2.3. Photo-identification

When a dolphin(s) is sighted, the vessel leaves the transect line and slowly approaches the group or individual. Attempts are made to photograph every individual sighted although close approaches to mother and calf pairs are not attempted. A digital SLR camera (Nikon D90) using long lenses (Nikor 80-200mm and fixed length 300mm) are used to obtain high resolution images. Effort is made to ensure consistency of image quality, e.g., no shadow and at an angle perpendicular to the dorsal fin. Polarising filters are used to minimise glare. In this manner, the best image clarity is achieved and image sorting and matching is more consistent. Images are sorted according to clarity and presence/absence of identifying features (nicks/cuts/deformities/injury/pigmentation). Only images deemed to be of suitable quality and as containing sufficient markings for unambiguous identification are included in the photo-identification catalogue.

2.4. Data Analyses

2.4.1. Distribution pattern analysis

Dolphin sightings data are mapped in the Geographic Information System (GIS) ArcView© 10.1.

2.4.2. Encounter rate analysis

For this report, the baseline encounter rates were re-calculated using the revised data provided (as presented in Annex I) rather than quoting directly from the baseline report. Calculation followed the EM&A Manuel methodology (“on-effort” sightings made during favourable weather and visibility conditions).

2.4.3. Quantitative grid analysis of habitat use

Quantitative grid analysis is performed by mapping both sighting and dolphin densities plotted onto 1kmx1km grid squares. Only “on effort” sightings made while on a transect line and under favourable conditions should be included in grid analyses. These densities are standardised by effort by calculating survey coverage in each line transect survey to determine the number of times the grid has been surveyed. Densities are calculated using the following formulae;

SPSE and DPSE:

SPSE = (S/E x 100)/SA%

DPSE = (D/E x 100)/SA%

Where;

S= total number “on effort” sightings

D = total number dolphins from “on effort” sightings

E = total number units survey effort

SA% = percentage of sea area

2.4.4. Behavioural analysis

When dolphins are sighted during vessel surveys, their behaviour is observed. Different activities are categorised (i.e. feeding, traveling, surface active, associated with boats, unknown) and recorded in the sighting data form of Logger. The sightings form is integrated with survey effort and positional data and can be subsequently mapped to examine distribution and behavioural trends. All sightings data (“on-effort” and “opportunistic”) are used in this analysis.

2.4.5. Ranging pattern analysis

Home ranges for individual dolphins can be calculated using a variety of software (Worton 1989). In the baseline monitoring report, the program Animal Movement Analyst Extension, created by the Alaska Biological Science Centre, USGS was used in conjunction with ArcView© 3.1 and Spatial Analyst 2.0. Using the fixed kernel method, kernel density estimates and kernel density plots are created using all sightings. In the baseline monitoring, data from other studies and from outside the baseline monitoring period were used to map individual ranges. It is important to maximize the number of sightings used as kernel analyses cannot be conducted unless more than 20 independent sightings are made for an individual although it is recommended that a minimum of 70 resightings are used before kernel analyses has any accuracy (Wauters et al 2007; Kauhala and Auttila 2010). AFCD Annual Reports use a minimum of 15 resightings for kernel analyses (AFCD 2012). To date, too few data on individual dolphins exist from impact monitoring alone, i.e., 15 or more independent resightings per individual, to map utilisation densities using the fixed kernel method. The most resightings for an individual dolphin in the baseline and impact monitoring period combined is thirteen (HZMB 054) split across baseline (seven sightings) and impact monitoring (6 sightings). A comparison of baseline and impact sightings using kernel analyses will require longer term data collection.

3. RESULTS AND DISCUSSIONS

3.1. Summary of survey effort and dolphin sightings

From September – November, 12 vessel surveys were conducted in NEL and NWL survey areas (Annex II). A total of 668.2 km of “on-effort” transect lines were conducted, of which 665.9km were under favourable conditions. Therefore, 99.7% of vessel surveys were conducted under favorable conditions (Annex III). Only those periods of “on-effort” survey conducted under favourable conditions were included in quantitative analyses. During September – November 2013, 42 groups of dolphins, numbering 133 (min 131: max 143^{^[4]}) individuals, were sighted from the vessel surveys. Of these, 28 groups were “on-effort” and the remaining 14 “opportunistic” (Annex IV).

Of the 42 sightings, 41 groups were located in NWL and 1 in NEL. The baseline report, conducted during September-November 2011, notes a total of 44 groups, 34 of which occurred in NWL and 10 in NEL. For period September – November 2012, a total of 71 groups were sighted, 53 of which were located in NWL and 18 in NEL. There are differences between the number of sightings made during baseline compared to the same period in 2012 and 2013. For NEL, the number of groups almost doubled between baseline (2011) and September – November 2012 and then decreased markedly in September – November 2013. For NWL, both September – November 2012 and 2013 recorded larger numbers of groups when compared to baseline monitoring (Table 2). Maps depicting location of sightings which have not been corrected for effort or survey track length are included as Figs. 3;4;5;6.

Table 2. A Comparison of Total Sightings Recorded in NEL and NWL Areas During Sep – Nov 2011; 2012 and 2013

Monitoring Period	Total Dolphin Sighting in NWL	Total Dolphin Sighting in NEL
Monitoring Period	Number of Groups	Number of Groups
Sep – Nov 2011* (Baseline Monitoring)	34	10
Sep – Nov 2012* (HKBCF Third Quarter)	53	18
Sep – Nov 2013* (HKBCF Seventh Quarter)	41	1

* All Surveys conducted once per month

As per the EM&A manual, only “on effort” sightings can be used for some analyses therefore the combined number of “on effort” sightings for all three periods was compared There is an increase in the total number of “on effort” sightings between baseline monitoring (2011) and impact monitoring (2012) but a decrease below both previous totals in September – November 2013 (Table 3). No correction for effort is made with these numbers, this is calculated in section 3.3.

Table 3. A Comparison of “On Effort” Sightings Recorded in NEL and NWL Combined During Sep – Nov 2011; 2012 and 2013.

Monitoring Period	Groups of Dolphin sighted in NEL and NWL
Sep - Nov 2011 (Baseline Monitoring)	44
Sep – Nov 2012 (HKBCF Third Quarter)	52
Sep – Nov 2013 (HKBCF Seventh Quarter)	28

3.2. Distribution

During the both the baseline survey and the same period the following year, approximately three quarters of all “on effort” sightings were made in NWL. For the period September – November 2013, however, no “on effort” sightings were made in NEL. The similarity between the 2011 and 2012 periods is noted, however, there is no correction for effort (Table 4). Throughout September – November 2013, the area of most use was the northern section of NWL, within and adjacent to the Shau Chau and Lung Kwu Chau Marine Park (SCLKCMP). A few groups were recorded at the Tai O area and only two groups at the E edge of the airport platform. Only one group was recorded in NEL at the east of the Brothers Islands (Fig. 6). These areas are highlighted consistently throughout AFCD annual monitoring reports as well as during pre construction monitoring. SCLKCMP is frequented all year round by dolphins and is perceived to be critical habitat whereas the use of NEL is regarded as more seasonal, however, the decrease in the number of groups occurring in NEL compared to the same season in previous years is noteworthy.

Table 4. A Comparison of “On Effort” Sightings Recorded in NEL and NWL During Sep – Nov 2011; 2012 and 2013.

