Geography Internal Assessment

How does water quality in the Silver Mine River

change with distance downstream?

A river field work study investigating downstream changes in river quality on the
Silver Mine River

Session: May 2021

Candidate code:
Word count: 2453
Contents

Introduction 3

Methodology 7

Primary Research Hypothesis 15

Secondary Research Hypothesis 27

Conclusion and Evaluation 30

Bibliography 32

Appendices 33

List of Abbreviations

SMR Silver Mine River

WQS Water Quality Score

SRCC Spearman’s Rank Correlation Coefficient

PMCC Product-Moment Correlation Coefficient

2
Introduction
This study aims to answer the research question:
How does water quality in the Silver Mine River (SMR) change with distance downstream?
Geography of Area of Study

Figure 1.1 Location Map of the SMR

3
The SMR is located Southeast of Lantau Island, Hong Kong. It flows through Mui Wo, a rural town with
a population of 6000 located on the Eastern coast of Lantau Island, Hong Kong1.

Relation to Syllabus
This study relates to Option A Freshwater, specifically Section 3 (Water Quality) and Section 4 (Water
Management)

Primary Research Hypothesis

!! - Primary Research Hypothesis - Water quality in the SMR will deteriorate with increasing distance
downstream.

!" - Null Hypothesis - Water quality in the SMR is independent of distance downstream.

Figure 1.2 The Bradshaw Model states that in theory,

load quantity and capacity is expected to increase with
distance downstream, allowing for more pollution to
accumulate. Moreover, increasing distance downstream
would expose the SMR to more point and nonpoint
sources of human pollution. Thus, my prediction is that
pollution will accumulate and build up from river source
to mouth, resulting in deteriorating water quality.

Figure 1.2 The Bradshaw Model2

1
"Fact Sheet for Tertiary Planning Units 961 - 2011 Hong Kong ...." https://www.census2011.gov.hk/en/district-
profiles/tpu/tpu961.html. Accessed: 5 Jan. 2021.
2
"Applied Geographical skills 2010 - Cool Geography." http://www.coolgeography.co.uk/A-
level/AQA/Year%2013/Fieldwork%20investigation/Fieldwork_Investigation_2012.htm. Accessed: 22 June. 2020.

4
Figure 1.3 Area of Study Map, showing point and nonpoint sources of pollution

5
Existing governmental research into SMR suggests that livestock farms and unsewered villages are major
sources of pollution3, indicating poor water quality. However secondary sources also suggested that tight
governmental regulations, including the Livestock Waste Control Scheme in Mui Wo4 might have
improved river water quality. In terms of geographical factors, secondary sources have stated that water
quality in SMR is generally poorer in wet seasons5.

Secondary Research Hypothesis

!! - Secondary Research Hypothesis - As dissolved oxygen levels increase, species WQS increases

!" - Null Hypothesis - There will be no significant relationship between dissolved oxygen levels and
species WQS in the SMR

Additionally, I hypothesized that there will be a relationship between dissolved oxygen levels and species
Water Quality Score (WQS). Dissolved oxygen levels are expected to decrease downstream due to
reduced riparian shading, which in turn rises the temperature of the river leading to decreased oxygen
solubility6. Moreover, biological oxygen demand is expected to increase downstream, reducing dissolved
oxygen levels. Furthermore, species WQS is expected to decrease downstream due to an accumulation of
pollution and decreased habitat availability for macroinvertebrates. All these reasons suggest that
dissolved oxygen levels might be correlated to species WQS.

Factors that might influence water quality include geographical factors such as river length, rainfall, and
time of the year, and human influences such as surrounding agricultural activity.

3
"River Water Quality in Hong Kong - EPD." https://www.epd.gov.hk/epd/misc/river_quality/1986-
2005/eng/6_lantau_content.htm. Accessed: 22 June. 2020.
4
"Success Stories | Environmental Protection Department - Epd." 1 Apr. 2020,
https://www.epd.gov.hk/epd/english/environmentinhk/water/beach_quality/story.html. Accessed: 22 June. 2020.
5
"Beach Pollution Source - HKU." http://www.waterman.hku.hk/beach/member/beachpollsnew.aspx?code=SIL.
Accessed: 22 June. 2020.
6
"Dissolved Oxygen - Environmental Measurement Systems." https://www.fondriest.com/environmental-
measurements/parameters/water-quality/dissolved-oxygen/. Accessed: 6 Feb. 2021.

