Gestão da Água. Modelação da qualidade ecológica e Modelos simples de gestão. J. Gomes Ferreira.

Transcription

1 Gestão da Água Modelação da qualidade ecológica e Modelos simples de gestão J. Gomes Ferreira Universidade Nova de Lisboa 28 de Novembro, 2017

2 Modelação da qualidade ecológica Abordagem diversificada Temas Principios gerais de modelação ecológica Modelos complexos (research models) Screening models para avaliar a eutrofização Modelos de diagnóstico para aquacultura integrada Síntese Perguntas diferentes, modelos diferentes.

3 Here is the best model Turn your brain on. Turn your computer off.

4 Changes in coastal systems Changes Tides Light Biomass Temperature Global events Local events Gradual change Time Hour Day Season Year Multiyear The noise in the distributions masks the signal of change

5 Model diversity Lab models Incubations for primary production or BOD GIS Spatial models Marine spatial planning, chlorophyll spatial distribution Mathematical models dc/dt = -kc (dynamic, time varying) Physical models Harbour scale models, toys When we talk The about other models, half sees % this... of the world sees this! Other models All models are wrong, but some are useful (George Box)

6 Ecological models are complex even for simple systems... How many state variables would you use in this system?

7 Why do we use models? Measure state, perform experiments, simulate... Our conceptual understanding of ecosystems is often illustrated as a set of boxes (state) linked by arrows (processes) Processes such as primary production or grazing form the links between boxes (state), e.g. phytoplankton biomass, nutrient concentration Experimental approaches can help quantify these processes (e.g. P-I curves) This quantification can be used to mathematically link the boxes, and simulate ecological changes in time and space No question, no model. A model is a tool, not an objective

8 Ecological Modelling A tool Measurement of chlorophyll (satellite), suspended matter (sampling), area of mussel culture (GIS) etc; Modelling of shellfish growth allows the determination of rates such as net phytoplankton removal, nutrient excretion, production, which often cannot be directly measured. State can be measured, processes can be modelled

9 Ecological Modelling - Objectives Description and support Test and validate mental models Support sampling design Describe and hindcast Support data interpretation (e.g. laboratory models) Forecasting Predict general behaviour of ecosystem Test and diagnose potential modifications Distinguish long-term signals from short-term variation Make your model as simple as possible but no simpler

10 Characteristics of models Four defining elements Models should be portable Generality Realism Accuracy Simplicity Loss of realism is expected Loss of accuracy due to smoothing, difficulty in accommodating stochastic events, etc Reduce complexity whenever possible (Occam s razor) Building a model is a trade-off among these four characteristics

11 Dimensions Ecological Modelling Different dimensions, different scales Statistical Zero-dimensional (time only) One-D (rivers, narrow estuaries) Two-D (non-stratified estuaries, coastal areas) Three-D (systems with pronounced horizontal and vertical gradients) Time and space scales Hydrodynamics - Small cells, short timestep and time scale (tidal cycles, spring-neap cycles, localised case studies) Ecology - Larger boxes, longer timestep and time scale (seasonal cycles, annual patterns, multiannual variation) Most people don t solve the problem, they change the problem into something they know how to solve. This does not solve the problem

12 Ecological modelling in coastal environments: At which spatial resolution do we need to represent an ecosystem?

13 General scheme of a simple ecological model River River flow Nutrients Other forcing functions BOX 1 Solar radiation I o Effluents Water temperature BOX n BOX 1 BOX 2 Atmosphere Water column Ocean I z Salt SPM GPP Phyto. Grazing Zooplankton Advection dispersion Ressuspension Deposition DIN NH + 4 NO - 2 NO - 3 Sloppy grazing Nitrification Water column Benthos Even simple ecosystems are complex to model

14 Model elements Ecological Modelling Elements and requirements State variables (nitrate, phytoplankton) Forcing functions (light, temperature) Processes (production, mineralization) Parameters (light extinction cofficient, half-saturation constants, grazing rate) Model requirements Physical framework (box volumes, areas, etc) Boundary conditions (concentration values at model limits) Initial conditions (starting values for model) Operational models (a.k.a. data assimilation) Re-initialised at appropriate time steps Conceptual framework + Physical framework = Model

15 Model Conception Ecological Models Development stages Objectives of the model Components of the model (variables, forcing functions) Scope of the model (time and space) Limitations and closure Model Implementation Problem decomposition, definition of appropriate sub-models Data handling and generation Model building (e.g. visual platform) Running and testing Model Calibration Tuning parameters and functions using field data Model Validation Testing against an independent dataset Re-use if possible, develop if necessary

16 Spreadsheets Excel, Lotus123 etc Data in rows and columns, only formula for active cell is visible Feedback mechanisms are eliminated to avoid circular references Ecological Models Spreadsheets and visual models Visual models InsightMaker, Stella etc Data (including data links) represented using visual elements Feedback is explicitly considered as a major factor in systems analysis Models are all about feedbacks

17 Ecological modelling Hands on learning by doing Visual modelling tools are the best way to acquire basic modelling skills.