Monitoring Period	No. of Dolphin Groups sighted in NWL	No. of Dolphin Groups sighted in NEL
Sep - Nov 2011 (Baseline Monitoring)	34	10
Sep – Nov 2012 (HKBCF Third Quarter)	39	13
Sep – Nov 2013 (HKBCF Seventh Quarter)	28	0

3.3. Encounter rate

As the survey periods have different transect lengths, variation in sightings occurrence was quantified by correcting for the different amount of effort (number and distance of transect lines surveyed, i.e., km spent “on-effort”), to obtain an encounter rate. The baseline study (Sep-Nov 2011) reports that a total of 545.6km^{^[5]} of survey effort was conducted under favourable conditions in the NEL and NWL survey areas. In NEL and NWL combined, 659.8km and 665.9km of track-line were conducted under favourable conditions during the periods September – November 2012 and 2013, respectively. In NEL, there is a slight increase in encounter rate between baseline and the same period the following year, however, for the period September – November 2013, there is a marked decrease in encounter rate. i.e., 6.3 to 0. For NWL, there is a continuous decline in encounter rate from the baseline period through the same period the next and subsequent years, i.e., 9.5 to 8.7 to 6.3 (Table 5).

Table 5. A Comparison of Encounter Rates^* in NEL and NWL Areas During September-November 2011; 2012 and 2013.

Monitoring Period	Encounter Rate NEL	*Encounter Rate NWL ()**
Sept-Nov 2011 (Baseline Monitoring)	5.4	9.5
Sep – Nov 2012 (HKBCF Third Quarter)	5.9	8.9
Sep – Nov 2013 (HKBCF Seventh Quarter)	0	6.3

The AFCD Annual Reports describe variation in spatial distribution between areas and between seasons in NEL and NWL. For the last sixteen years, it is reported that overall annual encounter rate for NEL varies between 1.6 and 6.2 and the annual encounter rate for NWL varies between 5.8 and 17.0. The encounter rate for NWL for all three periods (September – November 2011; 2012; 2013) is within the annual limits recorded for this area previously. For NEL, the encounter rates in September – November 2011 and 2012 are within the recorded annual norms for the area, however, the encounter rate of zero for the same period 2013 is not. Historically, there have been both up and down movements within these limits, however, the general trend in yearly encounter rate for dolphins in all areas of Hong Kong is that of significant decline over the last decade and prior to new development projects in the Lantau area (AFCD 2013). The known decline in the population, on top of the highly variable encounter rate noted historically, makes it problematic to discern any additional influence individual projects, such as HKBCF and others, may have on the dolphin population encounter rate. As the impact of the work at HKBCF extends in addition to new dredging and other projects being initiated in NEL, it is likely that these activities have effected NEL encounter rates.

3.4. Group size

During September – November 2013, group size of all sightings varied from 1 to 12 individuals with an average of 3.2 in NWL and 1 in NEL. For baseline monitoring, the NWL average group size was 4.5 and the NEL average group size was 3.5. For the period September 2012, the NWL average group size was 3.1 and in NEL it was 3.6 (Table 6). NEL shows a decreased number in groups size (although it is noted only one group was sighted in September – November 2013). In NWL, groups sizes between September – November 2012 and 2013 are approximately the same although both are lower than the baseline monitoring. A map depicting group size distribution shows that the two largest groups occurred at SCLKCMP and both of these groups contained calves (Fig. 7).

Table 6. A Comparison of Sightings Group Size Averages Recorded in NEL and NWL Areas During Sep – Nov 2011; 2012 and 2013

Monitoring Period	Average Group Size (NWL)	Average Group Size (NEL)
Sep - Nov 2011 (Baseline Monitoring)	4.5	3.5
Sep – Nov 2012 (HKBCF First Quarter)	3.1	3.6
Sep – Nov 2013 (HKBCF Seventh Quarter)	3.2	1.0

As encounter rate and group size are both subject to variation, the use of other more powerful analyses may be more appropriate to discern differences over the shorter term, such as multi-variate analyses (Taylor et al 2007). This is important so that project impact can be monitored over relevant time scales. Alternative analyses have been proposed and developed using the first year of impact monitoring data and the methodology is attached (Annex V). Considerable reformatting of baseline data and incorporation of environmental and habitat data from multiple sources is near completion and the models will be run over the next quarterly period ^{^[6]}

3.5. Habitat use

Quantitative grid analyses indicates that the most often frequented areas in NWL were the SCLKCMP, the western limit of NWL and one area to the north of the Hong Kong International Airport (HKIA) platform. In NEL, no “on effort” sightings occurred therefore no qualitative grid analyses can be conducted (Figs. 8; 9). The grid analyses from this quarter shows a similar distribution in NWL to that published in the AFCD long term monitoring reports and the baseline monitoring report. These areas of high use have been consistent in the long term and continue to be so. The decrease in dolphin sightings between September – November 2013 and the two previous autumn seasons is noted. It is also noted that the areas of DPSE and SPSE which were apparent in summer 2013 are also absent.

3.6. Mother-calf pairs

Seven of the groups sighted contained mother and calf pairs[7]. All groups were sighted in NWL (Fig. 10). Calves comprised 6.7% of all dolphins sighted, much higher than that reported in the last quarterly report (2.5%). Although calf mortality was highlighted in previous seasons, several new born dolphins have been sighted consistently in NWL this quarter.

3.7. Activities

Of the 42 groups sighted (using all sightings), 21 (50%) were engaged in feeding activities; three (7%) were travelling; 11 (26%) were feeding/travelling/surface active; two (5%) were milling (other) and it was not possible to define the behavior of five (12%) groups. Feeding was the predominant activity during daylight hours in September – November 2013 with travelling/feeding/surface active (multiple) behaviours being the next most dominant behavioural category (Fig. 11). In NWL, feeding occurred most often at east SCLKCMP and the western limits of NWL. (Fig. 12).

3.8. Photo-identification work

The photo-identification catalogue was regularly updated and re-sightings of dolphins previously identified were recorded. The project specific photo-identification catalogue for the impact monitoring period is presented in Annex VI. Not all dolphins sighted have sufficient scarring, injury or pigmentation uniqueness to be unambiguously identified. During the baseline survey, 96 individuals were noted in the NEL, NWL and WL areas. Of these, 57 were noted in the NEL and NWL area. No new dolphins which have been identified in the last quarter are from the baseline study, however, several well known individuals have been recorded throughout September – November 2013. There are five dolphins which have been sighted more than seven times, all of which are known from the AFCD catalogue (HZMB 002 [WL111]; HZMB 011 [EL01]; HZMB 041 [NL24]; HZMB 044 [NL98]; HZMB054 [CH34]). Two of these well known individuals were not seen during the baseline study (HZMB 002 AND HZMB 044). When both baseline and impact monitoring data is pulled, HZMB 54 has been seen the most in 14 different sighting groups. HZMB 041 has been sighted 11 times; HZMB 002 has been sighted ten times, HZMB 044 has been sighted nine times and HZMB 011 has been sighted eight times. Even when pooled with baseline data, the highest number of re-sightings is 14 (HZMB 054) and this does not consider independence of sightings, a critical assumption in kernel analyses. (Annex VI; Table1).

4. CONCLUSION

The data from September – November 2013 shows some consistencies with the results reported in the same period 2011 (baseline) and 2012. Habitat use, encounter rates, group size and behavioural trends all fall within those reported in AFCD Long Term Monitoring reports apart from the use of NEL which has dropped when compared to this season in previous years. It is noted from the previous quarterly report which summaries the summer seasons for the last three years, there was an increase in seasonal usage of NEL. Density distribution maps depicted key areas of frequent use within NWL, in particular, SCLKMP and Tai O. Behavioural patterns were broadly similar, with feeding behavior predominating all months. .