6
Methodology
Line systematic sampling will be used to enable comparison between locations of regular intervals, which
will allow us to measure downstream changes in water quality. However, due to safety and legal reasons
with site access, locations are not equidistant. The data will be collected during a fieldwork excursion to
the SMR on Friday, 9th October 2020.

Figure 2.1 Sampling Location Map

7
Primary Data Collection Methods

Water Quality
Water quality was measured by looking for indicator species, chemical data, riparian index, and levels of
dissolved oxygen. This is a Weighted Composite Index based on multiple indices.

Measuring Species WQS

Certain invertebrates are associated with high water quality. As invertebrates are bound to their
environment, by looking for indicator species, this would indicate water quality from a few days ago. No
creatures were harmed in the process of data collection.

Figure 2.2 Kick Test

8
Figure 2.3 Measuring Species WQS

Figure 2.4 Species Identification Key

9
Riparian Index Assessments
Riparian index is a measure of the ecological condition of riparian vegetation, which is crucial in filtering sediments, retaining pollutants,
and providing shelter for terrestrial and aquatic species. A lower riparian index score would reflect signs of human activity and pollution,
indicating low levels of water quality.

Figure 2.5 Riparian Index Assessment (Site 11)

10
Figure 2.6 Sample Riparian Index Calculation Sheet

11
Measuring Levels of Dissolved Oxygen
As previously mentioned, dissolved oxygen is a major indicator of habitat availability for
macroinvertebrates. Field photographs and observations will also be collected to identify factors
influencing levels of dissolved oxygen, such as eutrophication.

Figure 2.7 Dissolved Oxygen Meter

12
Collecting Chemical Data
Water samples were collected and tested for nitrates, phosphates, and dissolved oxygen. A sudden
anomaly of such chemicals would indicate sources of pollution such as illegal dumping of fertilizers.
Excessive nitrates would stimulate algae growth, which would indirectly cause oxygen deficiency and
reduced biodiversity7. Similarly, excessive phosphate leads to eutrophication, which reduces levels of
dissolved oxygen8.

Figure 2.8 Results of Chemical Tests (phosphate !"4 )

7
"Nitrates in rivers — European Environment Agency." 1 Mar. 2015, https://www.eea.europa.eu/data-and-
maps/explore-interactive-maps/nitrate-in-rivers. Accessed: 9 Nov. 2020.
8
"Phosphorus and Water - USGS." https://www.usgs.gov/special-topic/water-science-
school/science/phosphorus-and-water. Accessed: 9 Nov. 2020.

13
Figure 2.9 Testing for Presence of Nitrate (#"3 )

Secondary Data
To investigate factors which may influence downstream changes in river water quality, this will require
secondary data sources. Maps and satellite photographs throughout this investigation (unless otherwise
stated) are created using ArcGIS9 software by Esri.

9
"About ArcGIS | Mapping & Analytics Platform - Esri." https://www.esri.com/en-us/arcgis/about-arcgis/overview.
Accessed: 3 Jan. 2021.

14
Primary Research Hypothesis - Water quality with Distance Downstream

$! - Primary Research Hypothesis - Water quality in the SMR will deteriorate with increasing distance
downstream.

$" - Null Hypothesis - Water quality in the SMR is independent of distance downstream.

Weighted Composite Index

To examine the water quality of SMR, I designed a Weighted Composite Index based on multiple indices
in acknowledgment that multiple variables interact with each other. Looking at the data (Appendix 1),
there were no downstream changes in the levels of nitrates, hence I omitted it from my analysis. It is
important to note that such data suggests a lack of point and nonpoint sources of pollution, such as due to
illegal dumping of fertilizers.

