Water quality amelioration

Jane Turpie University of Cape Town & Anchor Environmental Consultants

Service & Benefits

1. Reduced impacts on water treatment costs or health costs

2. Reduced impacts on other downstream aquatic ecosystem services

AgricNatural

Downstream ecosystem/ reservoir

Urban

+

+-

--

final service

intermediate service

1. Limiting the area of polluting land uses

2. Assimilating some of the loads generated

the passive service

the active service

1/disservice of degradation

+ = generates excess nutrients- = removes excess nutrients

Nagle Dam Inanda

Dam

Durban case study (eThekwini Municipality)• Included active (nutrient assimilation) and passive (no LULC change) aspects; • Included final (treatment cost) and intermediate (estuary ES) values

These ecosystems save on water treatment costs

These ecosystems save prevent damages to

downstream rivers and estuaries

Understanding of treatment costs avoided

Land cover & use

River WQ

Reservoir WQ

Chemical use, backwashing & sludge removal at WTW

Water treatment costs

?

?

?

?

Two-stage valuation process

• Physical modelling • Change in nutrient load at each point of

interest (m3/year)• Mapping the removal back to service areas

(m3/ha/year) based on model coefficients

• Valuation • Avoided treatment costs at each treatment

plant based on empirical model – treatment costs = f(N, P entering dams)

• Map value to source areas based on removal rate ($/ha/year)

Land cover & use

River WQ

Reservoir WQ

Chemical use, backwashing & sludge removal at WTW

Water treatment costs

Physical model

• Whole catchment area modelled using PC-SWMM

• Production and uptake rates of nutrients for each land cover type based on literature

• Coupled with modelled flows to estimate the concentration and total load of N & P at relevant points

• Calibrated with WQ data from >40 monitoring stations.

eThekwininiMunicipality

uMgenicatchment

Mkomazicatchment

Mdloticatchment

Valuation model

• Treatment cost vs reservoir WQ: model had good fit

• However, physical model predicts inflowing river water quality

• (Too complex to model reservoir response)

• So used slightly weaker model of cost against river water quality

0,00

20,00

40,00

60,00

80,00

100,00

120,00

140,00

160,00

Jul.1

0

Okt

.10

Jän.

11

Apr.1

1

Jul.1

1

Okt

.11

Jän.

12

Apr.1

2

Jul.1

2

Okt

.12

Jän.

13

Apr.1

3

Jul.1

3

Okt

.13

Jän.

14

Apr.1

4

Jul.1

4

Okt

.14

Jän.

15

Apr.1

5

Trea

tmen

t cos

t (R/

ML)

Actual Treatment Cost Predicted Treatment Cost

0,0

20,0

40,0

60,0

80,0

100,0

120,0

140,0

Dez.

10

Feb.

11

Apr.1

1

Jun.

11

Aug.

11

Okt

.11

Dez.

11

Feb.

12

Apr.1

2

Jun.

12

Aug.

12

Okt

.12

Dez.

12

Feb.

13

Apr.1

3

Jun.

13

Aug.

13

Okt

.13

Dez.

13

Feb.

14

Apr.1

4

Jun.

14

Aug.

14

Okt

.14

Wat

er tr

eatm

ent c

ost (

R/M

L)

Actual Treatment Cost Predicted Treatment Cost

Model of cost vs reservoir water quality

Model of cost vs inflowing river water quality

Defining the baseline• Two scenarios were used to estimate service

provided:

1: removal of uptake capacity of natural habitats (a hypothetical construct)

2: replacing natural habitats with dense rural settlement as a likely alternative land use

• These provided lower and upper bound estimates of the magnitude of the service, depending on what is considered as the service

quantifies the active + passive services

quantifies the active service

Results• Big increases in P loads entering

dams • 193%-319% for Inanda Dam, • 193%-968% for Hazelmere and • 200%-509% for Nungwane Dam

• Estimated water treatment cost savings of R1 - R8.7 million per annum

• Large range of uncertainty, mainly from physical modelling and service definition

Empirical valuation based on land cover/use• Few studies directly relate catchment land

cover to water treatment costs• Many anecdotal reports, but little empirical

evidence • Some early studies have been criticized

• Vincent et al. 2015 analysed effect of forest cover on water treatment cost

• Rigorous approach using fixed effects and instrumental variable (IV) models

• 1% increase in virgin forest reduces costs by 1.16%

• ESAforD study used this approach to value water quality amelioration in South Africa, Sweden, Costa Rica, China, Kenya, Ethiopia & Tanzania

Land cover & use

River WQ

Reservoir WQ

Chemical use, backwashing & sludge removal at WTW

Water treatment costs

RSA study • Panel data analysis 2008-2014• Cost and output data from 14 treatment plants (TPs) • Climate and land cover data extracted each catchment

Model resultsVariables Random

EffectsFixedEffects Mundlak

Constant 740,405 -6.339e+07*** -1.273e+07***TP_capacity 15,690*** 15,395***Volume treated 0.156* 0.235*** 0.421***Rainfall 5,243* -521.1 300.9Rain sqr -8.5** 2.075 0.602Settlmnt -309.3 -18,543*** -450.6**Settlmnt sqr 0.00218 0.772*** 0.00335Subsistence 2,011*** 14,901*** 2,626***Orchards 659.5** 53,021*** 1,098***Plantations -350.9*** -8,599*** -622.2***Grasslands 5.882 1,215** 65.02*Natural_forest -2,830*** -39,942** -4,937***Woodland -1,252 -9,200*** 463.2Wetland -936.3** -4,133*** -2,243***Observations 92 92 92Number of TP 14 14 14R-squared 0.757

• Co-efficients translate to cost savings in R/ha/year

• Sensitive to model choice

• Some unexpected signs• Lack of data on

ecosystem condition

Discussion• Improving confidence

• More empirical modelling needed to refine modeling assumptions

• Needs larger datasets, more detailed land use• Empirical modelling may better isolate the active from

the overall service• Two/multi-stage empirical approach may be more

practical, achieve higher confidence

• Defining service • Should only assign active service value to ecosystems, • but should also assign negative (disservice) value to

polluting land uses (when added to production, this internalizes the externality)

Land cover & use

River WQ

Reservoir WQ

Chemical use, backwashing & sludge removal at WTW

Water treatment costs

Discussion

• Double counting?• run-of-river water – provisioning - maybe• treatment costs – sedimentation – no• if extrapolated to areas not serving users – yes – ie. Check demand

• Possibility of scaling up• Yes, but rigorous modelling will take time; must take spatial variation in

demand into account.• Once set up, highly repeatable.

• Thank you!

WQ amelioration by ecosystems• In landscape before reaching river systems, especially riparian zone• Within drainage system, particularly in wetlands