The decrease in encounter rate in NEL is noted although no link can be found between this and specific activities at HKBCF. Although it is likely that the increase in HKBCF activities is having an effect on dolphin encounter rates in NEL (although not in the three months that preceded this quarter), it is also noted that a dredging project was started in November 2013 and new project works also were initiated in NEL during the autumn of 2013. Increased activities in addition to the inherent variation apparent in dolphin distribution and habitat use make it challenging to discern specific sources of impact. A significant decline in dolphin throughout the last ten years prior to construction commencement has also been published by AFCD (2013). It is hoped that the fine scale density surface analysis presently being conducted will allow further light to be shed on the specific areas and environmental variables, including marine construction works, which impacts dolphin distribution throughout NEL and NWL.

References

Agriculture, Fisheries and Conservation Department (AFCD) 2012. Annual Marne Mammal Monitoring Programme April 2011-March 2012. ) The Agriculture, Fisheries and Conservation Department, Government of the Hong Kong SAR.

Buckland, S., Burnham, K., Laake, J., Borchers, D. and Thomas, L. 2001. Introduction to Distance Sampling. Oxford University Press.

Connor, R. Mann, J., Tyack, P. and Whitehead, H. 1998. Social Evolution in Toothed Whales. Trends in Ecology and Evolution 13, 228-232

Gillespie, D., Leaper, R., Gordon, J. and Macleod, K. 2010. An integrated data collection system for line transect surveys. J. Cetacean Res. Manage. 11(3): 217–227.

Kauhala, K. & Auttila, M. 2010: Estimating habitat selection of badgers - a test between different methods. - Folia Zoologica 59: 16-25.

Taylor, B., Martinez, M, Gerodette, T., Barlow, J and Hrovat, Y. 2007. Lessons from Monitoring Trends in Abundance of Marine Mammals. Marine Mammal Science 23(1):157-175.

Wauters, L., Preatoni, D., Molinari, A. and Tosi, G. 2007. Radio-tracking squirrels: Performance of home range density and linkage estimators with small range and sample size. Ecological Modelling 202(10):333-44

Worton, B. 1989. Kernel Methods for Estimating Utilization Distribution in Home Range Studies. Ecology 70(I):164-8

Annex I Summary of Data from the Baseline Monitoring and September – November 2012 and September – November 2013 (this study) and Calculated Encounter Rates

Date	Area	Groups	Trackline (KM)	Encounter rate
Sept – Nov 2011 (Baseline Monitoring)	NEL	10	175.7	5.4
Sept – Nov 2011 (Baseline Monitoring)	NWL	34	359.0	9.5
Sept-Nov 2012 (Third Quarter)	NEL	13	221	5.9
Sept-Nov 2012 (Third Quarter)	NWL	39	438.8	8.9
Sept-Nov 2013 (Seventh Quarter)	NEL	0	220.7	0
Sept-Nov 2013 (Seventh Quarter)	NWL	28	445.2	6.3

Annex II. Impact Monitoring Survey Schedule and Details (September – November 2013)

Date	Location	No. Sightings “on effort”	No. Sightings “opportunistic”	Total km “on effort”
17/09/2013	NE Lantau	0	0	33.5
19/09/2013	NW & NE Lantau	1	3	77.5
24/09/2013	NW Lantau	6	0	63.4
25/09/2013	NW & NE Lantau	0	0	47.6
15/10/2013	NW Lantau	6	3	59.7
17/10/2013	NW & NE Lantau	0	0	52.1
24/10/2013	NW Lantau	5	2	58.7
28/10/2013	NW & NE Lantau	0	0	51.8
01/11/2013	NE and NW Lantau	2	3	59.5
02/11/2013	NWL	1	3	52.2
07/11/2013	NWL	6	0	64.7
09/11/2013	NE and NW Lantau	1	0	47.5
	Total	28	14	668.2

All effort in all sea states is listed

Annex III. Impact Monitoring Survey Effort Summary (September – November 2013)

Date	Area	Beaufort	Effort (km)	Season	Vessel	Type
Date	Area	Beaufort	Effort (km)	Season	Vessel	Type
17/9/2013	NEL	1	9.2	AUTUMN	HKDW	IMPACT
17/9/2013	NEL	2	15.6	AUTUMN	HKDW	IMPACT
17/9/2013	NEL	3	7.7	AUTUMN	HKDW	IMPACT
17/9/2013	NEL	4	1	AUTUMN	HKDW	IMPACT
19/9/2013	NEL	2	3.5	AUTUMN	HKDW	IMPACT
19/9/2013	NWL	1	10.6	AUTUMN	HKDW	IMPACT
19/9/2013	NWL	2	44.7	AUTUMN	HKDW	IMPACT
19/9/2013	NWL	3	18.7	AUTUMN	HKDW	IMPACT
24/9/2013	NWL	1	23.6	AUTUMN	HKDW	IMPACT
24/9/2013	NWL	2	19.4	AUTUMN	HKDW	IMPACT
24/9/2013	NWL	3	20.4	AUTUMN	HKDW	IMPACT
25/9/2013	NWL	1	7.6	AUTUMN	HKDW	IMPACT
25/9/2013	NWL	2	2.7	AUTUMN	HKDW	IMPACT
25/9/2013	NEL	1	20.3	AUTUMN	HKDW	IMPACT
25/9/2013	NEL	2	17	AUTUMN	HKDW	IMPACT
15/10/2013	NWL	1	35.8	AUTUMN	HKDW	IMPACT
15/10/2013	NWL	2	23.9	AUTUMN	HKDW	IMPACT
17/10/2013	NWL	1	1.1	AUTUMN	HKDW	IMPACT
17/10/2013	NWL	2	7.4	AUTUMN	HKDW	IMPACT
17/10/2013	NWL	3	6	AUTUMN	HKDW	IMPACT
17/10/2013	NEL	1	9.2	AUTUMN	HKDW	IMPACT
17/10/2013	NEL	2	20.5	AUTUMN	HKDW	IMPACT
17/10/2013	NEL	3	7.9	AUTUMN	HKDW	IMPACT
24/10/2013	NWL	1	12.2	AUTUMN	HKDW	IMPACT
24/10/2013	NWL	2	32.7	AUTUMN	HKDW	IMPACT
24/10/2013	NWL	3	13.7	AUTUMN	HKDW	IMPACT
24/10/2013	NWL	4	0.1	AUTUMN	HKDW	IMPACT
28/10/2013	NWL	1	4.9	AUTUMN	HKDW	IMPACT
28/10/2013	NWL	2	10.2	AUTUMN	HKDW	IMPACT
28/10/2013	NEL	1	14.6	AUTUMN	HKDW	IMPACT
28/10/2013	NEL	2	10.7	AUTUMN	HKDW	IMPACT
28/10/2013	NEL	3	11.4	AUTUMN	HKDW	IMPACT
11/1/2013	NEL	1	35.4	AUTUMN	HKDW	IMPACT
11/1/2013	NWL	1	14.6	AUTUMN	HKDW	IMPACT
11/1/2013	NWL	2	9.5	AUTUMN	HKDW	IMPACT

Annex III. Impact Monitoring Survey Effort Summary (September – November 2013) (con)
Date	Area	Beaufort	Effort (km)	Season	Vessel	Type
11/2/2013	NWL	2	26.7	AUTUMN	HKDW	IMPACT
11/2/2013	NWL	3	24.3	AUTUMN	HKDW	IMPACT
11/2/2013	NWL	4	1.2	AUTUMN	HKDW	IMPACT
11/7/2013	NWL	1	43.4	AUTUMN	HKDW	IMPACT
11/7/2013	NWL	2	21.3	AUTUMN	HKDW	IMPACT
11/9/2013	NEL	1	10	AUTUMN	HKDW	IMPACT
11/9/2013	NEL	2	21.2	AUTUMN	HKDW	IMPACT
11/9/2013	NEL	3	6.5	AUTUMN	HKDW	IMPACT
11/9/2013	NWL	1	3.7	AUTUMN	HKDW	IMPACT
11/9/2013	NWL	2	6.1	AUTUMN	HKDW	IMPACT

Annex IV. Impact Monitoring Sighting Database (September – November 2013)

Project

Contract

Date

Sighting No.