Raw Data
To determine how to compose my index, I plotted a raw data graph (Figure 3.1). I noticed that the species
WQS fluctuated, suggesting that species WQS might be sensitive to changes in water quality. Thus, I
decided to give it more importance in the index. Moreover, Figure 3.1 shows that levels of phosphate in
SMR increases with distance downstream (a positive correlation). As we know from earlier research that
higher levels of phosphate indicate lower levels of water quality, it is important to process the data so that
phosphate levels and water quality are positively correlated (a higher score reflects a higher water
quality).

15
Figure 3.1 Raw data graph showing changes with distance downstream for different variable

16
Unweighted Composite Index
In order to create a weighted composite index for, I must first normalize the data. This is because variables
measured at different scales do not contribute equally to an index, which creates bias. Min-max
normalization was used, where I rescaled the data range to [0, 1].

Figure 3.2 Min-Max Normalization10

I have also processed the data so that phosphate levels are positively correlated with water quality by
simply subtracting the new value from 1. Afterwards, individual components were added together, and
the data was rescaled so that the Unweighted Composite Index has a max score of 100 (Appendix 2).

10
"Handbook on Constructing Composite Indicators - OECD." 21 Mar. 2005,
https://www.oecd.org/sdd/42495745.pdf. Accessed: 1 Jan. 2021.

17
Figure 3.3 Scatter graph showing WQS (unweighted) and its respective variables (scaled)

18
With all variables on the same scale, the degree of change (trendline gradient) and correlation becomes
more apparent. The composite index also reveals the anomaly [2.31, 4.8], which was removed from the
data set to preserve the trendline. Potential reasons for the anomaly will be discussed later in this study.
Additionally, outliers [1.40, 89.8] and [1.82, 64.2] are not considered as anomalies as they do not deviate
massively. It is important to recognize that they skew and reduce the correlation coefficient (as they are
still retained in the data set, as opposed to anomaly [2.31, 4.8]).

Weighting
I noticed that in addition to species WQS, phosphate levels also appear to fluctuate, suggesting that both
variables might be sensitive to changes in water quality. Hence, I decided to give both a heavier weighting
(× 1.25). I have also assigned levels of dissolved oxygen a lighter weighting (× 0.75), This is because
levels of dissolved oxygen are in theory reflected in the species WQS (as explained in the introduction).

By assigning correlated variables a lighter weighting, this prevents double-counting and increases the
sensitivity of my weighted composite index to different factors which might each influence water quality
in a different way11. Whether there exists an actual correlation between levels of dissolved oxygen and
species WQS will be the focus of my Secondary Research Hypothesis.

In a similar way to the unweighted WQS, individual components were then added together after weighting
and the data was rescaled so that the weighted composite index has a max score of 100 (Appendix 3).

11
"Handbook on Constructing Composite Indicators - OECD." 21 Mar. 2005,
https://www.oecd.org/sdd/42495745.pdf. Accessed: 31 Dec. 2020.

19
Figure 3.4 Scatter graph showing WQS (weighted) and its respective variables (scaled)

At first glance, the scatter graph for the weighted composite index appears to be largely similar to the unweighted one. However, the
differences will be better reflected when you consider the ratio between different variables (Figure 3.5)

20
Figure 3.5 Map with magnitude indicating WQS (weighted) and relative ratio of variables

21
Figure 3.5 provides us with a spatial understanding of WQS in the SMR. WQS appears to decrease with
distance downstream, as noted in Figure 3.4. The map suggests that this could be due to the presence of
rural settlements, which serves as major point and nonpoint sources of pollution. It is interesting to note
that as the river progresses downstream, in particular at Sites 10-12, the relative contribution from the
riparian index decreases almost entirely. This indicates a presence of human intervention on the riparian
corridor, suggesting potential nonpoint sources of pollution.

Inferential Statistics
Initial descriptive techniques have revealed a general negative trend between WQS and distance
downstream, supporting my primary research hypothesis. As an alternative to Spearman’s Rank
Correlation Coefficient (SRCC), the Product-Moment Correlation Coefficient (PMCC) was employed to
determine the strength and significance of the correlation. This was because ratio-level variables (WQS
and distance downstream) were measured, sample size was sufficient and that the relationship was
determined to be linear (via scatter graph, Figure 3.3). Moreover, PMCC is more statistically powerful
due to how SRCC reduces the level of measurement from ratio-level to ordinal-level, making it less
sensitive to outliers. As usual, anomaly [2.31, 4.8] was removed from the data set.

