM111-2-4000. Department of the Army, Washington D.C., December. U.S. Army Corps of Engineers 2016. HEC-RAS River Analysis System, User’s Manual, Version 5.0. U.S. Army corps of Engineers, Institute for Water Resources, Hydrologic Engineering Center, Davis, CA, February. USDA-SCS 1983: National engineering handbook, 2nd edition, Section 3, Sedimentation, Chapter 8 Sediment storage design criteria. US Department of Agriculture, Washington, DC. Verstraeten, G. and Poesen, J. 2000. Estimating trap efficiency of small reservoirs and ponds: methods and implications for the assessment of sediment yield. Progress in Physical Geography 24(2): 219-251. Walling, D.E. 1983: The sediment delivery problem. Journal of Hydrology 65: 209–37. Walling, D.E. 1999: Linking land use, erosion and sediment yields in river basins. Hydrobiologia 410: 223–40. Walling, D.E. and Webb, B.W. 1996. Erosion and sediment yield: a global overview. Erosion and Sediment Yield: Global and Regional Perspectives, Proceedings of the Exeter Symposium, July 1996. IAHS Publ. no. 236, pp. 3-19. Wang, C.Y., Li, C.X., Wei, H., Xie, Y.Z. and Han, W.J. 2016. Effects of Long-Term Periodic Submergence on Photosynthesis and Growth of Taxodium distichum and Taxodium ascendens Saplings in the Hydro-Fluctuation Zone of the Three Gorges Reservoir of China. PLoS One 11(9): e0162867, doi: 10.1371/journal.pone.0162867.

Wang, Q, Yuan, X.Z. and Liu, H. 2014. Influence of the Three Gorges Reservoir on the vegetation of Its drawdown area: effects of water submersion and temperature on seed germination of Xanthium sibiricum (Compositae). Polish J. of Ecology 62(1): 25-36. Williams, J.R. and Berndt, H.D. 1972. Sediment yield computed with universal equation. Journal of Hydraulic Division, ASCE 98: 2087-2098. Wolman, M.G. 1954. A method of sampling coarse river-bed material. Transactions of the American Geophysical Union 35(6): 951–956. Yang, C.T. 1996. Sediment transport: Theory and practice. McGraw-Hill, New York. Yang, F., Liu, W.W., Wang, J., Liao, L. and Wang, Y. 2012. Riparian vegetation’s responses to the new hydrological regimes from the Three Gorges Project: Clues to revegetation in reservoir water-levelfluctuation zone. Acta Ecologica Sinica 32(2): 89–98.