Time

Group Size

Area

Beaufort

PSD

Effort

Type

Latitude

Longitude

Season

Boat Assoc

HKBCF

HY/2010/02

9/19/2013

791

9:42

NEL

Opp

Impact

22.33573

113.9952

Autumn

HKBCF

HY/2010/02

9/19/2013

792

11:31

NWL

Opp

Impact

22.35992

113.9283

Autumn

HKBCF

HY/2010/02

9/19/2013

794

14:13

NWL

Impact

22.39786

113.8874

Autumn

HKBCF

HY/2010/02

9/19/2013

795

14:31

NWL

Opp

Impact

22.39240

113.8843

Autumn

HKBCF

HY/2010/02

9/24/2013

798

9:19

NWL

243

Impact

22.28493

113.8701

Autumn

HKBCF

HY/2010/02

9/24/2013

799

10:02

NWL

279

Impact

22.29694

113.8703

Autumn

HKBCF

HY/2010/02

9/24/2013

800

10:15

NWL

Impact

22.30655

113.8702

Autumn

HKBCF

HY/2010/02

9/24/2013

802

10:42

NWL

Impact

22.34161

113.8700

Autumn

HKBCF

HY/2010/02

9/24/2013

803

13:33

NWL

221

Impact

22.35825

113.8877

Autumn

HKBCF

HY/2010/02

9/24/2013

804

14:41

NWL

Impact

22.37340

113.8975

Autumn

HKBCF

HY/2010/02

10/15/2013

812

9:48

NWL

365

Impact

22.33154

113.8662

Autumn

HKBCF

HY/2010/02

10/15/2013

813

9:51

NWL

347

Impact

22.32847

113.8704

Autumn

HKBCF

HY/2010/02

10/15/2013

814

9:53

NWL

Impact

22.32959

113.8702

Autumn

HKBCF

HY/2010/02

10/15/2013

815

10:11

NWL

Impact

22.33614

113.8703

Autumn

HKBCF

HY/2010/02

10/15/2013

817

10:43

NWL

Opp

Impact

22.34800

113.8630

Autumn

HKBCF

HY/2010/02

10/15/2013

818

10:44

NWL

Opp

Impact

22.34564

113.8637

Autumn

HKBCF

HY/2010/02

10/15/2013

819

11:15

NWL

Impact

22.37013

113.8700

Autumn

HKBCF

HY/2010/02

10/15/2013

820

12:15

NWL

Impact

22.40074

113.8776

Autumn

HKBCF

HY/2010/02

10/15/2013

821

16:33

NWL

Opp

Impact

22.37510

113.9123

Autumn

HKBCF

HY/2010/02

10/24/2013

827

11:00

NWL

441

Impact

22.36340

113.8746

Autumn

HKBCF

HY/2010/02

10/24/2013

829

13:24

NWL

Impact

22.36996

113.8874

Autumn

HKBCF

HY/2010/02

10/24/2013

830

14:04

NWL

Impact

22.40109

113.8975

Autumn

HKBCF

HY/2010/02

10/24/2013

831

14:34

NWL

848

Impact

22.39506

113.8968

Autumn

HKBCF

HY/2010/02

10/24/2013

832

15:07

NWL

Opp

Impact

22.39703

113.9011

Autumn

HKBCF

HY/2010/02

10/24/2013

833

15:11

NWL

Opp

Impact

22.39688

113.8988

Autumn

HKBCF

HY/2010/02

10/24/2013

834

15:21

NWL

Impact

22.37427

113.8976

Autumn

Annex IV. Impact Monitoring Sighting Database (September – November) (con)

Project

Contract

Date

Sighting No.

Time

Group Size

Area

Beaufort

PSD

Effort

Type

Latitude

Longitude

Season

Boat Assoc

HKBCF

HY/2010/02

11/1/2013

837

13:34

NWL

Opp

Impact

22.35255

113.9363

Autumn

HKBCF

HY/2010/02

11/1/2013

838

14:25

NWL

Opp

Impact

22.36760

113.9164

Autumn

HKBCF

HY/2010/02

11/1/2013

839

15:26

NWL

112

Impact

22.36538

113.9076

Autumn

HKBCF

HY/2010/02

11/1/2013

841

16:17

NWL

173

Impact

22.35673

113.9079

Autumn

HKBCF

HY/2010/02

11/1/2013

842

17:08

NWL

Opp

Impact

22.32953

113.9332

Autumn

HKBCF

HY/2010/02

11/2/2013

845

11:44

NWL

594

Impact

22.34556

113.8780

Autumn

HKBCF

HY/2010/02

11/2/2013

847

14:38

NWL

Opp

Impact

22.26560

113.8617

Autumn

HKBCF

HY/2010/02

11/2/2013

849

15:01

NWL

Opp

Impact

22.25650

113.8488

Autumn

HKBCF

HY/2010/02

11/2/2013

850

15:43

NWL

Opp

Impact

22.26079

113.8516

Autumn

HKBCF

HY/2010/02

11/7/2013

853

9:08

NWL

141

Impact

22.27083

113.8703

Autumn

HKBCF

HY/2010/02

11/7/2013

854

9:46

NWL

109

Impact

22.30234

113.8706

Autumn

HKBCF

HY/2010/02

11/7/2013

855

10:38

NWL

Impact

22.36685

113.8701

Autumn

HKBCF

HY/2010/02

11/7/2013

856

12:13

NWL

169

Impact

22.35167

113.8778

Autumn

HKBCF

HY/2010/02

11/7/2013

857

15:31

NWL

Impact

22.37546

113.9072

Autumn

HKBCF

HY/2010/02

11/7/2013

858

16:31

NWL

Impact

22.35391

113.9170

Autumn

HKBCF

HY/2010/02

11/9/2013

860

10:49

NWL

Impact

22.33698

113.9463

Autumn

Annex V

METHODS PROPOSAL FOR DENSITY SURFACE MODELLING AND POWER ANALYSES RELATED TO THE HONG KONG-ZHUHAI-MACAO BRIDGE (HZMB)

OVERVIEW

This document outlines the proposed statistical analysis of data concerning the Chinese White Dolphin (CWD) found in and around the HZMB area. The proposal involves the analysis of baseline monitoring data and data collected during and post construction.

This proposal outlines statistical analyses for comparing differences in dolphin densities between baseline and impact monitoring - as per Section 9.5.3 of the Contract Specific EM & A Manual.

