[PPT] - Analytical uncertainties and model output Anders Grimvall 1 , PowerPoint Presentation

SLIDE 1

Analytical uncertainties and model output

Anders Grimvall1, Claudia von Brömssen2 and Göran Lindström3 1) Swedish Institute for the Marine Environment 2) Swedish Univerity of Agricultural Sciences 3) Swedish Meteorological and Hydrological Institute

Uppsala 18 May 2015

SLIDE 2

Monitoring and modelling riverine loads of substances

Measured water quality and discharge Process-based catchment model (e.g. HYPE)

High quality monitoring is a prerequisite for high quality modelling!

SLIDE 3

Monitoring and modelling river water quality

Measured water quality and discharge Process-based catchment models (e.g. HYPE)

Can process-based models help to reduce the uncertainty of measured data?

SLIDE 4

Principle premise

Weather- driven temporal variation Process-based catchment model (e.g. HYPE) Constant land- use and point emissions Observed temperature and rainfall The impact of human interventions in the catchment will emerge more clearly If weather-driven fluctuations are removed from observed data

SLIDE 5

Adjustment methods

Simplistic method

Substract calculated weather-driven fluctuations from observed values

Slightly more advanced method

Regress observed values on modelled values and compute residuals. Focus on the feasibility to predict the irregular components (anomalies) of time series

SLIDE 6

HYPE

HYdrological Predictions for the Environment

Uppsala 18 May 2015

The soil is modelled as several layers which may have different thickness for each soil class

SLIDE 7

S-HYPE simulations

S-HYPE is a HYPE setup that covers the whole of

Sweden

Sweden was divided into 37786 sub-basins
The anthropogenic forcing was kept fixed to the

conditions prevailing in 2005

The physical forcing consisted of time series of

meteorological data from 1992 to 2010.

Uppsala 18 May 2015

We used monthly mean model outputs for water discharge and nitrogen and phosphorus concentrations at major river mouths

SLIDE 8

Observational data

Uppsala 18 May 2015

Monthly sampling for water quality. Daily water discharge.

River Sampling site Sampling period Sea area River Sampling site Sampling period Sea area

Norrström Stockholm Centralbron 1996–2010 Baltic Proper Örekilsälven Munkedal 1992–2010 Skagerrak Norrström Stockholm Norrström 1992–2002 Baltic Proper Enningdalsälven

N. Bullaren

1992–2010 Skagerrak Nyköpingsån Spånga 1992–2010 Baltic Proper Forsmarksån Johannisfors 1992–2010 Gulf of Bothnia Motala ström Norrköping 1992–2010 Baltic Proper Dalälven Älvkarleby 1992–2010 Gulf of Bothnia Ljungbyån Ljungbyholm 1992–2010 Baltic Proper Gavleån Gävle 1992–2010 Gulf of Bothnia Alsterån Getebro 1992–2010 Baltic Proper Ljusnan Ljusne Strömmar 1992–2010 Gulf of Bothnia Emån Emsfors 1992–2010 Baltic Proper Delångersån Iggesund 1992–2010 Gulf of Bothnia Botorpström Brunnsö 1992–2010 Baltic Proper Ljungan Skallböleforsen 1992–2010 Gulf of Bothnia Gothemsån Hörsne 1992–2010 Baltic Proper Indalsälven Bergeforsen 1992–2010 Gulf of Bothnia Mörrumsån Mörrum 1992–2010 Baltic Proper Ångermanälven Sollefteå 1992–2010 Gulf of Bothnia Lyckebyån Lyckeby 1992–2010 Baltic Proper Gide älv Gideåbacka 1992–2010 Gulf of Bothnia Skivarpsån Skivarp 1992–2010 Baltic Proper Lögde älv Lögdeå 1992–2010 Gulf of Bothnia Kävlingeån Högsmölla 1996–2010 Baltic Proper Öre älv Torrböle 1992–2010 Gulf of Bothnia Helgeån Hammarsjön 1992–2010 Baltic Proper Ume älv Stornorrfors 1992–2010 Gulf of Bothnia Råån Helsingborg 1992–2010 Öresund Rickleån Rickleån outflow 1992–2010 Gulf of Bothnia Rönneå Klippan 1992–2010 Kattegat Skellefte älv Kvistforsen 1992–2010 Gulf of Bothnia Smedjeån