PMCC showed a statistically significant result: r(10) = -0.90, p < 0.01 (Appendix 4). Since the probability
of obtaining this result by chance is less than 0.01, the study successfully rejects the null hypothesis: There
is a strong negative correlation between water quality in the SMR and distance downstream.

Possible Reasons
The potential reasons for this strong negative correlation can be explained through the initial secondary
research and satellite photographs (in the Introduction, in particular, Figure 1.3) where numerous point
and nonpoint sources of pollution were identified. Figures 3.6-9 elaborates on this.

22
Figure 3.6 Annotated map explaining results obtained (Sites 1-2)

23
Figure 3.7 Annotated map explaining results obtained (Sites 3-6)

24
Figure 3.8 Annotated map explaining results obtained (Sites 7-9)

25
Figure 3.9 Annotated map explaining results obtained (Sites 10-12

26
Satellite photographs and field observations have indicated the presence of agricultural land use, in
particular smallholder, indigenous, and hobby farms. Fertilizers applied to the field creates nonpoint
sources of pollution. We have observed that this is particularly significant in SMR as smallholders tend
to not follow government guidelines and lack an understanding of fertilizer damage (Figure 3.8).
Moreover, isolated housing and rural settlements along the river would result in an accumulation of
pollution downstream (Figure 3.7), which explains the sudden decrease in WQS after Site 3. Additionally,
feral water buffalos and cows that roam the area would result in the deposition of animal waste,
contributing to the deteriorating water quality (Figure 3.8-9). The anomaly present at Site 11 could
potentially be due to its low discharge (as indicated by the presence of human activity), resulting in an
accumulation of pollutants. The influence of discharge on the concentration of pollutants could be a point
of investigation in the future.

Secondary Research Hypothesis - Dissolved Oxygen and Species WQS

!! - Secondary Research Hypothesis - As dissolved oxygen levels increase, species WQS increases

!" - Null Hypothesis - There will be no significant relationship between dissolved oxygen levels and
species WQS in the SMR

27
Figure 3.10 Dual series graph showing the downstream changes of Species WQS and Dissolved Oxygen Levels

Figure 3.10 shows a strong negative linear trend for both species WQS and dissolved oxygen levels, suggesting a probable relationship
between the two. However, the fluctuating species WQS data and high variance challenges the strength of the correlation. A raw data
graph between dissolved oxygen levels and species WQS is necessary to determine the nature of the relationship.

28
Figure 3.11 Raw data graph showing the relationship between Species WQS and Dissolved Oxygen Levels (mg/L)

Figure 3.11 revealed a moderately strong positive trend between species WQS and dissolved oxygen. However, the correlation appears
to be weak due to the one-to-many relation between variables and high variance.

29
Inferential Statistics
As the relationship could be non-linear (suggested by Figure 3.11), SRCC is used due to its sensitivity to
monotonic relationships (as opposed to linear). SRCC showed a statistically significant result: !! (10) =
0.64, p < 0.05 (Appendix 5). Since the probability of obtaining this result by chance is less than 0.05, the
study successfully rejects the null hypothesis: There is a moderately strong positive correlation between
dissolved oxygen levels and species WQS.

It is important to note that although the positive correlation was of moderate strength, the one-to-many
relationship suggests a lack of causation between the two variables-- there are external factors responsible
for the relationship. Nevertheless, such findings reveal the interrelationships between dissolved oxygen
levels and species WQS, demonstrating the importance of dissolved oxygen for biological processes.

Conclusion and Evaluation

This study aimed to answer the research question: How does water quality in the SMR change with
distance downstream?

Summary of Results
● Water quality in the SMR deteriorates with increasing distance downstream.
● As dissolved oxygen levels increase, species WQS increases.

Primary Hypothesis - Water Quality and Distance Downstream

Water quality was operationalized as a weighted composite index consisting of species WQS, riparian
score, dissolved oxygen levels, and phosphate levels (Figure 3.4-5). In accordance with my hypothesis,
the study found a strong and statistically significant negative correlation between water quality in the
SMR and distance downstream. This is consistent with The Bradshaw Model, which states that in theory,
load quantity and capacity is expected to increase with distance downstream, allowing for more pollution
to accumulate.