This document also serves a form of Statistical Analysis Plan (SAP) that permits vetting of the statistical methods. The technical details necessarily pre-suppose familiarity with statistical models, in particular Generalized Linear Models (GLMs) or Generalized Additive Models (GAMs).

ASSESSING THE DISTRIBUTION OF DOLPHINS IN THE SURVEYED AREA

The distribution of dolphins through space and time will be described by statistical models fitted to dolphin observation data. Relevant drivers of dolphin distributions, such as oceanographic features, will also be included in the modelling process. The modelling methods employed will be appropriate for the problem, representing the most recent developments in this area and currently accepted UK standards for species distribution modelling for the purposes of Environmental Impact Assessments. The methods are described briefly here.

Flexible modelling of animal distributions

In many surveyed sites, the way the animals distribute themselves can vary a great deal. For example, some areas (e.g. close to the coast) may require a very flexible modelling surface which can potentially change quickly with local features, while areas with deeper water may exhibit less local variability and require less flexibility.

It is important to target model flexibility to ensure important local features are not missed and 'smoothed-out', such as areas in and around a potentially impacted site, and to ensure the spatial range of any local effects that do exist are not exaggerated. For instance, 'smoothing-out' local features will result in under-reporting of any impacts in and around the site(s) of interest and also result in extending the range of the impact into areas which are, in truth, unaffected.

Targeting model flexibility also helps ensure that some areas of the surface are not unduly variable, and therefore natural fluctuations in dolphin numbers are not mistaken for genuine changes in the underlying system. This is particularly relevant in impact studies where it is important not to falsely attribute natural variability in animal numbers to an impact effect.

The methods we propose to use for the HZMB data are 'spatially-adaptive' and allow model flexibility to be targeted. Additionally, these methods are developed with impact assessment in mind, to allow a particular focus on special areas of interest e.g. in and around the potentially impacted site. The methods proposed for the analysis reflect the most recent research in this area^{^[8]}. They were recently presented as an invited talk for the UK government and industry 'Sharing Good Practice' event for Scottish Natural Heritage (November, 2011)^{^[9]} and are currently being used to analyse an extensive international data set collected over 30 years as a part of the recent Joint Cetacean Protocol Project (JCP, commissioned by the UK government; JNCC)^{^[10]}. These methods have also been used successfully to model many renewable projects for a variety of large UK and international companies with stakes in renewables (Forewind, Centrica plc, Royal Haskoning), including one of the world's largest offshore wind farms^{^[11]}.

As an example of modelling output, 'difference-maps' illustrating where statistically significant changes/redistribution of animals across a surveyed area can be supplied with reference to an impact site (Figure 1). This allows any differences over time to be geo-referenced, helping the end-user to exercise judgement about whether differences are both significant and related to the speculative impact. Maps of this nature will be extended to the current region (Figure 2).

Inclusion of Covariate data

Any readily available data that might be useful in the prediction of dolphin distributions can and should be considered for selection in these models. The utility of these covariates for predicting dolphin distributions can then be determined during the modelling phase. The covariates might include oceanographic features such as tidal flows and bathymetry, whose temporal & spatial resolution can be variable e.g. daily/ weekly /monthly. The finest available resolutions will be considered as a starting point; however the most sensible resolutions are also determined during the modelling process. For example, some covariate data may need to be coarsened to match the resolution of the dolphin observation data.

Covariates for consideration

The following outlines the available covariates for a priori consideration in the models, with justification. These may not be represented in the final model, subject to model selection results.

Time of day, day-of-year, month, year: These temporal measures will all be included as potential covariates in the model to capture patterns of dolphin distribution not evident in other covariates. For example, month has a tendency to capture seasonal variations in animal distributions; the time of day may capture diurnal patterns beyond tidal patterns.

Tidal state: marine animals typically show patterns of activity associated with tidal cycles. This can be due to changes in water depth, tidal flows, prey movements, etc. Suitable summaries of the tidal state will be considered in the pool of covariates.

Salinity/Turbidity/DO & DO saturation/pH: while these may not be directly influential on dolphins, they may present as proxies for prey distributions or habitats which influence dolphin distributions.

Depth: this frequently serves as a strong predictor of the distributions of marine animal distributions. There are many potential reasons for this e.g. in the case of diving animals this indicates the accessibility of the sea floor. It is also a strong indicator of particular prey species and is often indicative of the proximity of land masses.

Water temperature: as above, this is frequently a strong predictor of marine animal distributions, whether as a direct driver on the animals or via its effects on productivity and prey distributions.

Suspended solids: given the nature of the construction activities, patterns in animal distributions relating to solids in the water column will be of clear interest.

Details of construction activities per day: These are essential, as they will provide the basis of variables that code for potential construction–related impacts. These will be translated into an appropriate small set of categorical variables and potentially durations, depending on the available information.

Data resolution

The maximum native resolution will be used in the first instance where possible to retain the maximum information. The resolutions need not be matched either in time or space and can be post-processed where needed. The resolutions of the covariate data identified above are:

Native resolutions of the data….

The resolution of the data may be coarsened in some cases. This cannot be determined with certainty prior to analysis of the data, but the following scenarios are common based on previous work in this area:

Dolphin observation data is frequently very spatially and temporally sparse i.e. dominated by zero observations. The data may be condensed to quarter-hour to hour temporal resolution and some 100s of metre spatial resolution.

All subsequent covariate data will be formatted to match the dolphin observation data. For example, near-continuous time covariates, such as tide height, will be condensed to match the temporal resolution of the observations data.

Model structure/equations

To avoid unnecessary assumptions, the models proposed are:

adaptive in their systematic form,
have a range of potential error distributions,
can accommodate various types of non-independence in the errors.

The modelling process will involve the selection of model terms, optimisation of covariate-response functions and selection of models for stochastic components.

Hence, the specific model equation cannot be known with certainty prior to analysis. This is common to any modelling process that involves optimisation of complexity e.g. regression with model selection, GAMs with optimised smoothing terms. However the broad model structure can be considered, being a type of GAM fitted with GEEs.

Generalized Additive Models (GAMs)

Broadly we seek to estimate f, where the response is in some way functionally related to a set of covariates :

The actual observed response will consist of some inexplicable noise, modelled by some error distribution. The approach is similar to GLMs, so the mean response is modelled on the scale of a link-function , with errors being governed by probability density functions of the exponential family e.g. Gaussian, Gamma, Poisson or quasi- variants. In the current context, may be the expected numbers or densities of dolphins, with the being relevant covariates including spatial coordinates.

GLMs approximate the link-scale f with a simple linear form, , where is the matrix of measured covariates and the associated parameters to be estimated. GAMs differ by seeking to approximate f as a linear combination of smooths on the link-scale, whose individual forms are estimated from the data. As an indication:

Where we can have a range of marginal/single-covariate smooth terms , and multidimensional smooth terms , for groups of covariates , as required. The smooth terms may be very smooth, i.e. linear terms, meaning that GLMs are a special case. Categorical variables can be included as per ordinary GLMs, not included as smooths per se, but via dummy variables.

The fitting of this model operates under assumed error distributions, such as found in GLMs. The smooths themselves can be created in a variety of ways, with splines being commonplace. The complexity of the smooth terms is either set in advance or estimated. The various smooth terms to be included in the model will be subject to selection methods i.e. parsimonious models are selected with insignificant terms dropped.