V. Mellby

1992–2010 Kattegat Pite älv Bölebyn 1992–2010 Gulf of Bothnia Lagan Laholm 1992–2010 Kattegat Alterälven Norrfjärden 1992–2010 Gulf of Bothnia Nissan Halmstad 1992–2010 Kattegat Lule älv Luleå 1992–2010 Gulf of Bothnia Ätran Falkenberg 1992–2010 Kattegat Kalix älv Karlsborg 1992–2010 Gulf of Bothnia Viskan Åsbro 1992–2010 Kattegat Töre älv Bölträsket inflow 1992–2010 Gulf of Bothnia Göta älv Alelyckan 1992–2010 Kattegat Torne älv Mattila 1992–2010 Gulf of Bothnia Göta älv Trollhättan 1992–2010 Kattegat Råne älv Niemisel 1992–2010 Gulf of Bothnia Bäveån Uddevalla 1992–2010 Skagerrak

SLIDE 9

S-HYPE performance: annual means

Uppsala 18 May 2015

SLIDE 10

S-HYPE performance: seasonal components for total N

Uppsala 18 May 2015

SLIDE 11

S-HYPE performance: seasonal components for total P

Uppsala 18 May 2015

SLIDE 12

Prediction of anomalies, i.e the irregular components of a time series

Uppsala 18 May 2015

Which is the best predictor of observed anomalies? S-HYPE anomalies or water discharge anomalies?

SLIDE 13

S-HYPE performance:

monthly anomalies for total N

Uppsala 18 May 2015

Measured anomalies are sometimes slightly better correlated to modelled anomalies than to water discharge anomalies

SLIDE 14

S-HYPE performance:

annual anomalies for total N

Uppsala 18 May 2015

Measured anomalies are sometimes better correlated to modelled anomalies than to water discharge anomalies

SLIDE 15

S-HYPE performance:

annual anomalies for total P

Uppsala 18 May 2015

With few exceptions, measured anomalies are poorly correlated to both modelled anomalies and water discharge anomalies

SLIDE 16

Weather-normalized annual anomalies

f total N

Uppsala 18 May 2015

Red solid line: modelled anomalies Blue dashed line: observed anomalies

SLIDE 17

Mann-Kendall tests for monotone trends:

unadjusted and weather-normalized total N data

Uppsala 18 May 2015

River/Sea area p-value (twosided) Significance code Slope (change/year) p-value (twosided) Significance code Slope (change/year) Mean total N (mg/l) Enningdalsälv 0.1154

4.4

0.2208

2.9

623 Bäveån 0.0543

8.3

0.2483

2.6

939 Örekilsälven 0.0029

-
11.8

0.3449

4.1

1013 Skagerrak 0.0116

7.4

0.1695

3.0

Göta älv, Trollhättan 0.0001

- -
8.2

0.3103

1.8

793 Göta älv, Alelyckan 0.0071

-
8.3

0.8065

0.2

833 Lagan 0.7004 0.5 0.9721 0.1 908 Nissan 0.8611 1.3 0.5520 1.9 1029 Ätran 0.0107

15.4

0.2781

4.3

1183 Viskan 0.0637

8.6

0.3818

4.2

1281 Rönneå 0.0037

-
30.3

0.0744

15.2

2316 Smedjeån 0.0001

- -
89.5

0.0046

-
65.6

4438 Kattegat 0.0023

-
9.7

0.1099

2.6

Kävlingeån 0.0001

- -
168.6

0.0001

- -
168.3

4215 Råån 0.0000

- -
272.2

0.0000

- -
260.6

6714 Öresund 0.0000

- -
208.6

0.0000

- -
204.2

Annual measured anomaly (Total N: mg/l) Annual regression residual (Total N: mg/l)