Secondary Hypothesis - Species WQS and Dissolved Oxygen levels

The study determined a moderately strong positive correlation between dissolved oxygen levels and
species WQS. This is in accordance with my hypothesis, which explains the interrelationships between
habitat availability for macroinvertebrates and levels of dissolved oxygen.

30
Evaluation

Limitation Impact Method of Improvement

The study does Rainfall prior to data collection would affect the discharge of Any rainfall a few days prior to the data collection should be
not account for the river, which prevents the accumulation of pollutants thus noted. This allows us to account for increased discharge due
geographical influencing the reliability of the data. to rainfall thus increasing the reliability of the study.
factors

Sample size was The lack of equidistant locations greatly diminishes data Ideally, data should be collected on equally spaced-out
small and accuracy as we could only collect limited data from specific locations, allowing us to improve on the accuracy of our
locations were not locations due to access restrictions. The limited data is likely to findings and make better generalizations.
equidistant be unreliable and does not provide us with a holistic
understanding of the entire river.

Weightings for This could heavily influence the results of our findings. For The lack of objectivity in the Weighted Composite Index does
the Weighted instance, if a variable that strongly deviated from the trend was not mean that we should reject its validity. As long as the
Composite Index assigned a heavier weight, this could influence the correlation process was justified and transparent, this could be useful in
were subjective and statistical significance of our findings, potentially (and modeling and approximating the real conditions of the river.
incorrectly) refuting the hypothesis.

Correlation Despite the correlation between species WQS and dissolved Additional data, such as river discharge and chemical data
between variables oxygen levels, the one-to-many relationship suggests a lack of could be collected to identify such external factors, providing
does not imply causation between the two variables-- there are external factors us with a more holistic understanding of the river.
causation responsible for the relationship.

31
Bibliography

Maps and satellite images throughout this investigation (unless otherwise stated) were created using ArcGIS
software by Esri, www.esri.com.

“Fact Sheet for Tertiary Planning Units 961 - 2011 Hong Kong ...." https://www.census2011.gov.hk/en/district-
profiles/tpu/tpu961.html. Accessed: 5 Jan. 2021.
"Applied Geographical skills 2010 - Cool Geography." http://www.coolgeography.co.uk/A-
level/AQA/Year%2013/Fieldwork%20investigation/Fieldwork_Investigation_2012.htm. Accessed: 22
June. 2020.
"River Water Quality in Hong Kong - EPD." https://www.epd.gov.hk/epd/misc/river_quality/1986-
2005/eng/6_lantau_content.htm. Accessed: 22 June. 2020.
"Success Stories | Environmental Protection Department - Epd." 1 Apr. 2020,
https://www.epd.gov.hk/epd/english/environmentinhk/water/beach_quality/story.html. Accessed: 22 June.
2020.
"Beach Pollution Source - HKU." http://www.waterman.hku.hk/beach/member/beachpollsnew.aspx?code=SIL.
Accessed: 22 June. 2020.
"Dissolved Oxygen - Environmental Measurement Systems." https://www.fondriest.com/environmental-
measurements/parameters/water-quality/dissolved-oxygen/. Accessed: 6 Feb. 2021.
"Riparian Quality Index - Asociación Ibérica de Limnología."
http://www.limnetica.net/documentos/limnetica/limnetica-30-2-p-235.pdf. Accessed: 3 Jan. 2021.
"Nitrates in rivers — European Environment Agency." 1 Mar. 2015, https://www.eea.europa.eu/data-and-
maps/explore-interactive-maps/nitrate-in-rivers. Accessed: 9 Nov. 2020.
"Phosphorus and Water - USGS." https://www.usgs.gov/special-topic/water-science-school/science/phosphorus-
and-water. Accessed: 9 Nov. 2020.
"About ArcGIS | Mapping & Analytics Platform - Esri." https://www.esri.com/en-us/arcgis/about-
arcgis/overview. Accessed: 3 Jan. 2021.
"Handbook on Constructing Composite Indicators - OECD." 21 Mar. 2005,
https://www.oecd.org/sdd/42495745.pdf. Accessed: 1 Jan. 2021.
"Pearson Correlation Coefficient - Magoosh Statistics Blog." 9 Apr. 2018, https://magoosh.com/statistics/pearson-
correlation-coefficient/. Accessed 5 Jan. 2021.
"Table of Critical Values: Pearson Correlation - Statistics ...." https://www.statisticssolutions.com/table-of-
critical-values-pearson-correlation/. Accessed: 3 Jan. 2021.
"Spearman's Rank Correlation Coefficient." 1 Sept. 2020, https://geographyfieldwork.com/SpearmansRank.htm.
Accessed: 3 Jan. 2021.
"Significance of Spearman's Rank Correlation Coefficient." 1 Sept. 2020,
https://geographyfieldwork.com/SpearmansRankSignificance.htm. Accessed: 3 Jan. 2021.