GAMs are a well-established flexible modelling tool with extensive theoretical treatment – for example refer to Hastie & Tibshirani (1989) or Wood (2006).

Accounting for correlated data/errors

The models we propose to use for this analysis are designed for data collected across space and time, which is true of the type of survey data considered here. This methodology allows for reliable geo-referenced confidence intervals across the fitted surface - which can be interpreted as best and worst case scenarios for any changes over time. Specifically, the modelling framework involves Generalized Estimating Equations (GEEs^{^[12]}) to account for the spatio-temporal autocorrelation.

Most basic statistical modelling methods operate under the simple assumption of independence of errors. However many data collection scenarios will impose some correlation in the errors for these models, due to sampling in a patterned way through space and/or repeated measures on areas through time. The usual consequence of ignoring this correlation when fitting models, is that the variance estimates will be incorrect (typically too small as positive correlation is most common). This has a multitude of inferential consequences – in particular, covariates will be incorrectly determined as significance/insignificant. Generally the complexity of the models is incorrectly determined.

There are a variety of modelling methods to account for this, whereby the correlation structure of the errors is also specified and estimated. We favour here the use of GEEs, which produce empirical adjustments to the variance estimates and are robust to initial misspecification of the error correlation structure.

Any analysis applied to this type of data that does not account for correlated errors will provide spurious results. Most basic statistical tools operate under this assumption of the independence of errors, which while providing simple analyses, produces indefensible results.

Model assumptions

The model and fitting method indicated in section 0 are a form of Generalized Additive Model, fitted using Generalized Estimating Equations. These have been chosen to avoid unreasonable or unnecessary assumptions.

Assumptions of the systematic component

As indicated in section 0, the systematic component of the model consists of adaptive spline models for continuous covariates, or factors for categorical covariates. For the spatial component, the function used is a geodesic smoother with adaptive selection of knots and basis functions, as described in Scott-Hayward et al (2013). For the modelling of individual continuous covariates these are splines with adaptive knot selection, as described in Walker et al (2011). The range of functions that can be approximated by this scheme is very rich.

Common to other regression methods, it is assumed that important predictors are included in the model. However this is mitigated against, in part, by the inclusion of a spatial smoothing term. Un-measured covariates may be represented in by proxy, if they have spatial structure themselves. Failing to include important drivers in the model may produce models with poor predictive power.

The assumptions regarding the systematic component are minimal:

Due to the adaptive model specification, relatively little is assumed for the functional form of the relationship between the response and continuous covariates, beyond continuity and smoothness.

The level of complexity is correctly determined i.e. under- and over- fitting are to be avoided, which covariates are important and to be included, the level of smoothing applied. These are determined by various information criteria (AIC, QIC, BIC) and/or cross-validation.

It is assumed that the covariates are not excessively correlated with one another. These are checked as a matter of course, so high collinearity will be mitigated against.

Assumptions of the stochastic component

Common to GLMs, the models here permit a range of error distributions. The exact distribution assumed is determined during analysis, however the nature of the response suggests quasi-Poisson. The parameter estimates are therefore from maximising either likelihoods or quasi-likelihoods.

Taken collectively, the assumptions implied from our treatment of errors are minimal:

We do not assume independence of errors, which is a priori unlikely. The fitting of GEEs is known to be robust to the choice of working correlation structure, so while this is assumed to be correctly specified, it is not largely influential on the results.

We do assume that the choice of error distribution is correct, however there is a wide range of distributions to choose between and the choice is scrutinised.

Determination of impact

Definition of impact

Impact here is restricted to statistically significant differences in dolphin abundance or spatial distribution that are attributable to construction activities. Other definitions based on the magnitude of effect can be employed if such criteria exist. The models described here provide estimates of the size of effect, along with confidence intervals – these permit alternative impact assessments.

Impact will be sought via statistical models by:

Statistically significant changes in the absolute or relative abundance of dolphins within the study area, when comparing pre- to post-development data. This is after account has been taken of important covariates.

Statistically significant changes in the distribution of animals over the study area when comparing pre- to post-development data - again after the effect of important covariates has been accounted for. The model contains a spatial smoothing component, providing estimated surfaces and confidence intervals for the dolphin distributions through time, with other covariates partialled. Changes in dolphin distributions which are strongly associated with the areas of highest development activity will provide evidence of impact.

It is difficult to determine with certainty that a statistically significant change is due only to development activities. Any change in conditions coincident in time with the development, that is not included as a covariate in the model, serves as an alternative explanation. For example, a disease that is coincident with the development phase.

However, significant distributional changes that are coincident both temporally and spatially with the development will provide a compelling case. For example, if there is a significant shift in distribution around the development site at the same time as development, after other important covariates are accounted for.

Impact hypotheses

The nature of the questions, data and the statistical methods necessitated by these, precludes simple a priori hypotheses as might be presented for simple, inappropriate analysis. However some formalisation can be attempted.

The null hypothesis is one of no impact on the dolphin distribution, or abundance, attributable to development activities.

The alternative hypothesis is one of some impact on the dolphin distribution, or abundances, due to development activities.

The impact here is determined by statistically significant parameters, or changes parameters, that compare the pre- to post-development states. The models are subject to selection in light of the data, so their specific form cannot be known a priori.

Nonetheless, a simple variable can be included in the model, along with the model-selected covariates, that codes for different development states e.g. before/during/after. Significant changes in the general dolphin abundance will be detectible from tests and confidence intervals on these parameters.

Relatedly, the spatial smoothing method has a set of parameters that relate to locally restricted abundances. Tests and confidence intervals for these parameters over the development phases will also identify spatially explicit shifts in dolphin abundances.

More generally, parametric bootstraps will be employed to provide empirical confidence intervals for abundances and their spatial distribution, for the distinct development phases. Any shifts in abundance and distribution over these phases, that are not explicable by natural variation and other covariates, will suggest impacts.

AN ASSESSMENT OF THE POWER OF THE SURVEY (AND DATA) AT DETECTING CHANGE

As a part of this work, the power of the monitoring regime to detect real change in dolphin numbers can also be quantified, under a range of scenarios agreed with the client. For example, this might consist of determining the power to detect a general decline in animal numbers ranging over 5%-50% and/or a spatially restricted decline about the point source.

The power analysis will be carried out using a simulation based approach which includes relevant complexities in the data to ensure realistic results. Specifically, the simulations for power analysis are bespoke to quantify the power in detecting an 'impact' while addressing the following important features of the survey data:

a) Non-linear relationships.

b) Correlated and/or over dispersed observations.

c) Spatially explicit impact/changes.

Briefly, the simulation-based power analysis works as follows:

A model for the species distribution, including any environmental covariates, is constructed.

This model, that has been developed on historical data, is assumed to hold true for the future monitoring periods, but with some general reduction in relative animal abundance (e.g. attributable to a specific impact).

Simulated data is generated from this process with noise consistent with the historical recordings.

The current modelling process is applied to the multiple sets of simulation data, but additionally estimating an impact effect.

Various sizes/types of impact are simulated, and detection of a statistically significant effect is sought from the models under scenarios of interest, for example under differing intensities of monitoring.

The inherent noise in the system may mean small installation effects might be difficult to detect over small time-periods, but these effects will become more detectable with more data. Large effects should be detectible sooner. The simulation process described allows quantification of the probability of detecting an effect, for various effect sizes and periods of additional monitoring.