SLIDE 18

Mann-Kendall tests for monotone trends:

unadjusted and weather-normalized total N data

Uppsala 18 May 2015

River/Sea area p-value (twosided) Significance code Slope (change/year) p-value (twosided) Significance code Slope (change/year) Mean total N (mg/l) Norrström, Centralbron 0.9605

0.6

0.9605 0.2 672 Alsterån 0.3818 3.3 0.4210 3.5 700 Norrström, Norrström 0.0240 + 24.5 0.1857 13.6 726 Botorpström 0.0865 6.0 0.0230 + 7.7 794 Mörrumsån 0.0191 + 6.8 0.0230 + 6.6 835 Motala ström 0.9164

0.6

0.8065 1.0 869 Emån 0.0461 + 7.4 0.1001 5.4 945 Lyckebyån 0.0637 7.0 0.1515 5.2 996 Nyköpingsån 0.3103

4.3

0.4210

2.5

1046 Helgeån 0.4625

7.0

0.6492

2.5

1663 Ljungbyån 0.0107

43.8

0.0230

43.2

1875 Gothemsån 0.1325

39.4

0.0543

45.7

3641 Skivarpsån 0.0037

-
136.1

0.0005

- -
126.2

5421 Baltic Proper 0.8746 0.6 0.9480 0.2 Annual measured anomaly (Total N: mg/l) Annual regression residual (Total N: mg/l)

SLIDE 19

Mann-Kendall tests for monotone trends:

bserved total-P anomalies and residuals after

regressing on S-HYPE anomalies

Uppsala 18 May 2015

River/Sea area p-value (twosided) Significance code Slope (change/year) p-value (twosided) Significance code Slope (change/year) Mean total P (mg/l) Enningdalsälv 0.0540

0.08

0.1724

0.06

11.6 Bäveån 0.2340

0.23

0.1724

0.19

35.9 Örekilsälven 0.0328

0.46

0.0191

0.49

37.4 Skagerrak 0.0128

0.17

0.0194

0.17

Göta älv, Trollhättan 0.0106

0.19

0.0865

0.12

12.4 Göta älv, Alelyckan 0.3627

0.10

0.5520

0.05

18.8 Lagan 0.2208

0.17

0.2483

0.15

20.2 Ätran 0.1614

0.28

0.0543

0.28

21.5 Nissan 0.0041

-
0.36

0.0461

0.23

25.1 Viskan 0.0003

- -
1.07

0.0007

- -
1.07

36.9 Rönneå 0.9721 0.03 0.7004 0.21 51.8 Smedjeån 0.7794 0.20 0.3103 0.29 60.6 Kattegat 0.0172

0.23

0.0736

0.17

Kävlingeån 0.1815

0.69

0.6560

0.26

81.9 Råån 0.0002

- -
3.46

0.0000

- -
3.19

118.3 Öresund 0.0008

- -
2.70

0.0003

- -
2.49

Alsterån 0.4833

0.02

0.3449

0.05

14.6 Emån 0.6740 0.05 0.4625 0.06 18.4 Botorpström 0.0251 + 0.32 0.1724 0.14 19.0 Mörrumsån 0.0037 + + 0.37 0.0018 + + 0.39 24.4 Lyckebyån 0.7526

0.01

0.9721

0.01

25.7 Ljungbyån 0.3103

0.28

0.3103

0.24

25.7 Norrström, Centralbron 0.0197

0.30

0.0200

0.22

28.8 Motala ström 0.0143

0.50

0.0130

0.51

37.1 Helgeån 0.4210 0.27 0.7004 0.19 37.1 Norrström, Norrström 0.5858

0.58

0.3115

0.69

37.9 Nyköpingsån 0.0390

0.48

0.1001

0.39

43.3 Gothemsån 0.0637

0.96

0.0543

1.04

67.1 Skivarpsån 0.0158

1.51

0.0158

1.23

136.4 Baltic Proper 0.2107

0.08

0.1618

0.09

Annual regression residual (Total P: mg/l) Annual measured anomaly (Total P: mg/l)