32
Appendices

Appendix 1: Raw Data Table

Site Latitude Longitude Distance Species WQS Dissolved !2 Riparian Nitrate "!2 Nitrate "!3 Phosphate
downstream (mg/L) Score #!4 (mg/L)
(km)
1 22.275899 113.988312 1.05 4.0 1.7 51 0 0 0.00
2 22.275705 113.988406 1.07 6.1 1.5 45 0 0 0.50
3 22.273843 113.989824 1.40 6.7 1.7 46 0 0 0.50
4 22.273976 113.989873 1.42 4.0 1.4 40 0 0 0.25
5 22.273965 113.990918 1.58 4.7 1.4 35 0 0 0.25
6 22.273858 113.991208 1.62 4.0 1.5 40 0 0 1.00
7 22.272771 113.991905 1.78 4.0 1.2 41 0 0 1.00
8 22.272458 113.992066 1.82 4.8 1.2 41 0 0 0.00
9 22.272370 113.992521 1.88 5.0 1.2 40 0 0 1.25
10 22.268915 113.992178 2.29 3.1 1.0 29 0 0 0.50
11 22.268721 113.992419 2.31 2.9 1.0 19 0 0 2.00
12 22.268868 113.993495 2.48 2.0 1.0 25 0 0 1.00

33
Appendix 2: Unweighted Composite Index Calculations

34
35
Appendix 3: Weighted Composite Index Calculations

36
Appendix 4: PMCC Calculations (WQS and distance downstream)

PMCC Formula12:

Where:
● n is the sample size
● !! , "! are the x and y component of the #"# data point
● mean(x) and mean(y) are the mean of x components and mean of y components of the
data set respectively.

12
"Pearson Correlation Coefficient - Magoosh Statistics Blog." 9 Apr. 2018,
https://magoosh.com/statistics/pearson-correlation-coefficient/. Accessed: 3 Jan. 2021.

37
●

38
Table of Critical Values (PMCC)13:

Where degrees of freedom = !"#$%& () *+,&- ,! .ℎ% -+#*/% − 2

The critical value for PMCC at * = 0.01 for ! = 12 is 0.708. Since 0.90 > 0.708, the result is statistically
significant.

13
"Table of Critical Values: Pearson Correlation - Statistics ...." https://www.statisticssolutions.com/table-of-
critical-values-pearson-correlation/. Accessed: 3 Jan. 2021.

39
Appendix 5: SRCC Calculations (Species WQS and Dissolved Oxygen Levels)

SRCC Formula14:

Where:
● 3! is the difference in rank of the ,"# data point
● n is the number of pairs in a sample

14
"Spearman's Rank Correlation Coefficient." 1 Sept. 2020,
https://geographyfieldwork.com/SpearmansRank.htm. Accessed: 3 Jan. 2021.

40
Table of Critical Values (SRCC)15:

SRCC showed a statistically significant result: &$ (10) = 0.64, p < 0.05. Since the probability of obtaining
this result by chance is less than 0.05, the result is statistically significant.

15
"Significance of Spearman's Rank Correlation Coefficient." 1 Sept. 2020,
https://geographyfieldwork.com/SpearmansRankSignificance.htm. Accessed: 3 Jan. 2021.

41