Map_ver3

Figure 1: An example output showing the significant decreases (white horizontal lines) and significant increases (solid black crosses) post impact. In this particular case, the scale moves from a deep blue colour (which indicates a decrease in number) to a red colour (which indicates an increase in number) and the animals have moved out along the 10-15m contour in depth, post impact.

Figure 2: the region over which the methods will be applied, ultimately providing a model surface and infrerence similat to Figure 1. Model predictions can be provided on a range of resolutions.

References

Hardin, J. W. & Hilbe, J. (2003) Generalized Estimating Equations. Chapman & Hall/Crc Press.

Hastie, T. & Tibshirani, R. (1989) Generalized Additive Models. Volume 43 of Monographs on Statistics and Applied Probability. Chapman & Hall, 1990. ISBN 0412343908. 335pp.

Paxton, C., Mackenzie, M. L., Burt, L., Rexstad, E. & Thomas, L. (2011) Phase II Data Analysis Of Joint Cetacean Protocol Data Resource, Http://Jncc.Defra.Gov.Uk/Pdf/Jcp_Phase_Ii_Report.Pdf

Petersen, I. K., MacKenzie, M. L., Rexstad, E., Wisz, M. S. and Fox, A. D. (2011) Comparing Pre- And Post-Construction Distributions Of Long-Tailed Ducks Clangula Hyemalis In And Around The Nysted Offshore Wind Farm, Denmark : A Quasi-Designed Experiment Accounting For Imperfect Detection, Local Surface Features And Autocorrelation. Http://Hdl.Handle.Net/10023/2008

Scott-Hayward, L. A. S., Mackenzie, M. L. ,Donovan, C. R., Walker, C. & Ashe, E. (2013), Complex Region Spatial Smoother (Cress). Journal of Computational and Graphical Statistics. In press

Wood, S. (2006) Generalized Additive Models: An Introduction with R. Volume 66 of Texts in Statistical Science. Chapman and Hall/CRC Texts in Statistical Science Series. ISBN1584884746. 391 pp.

Annex VI

March 2012– November 2013

(and Baseline September – November 2011)

Photo Identification Information

Table 1. Sightings of Individually Identified Chinese White Dolphin (Sousa chinensis) between March 2012 – November 2013 and during the Baseline Survey (September - November 2011)

Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 114		2013-10-24	827	NWL
HZMB 113		2013-10-24	827	NWL
HZMB 112		2013-10-15	815	NWL
HZMB111		2013-10-15	815	NWL
HZMB 110		2013-10-15	812	NWL
HZMB 108		2013-08-30	780	NEL
HZMB 107		2013-08-21	770	NWL
HZMB 106		2013-08-21	769	NWL
HZMB 105		2013-07-08	711	NWL
HZMB 104		2013-07-08	711	NWL
HZMB 103		2013-07-08	711	NWL
HZMB 102		2013-07-08	706	NWL
HZMB 101		2013-07-08	706	NWL
HZMB 100		2013-07-08	706	NWL
HZMB 099		2013-06-13 2013-06-13	681 680	NWL NWL
HZMB 098	NL104	2013-11-02 2013-11-02 2013-10-24 2013-07-08 2013-05-24	849 845 831 711 659	NWL NWL NWL NWL NWL
HZMB 097		2013-05-09	647	NWL
HZMB 096		2013-04-01	621	NWL
HZMB 095		2013-08-30 2013-06-25 2013-06-13 2013-04-01	780 697 682 621	NEL NWL NWL NWL
Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 094		2013-06-26 2013-06-25 2013-03-18	703 698 601	NWL NWL NWL
HZMB 093		2013-05-24 2013-02-21	657 587	NWL NWL
HZMB 092		2013-02-21 2013-02-15	589 581	NWL NWL
HZMB 091		2013-02-15	579	NWL
HZMB 090		2013-06-25 2013-06-13 2013-02-15	697 682 579	NWL NWL NWL
HZMB 089		2013-02-15	579	NWL
HZMB 088		2013-02-15	579	NWL
HZMB 087		2013-02-15	579	NWL
HZMB 086	NL242	2013-05-09 2013-02-15 2011-10-10	642 579 Baseline	NWL NWL NWL
HZMB 085		2013-06-26 2013-02-15	703 579	NWL NWL
HZMB 084		2013-02-14	575	NWL
HZMB 083	NL136	2013-03-28 2013-02-15 2013-01-28 2012-01-28	607 579 568 564	NWL NWL NWL NWL
HZMB 082		2013-02-21 2013-02-15 2013-01-28	587 579 563	NWL NWL NWL
HZMB 081		2013-01-28 2013-01-28	559 557	NWL NWL
Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 080		2013-01-28	556	NWL
HZMB 079		2013-01-28	556	NWL
HZMB 078		2013-02-15 2013-01-08	579 552	NWL NWL
HZMB 077		2013-07-08 2012-12-11	706 541	NWL NWL
HZMB 076		2013-07-08 2012-12-11	706 541	NWL NWL
HZMB 075		2012-12-06	525	NEL
HZMB 074		2013-05-09 2013-04-01 2013-04-01 2013-02-21 2012-12-10 2012-12-06	647 623 621 594 529 525	NWL NWL NWL NEL NEL NEL
HZMB 073		2013-05-09 2013-04-01 2013-04-01 2013-02-21 2012-12-10 2012-12-06	647 623 621 594 529 525	NWL NWL NWL NEL NEL NEL
HZMB 072		2012-10-24	476	NWL
HZMB 071		2012-10-24 2012-10-12	475 466	NWL NWL
HZMB 070		2012-10-24	476	NWL

Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 069		2013-08-21 2013-07-08 2012-10-24	774 711 476	NWL NWL NWL
HZMB 068		2013-11-01 2012-10-24	839 476	NWL NWL
HZMB 067		2012-10-24	475	NWL
HZMB 066	NL93	2013-01-28 2012-12-11 2012-10-24 2012-10-12	559 537 475 466	NWL NWL NWL NWL
HZMB 064		2013-05-09 2013-01-28 2012-10-24 2012-10-12	647 561 475 466	NWL NWL NWL NWL
HZMB 063		2013-05-09 2012-10-12	647 466	NWL NWL
HZMB 060		2012-09-18	447	NWL
HZMB 059		2013-02-21 2012-09-18	591 445	NWL NWL
HZMB 057		2012-09-18	440	NWL
HZMB 056		2012-09-18 2012-09-05	442 433	NWL NEL
HZMB 055		2012-09-04	425	NWL

Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 055		2012-09-04	425	NWL
HZMB 054	CH34	2013-11-07 2013-11-02 2013-10-24 2013-08-30 2013-07-08 2012-09-18 2012-09-05 2011-11-07 2011-11-05 2011-11-02 2011-11-01 2011-11-01 2011-10-28 2011-10-06	854 845 831 780 711 448 432 Baseline Baseline Baseline Baseline Baseline Baseline Baseline	NWL NWL NWL NEL NWL NWL NEL NWL NWL NWL NEL NEL NWL NWL
HZMB 053		2012-09-04	425	NWL
HZMB 052		2012-09-04	423	NWL
HZMB 051	NL213	2013-05-09 2013-04-01 2013-02-15 2013-02-15 2013-01-28 2013-01-28 2012-09-04	644 622 582 581 559 556 422	NWL NWL NWL NWL NWL NWL NWL
HZMB 050		2013-02-15 2012-09-04	579 421	NWL NWL
HZMB 049		2012-09-03	419	NWL
HZMB 048		2012-09-03	419	NWL
HZMB 047		2012-09-03	412	NWL
Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 046		2012-09-03	412	NWL
HZMB 045		2013-06-13 2013-02-15 2012-11-01	682 579 495	NWL NWL NWL
HZMB 044	NL98	2013-11-02 2013-11-01 2013-10-15 2013-05-09 2013-05-09 2013-04-01 2013-04-01 2013-02-15 2012-11-01	845 842 819 648 647 623 621 579 495	NWL NWL NWL NWL NWL NWL NWL NWL NWL
HZMB 043		2012-09-03	407	NWL
HZMB 042	NL260	2012-11-01 2011-11-07	495 Baseline	NWL NWL
HZMB 041	NL24	2013-11-02 2013-05-09 2013-05-09 2013-04-01 2013-04-01 2013-02-15 2012-11-01 2011-11-06 2011-11-05 2011-11-05 2011-10-10	845 648 647 623 621 579 495 Baseline Baseline Baseline Baseline	NWL NWL NWL NWL NWL NWL NWL NEL NWL NWL NWL
HZMB 040		2013-10-15 2013-07-08 2013-07-08 2013-02-21 2012-11-01	821 714 711 589 493	NWL NWL NWL NWL NWL
HZMB 038		2012-11-01	490	NWL
HZMB 037		2012-11-01	490	NWL
HZMB 036		2012-09-03 2012-11-01	407 490	NWL NWL
Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 035		2013-02-15 2012-11-01	579 490	NWL NWL
HZMB 034		2012-11-01	493	NWL
HZMB 028		2013-04-01 2012-08-06	625 373	NWL NWL
HZMB 027		2013-02-15 2013-01-28 2013-01-28 2012-06-14	579 568 564 299	NWL NWL NWL NWL
HZMB 026		2013-06-25 2013-05-09 2013-01-28 2012-06-13	697 642 561 295	NWL NWL NWL NEL
HZMB 025		2013-02-22 2013-02-21 2012-12-06 2012-10-11 2012-06-13	596 591 525 457 295	NEL NWL NEL NWL NEL
HZMB 024		2013-03-18 2012-06-13	601 295	NWL NEL
HZMB 023		2013-07-08 2013-07-08 2013-04-01 2013-02-21 2013-02-15 2012-07-10	715 711 619 589 579 330	NWL NWL NWL NWL NWL NWL
HZMB 022		2013-10-24 2013-07-08 2013-07-08 2013-04-01 2013-02-21 2013-02-15 2012-07-10	827 715 711 619 589 579 330	NWL NWL NWL NWL NWL NWL NWL
HZMB 021	NL37	2012-07-10 2011-09-16	330 Baseline	NWL NWL
Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 020		2012-07-10	330	NWL
HZMB 019		2012-07-10	330	NWL
HZMB 018		2013-05-09 2013-02-21 2012-12-10 2012-07-10	647 594 529 330	NWL NEL NEL NWL
HZMB 017		2012-07-10	330	NWL
HZMB 016		2013-07-08 2012-12-11 2012-09-18 2012-09-04 2012-07-10	706 539 446 421 330	NWL NWL NWL NWL NWL
HZMB 015		2012-07-10	330	NEL
HZMB 014	NL176	2012-08-06 2012-06-13 2011-11-06 2011-11-01 2011-11-01	373 295 Baseline Baseline Baseline	NWL NEL NEL NEL NEL
HZMB 013		2012-05-28	281	NWL
HZMB 012		2012-05-28	281	NWL

Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 011	EL01	2013-02-22 2013-02-21 2013-02-14 2012-11-06 2012-09-19 2012-03-31 2011-11-02 2011-11-01	597 592 572 517 452 261 Baseline Baseline	NEL NEL NEL NEL NWL NEL NWL NEL
HZMB 009		2012-05-28	281	NWL
HZMB 008		2012-05-28	281	NWL
HZMB 007	NL246	2012-12-10	529	NEL
HZMB 006		2013-02-21 2012-12-11 2012-11-01 2012-03-29	594 539 495 250	NEL NWL NWL NWL
HZMB 005		2013-11-09 2013-11-07 2013-10-15 2012-12-10 2012-08-06 2012-05-28	860 858 813 532 374 287	NWL NWL NWL NWL NWL NWL
HZMB 004		2012-09-04 2012-03-31	421 262	NWL NWL
HZMB 003	NL179	2014-10-15 2013-06-25 2012-12-10 2012-03-31 2011-11-06 2011-09-16	812 697 529 261 Baseline Baseline	NWL NWL NEL NWL NEL NWL
Identification Number	Baseline Identification Number	Date (YYYY-MM-DD)	Sighting Number	Area Sighted
HZMB 002	WL111	2013-11-01 2013-10-15 2013-09-24 2013-02-14 2012-12-11 2012-12-11 2012-10-12 2012-10-24 2012-05-28 2012-03-29	839 819 798 573 536 535 466 475 281 250	NWL NWL NWL NWL NWL NWL NWL NWL NWL NWL
HZMB 001	WL46	2013-08-21 2013-06-13 2013-04-01 2013-02-14 2012-03-29	771 681 617 573 250	NWL NWL NWL NWL NWL
	CH98	2011-11-02	Baseline	NWL
	NL11	2011-11-02 2011-11-07	Baseline Baseline	NWL NWL
	NL12	2011-11-02	Baseline	NWL
	NL33	2011-09-23 2011-11-01 2011-11-05 2011-11-07	Baseline Baseline Baseline Baseline	NWL NEL NWL NWL
	NL37	2011-09-16	Baseline	NWL
	NL46	2011-10-28	Baseline	NWL

[2] Logger is purpose built software which automatically collects and stores GPS data and contains a user configurable interface for the manual entry of the data required for line transect and other cetacean research studies (Gillespie et al 2010).

[3] Group size is defined as an aggregation of dolphins within 100m of each other involved in similar behaviour (Connor et al 1998).

[4] During sightings a minimum, maximum and best estimate of group size is noted; the range stated represents the minimum and maximum numbers estimated)

[6] The provision of data from multiple government agencies and the private sector is acknowledged

[7] Please note: the monthly report for October 2013 erroneously reported no calves were seen during this month. There were four sightings of calves during this month.

[8] Scott Hayward, L. A. S. , Mackenzie ,M. L. , Donovan, C R , Walker, C. & Ashe, E. (2013), Complex Region Spatial Smoother (Cress). Journal of Computational and Graphical Statistics. In press

[10] Paxton, C., Mackenzie, M. L., Burt, L., Rexstad, E. & Thomas, L. (2011) Phase II Data Analysis Of Joint Cetacean Protocol Data Resource, Http://Jncc.Defra.Gov.Uk/Pdf/Jcp_Phase_II_Report.Pdf

[11] Comparing Pre- and Post-Construction Distributions of Long-Tailed Ducks Clangula Hyemalis in and Around the Nysted Offshore Wind Farm, Denmark : A Quasi-Designed Experiment Accounting for Imperfect Detection, Local Surface Features and Autocorrelation. Http://Hdl.Handle.Net/10023/2008

[12] Hardin, J. W. and Hilbe, J., Generalized Estimating Equations. 2003. Chapman & Hall/CRC Press.

	Hong Kong-Zhuhai-Macao Bridge Hong Kong Boundary Crossing Facilities-Reclamation Works

September – November 2013 Quarterly Report	Dolphin Impact Monitoring