SLIDE 20

Mann-Kendall tests for monotone trends:

bserved total-P anomalies and residuals after

regressing on S-HYPE anomalies

Uppsala 18 May 2015

River/Sea area p-value (twosided) Significance code Slope (change/year) p-value (twosided) Significance code Slope (change/year) Mean total P (mg/l) Indalsälven 0.0005

- -
0.22

0.0011

-
0.23

6.3 Skellefte älv 0.0000

- -
0.25

0.0000

- -
0.25

7.2 Lule älv 0.0000

- -
0.25

0.0001

- -
0.25

7.4 Ångermanälven 0.0013

-
0.21

0.0009

- -
0.22

8.1 Ume älv 0.0015

-
0.24

0.0015

-
0.23

8.5 Delångersån 0.0071

-
0.19

0.0107

0.16

9.6 Ljungan 0.0029

-
0.41

0.0018

-
0.45

10.6 Ljusnan 0.0275

0.27

0.0230

0.25

11.5 Pite älv 0.7794

0.03

0.8611 0.01 11.9 Råne älv 0.0032

-
0.31

0.0107

0.24

14.3 Gide älv 0.0004

- -
0.41

0.0007

- -
0.42

14.5 Dalälven 0.0045

-
0.26

0.0023

-
0.25

14.9 Kalix älv 0.0045

-
0.33

0.4210

0.08

15.6 Rickleån 0.0063

-
0.23

0.0071

-
0.24

16.0 Öre älv 0.3449

0.17

0.9721

0.01

18.4 Torne älv 0.2936

0.17

0.7529 0.07 18.6 Forsmarksån 0.7000

0.04

0.8065

0.02

19.0 Lögde älv 0.0637

0.45

0.0865

0.41

20.2 Gavleån 0.1235

0.18

0.1154

0.17

26.7 Alterälven 0.2781

0.15

0.2483

0.22

27.5 Töre älv 0.1001

0.39

0.1154

0.35

31.3 Gulf of Bothnia 0.0004

- -
0.24

0.0007

- -
0.22

Annual measured anomaly (Total P: mg/l) Annual regression residual (Total P: mg/l)

SLIDE 21

Level shifts in particulate and soluble P: in low total P rivers

Uppsala 18 May 2015

SLIDE 22

Statistical break-point analysis of particulate P anomalies

Uppsala 18 May 2015

Mixed linear model with quadratic mean function and a level shift between 2001 and 2002

Mean concentration

f total

phosphorus (mg/l) Level shifts in standardized concentration anomalies Level shifts in standardized regression residuals Magnitude Two-sided p-value Magnitude Two-sided p-value 0–10 –0.2143 0.0103 –0.2586 0.0017 10–20 –0.2085 0.0021 –0.2143 0.0013 20–30 –0.1482 0.0413 –0.1764 0.0124 30–40 –0.2021 0.0096 –0.2511 0.0016 > 40 –0.0411 0.6134 –0.0477 0.5319 0–100 –0.1684 0.0039 –0.1908 0.0009

SLIDE 23

Conclusions

Joint analysis of model outputs and measured total N and total P has:

– Helped to attribute trends in total N to measures in agriculture – Helped to reveal a break-point in particulate P – Strengthened the need for better feedback from modelling yo monitoring – Raised the question if flow adjustment of riverine loads can be substituted for procedures involving joint analysis of measured data and model outputss

Uppsala 18 May 2015

SLIDE 24

Reference

Grimvall, A., von Brömssen, C. and Lindström, G.

(2014) Using process-based models to filter out natural variability in observed concentrations of nitrogen and phosphorus in river water. Environmental Monitoring and Assessment. 186:5135-5152.

Uppsala 18 May 2015