18 Ecological modelling Community modelling platform The SIR model, by (Sir) Robert May and Roy Anderson: a classic disease model

19 Ecological models Research models and screening models Characteristics Research models Screening models Resolution High spatial and temporal resolution Low resolution, or integrated in space and/or time Complexity Several-many state variables Focus on a few diagnostic features Difficulty of use Cost Application Substantial, usually have a champion group/groups High due to typical data requirements and complexity Detailed management support, usually supplied as a service Minimal, require few parameters Low cost Broad compliance analysis, scoping work, more a product than a service Target audience Academics, consultancy Managers, public Integrity Hard to verify, hard to modify Easy to do both, more prone to misuse Both types of models play important roles in water quality management

20 What is the question? Ecological models Integrated management Ireland and Northern Ireland must implement the WFD What is the eutrophication status of Lough Foyle? What land-based nutrient control measures are possible? How can shellfish aquaculture contribute to eutrophication abatement? How can regulatory ecosystem services of shellfish be quantified? Relevance: shellfish aquaculture and top-down eutrophication control

21 EcoWin.NET Lough Foyle Model General framework for ecosystem modelling Multi-model frameworks are complex to develop, but they make the link between catchment and coast, and have great potential for management support.

22 SWAT application for the Foyle catchment General objectives 1. Estimate the terrestrial loads to Lough Foyle Loads: water, sediment, dissolved N and P, particulate organic matter (POM) Sources: WWtWs, agriculture Reference year: 2014 Time-step: daily Methodology Direct WWtW loads Measured data 2. Build a modelling tool capable of apportioning sources and assessing scenarios The SWAT eco-hydrological model was applied and calibrated for the Foyle watershed Foyle watershed SWAT model Bann watershed Measured data Integration of catchment modelling is critical to deal with European directives such as the Water Framework Directive.

23 Direct input points (WWtW, Bann) and SWAT simulation area 29 watersheds 330 HRUs Hydrological Response Units 11 WWtW Waste Water treatment Works Measured loads Moville: untreated Faughan Roe Deele Burndennet Finn Mourne Enhanced SWAT application for the Foyle catchment J.P. NUNES

24 SWAT outputs Nutrient loads in time: 2014 Coastal Lough DIN (ton N/dy) DIN Coastal Lough Foyle River Foyle River Jan/2014 Feb/2014 Mar/2014 Apr/2014 May/2014 Jun/2014 Jul/2014 Aug/2014 Sep/2014 Oct/2014 Nov/2014 Dec/2014 DRP (ton P/dy) Jan/2014 Feb/2014 Mar/2014 Apr/2014 May/2014 Jun/2014 Jul/2014 Aug/2014 Sep/2014 Oct/2014 Nov/2014 Dec/2014 DRP Coastal Lough Foyle River Foyle Daily loads are used to drive the nutrient input for the carrying capacity model, with simulated inputs to appropriate model boxes.

25 SWAT - source apportionment (Foyle only) Main DIN source: diffuse pasture follows fertilizer application DRP sources: point-source & diffuse Diffuse: low erosion rates leads to exports at background values DIN & DRP results broadly agree with Foy and Girvan, Area (Km2) Cropland POM sources: diffuse follows landuse Diffuse: low erosion rates leads to exports at background values Point source: negligible exports due to WW treatment Rangeland Pasture DIN (ton N/yr) DRP (ton P/yr) POM (ton/yr) WWTW - direct WWTW - streams Cropland WWTW - direct WWTW - streams Cropland WWTW - direct WWTW - streams Cropland Rangeland Pasture Rangeland Pasture Rangeland Pasture Without source apportionment, management is a black art. Source control in systems such as Lough Foyle is complex, costly, and socially challenging.

26 Hydrodynamic modelling for detailed circulation patterns Delft3D - Flow Atmospheric forcing River flow Tide Flow Water Quality / Ecological Modelling Free and open source, tidal response, drying and flooding, evaporative processes, inner shelf circulation, shelf stratification.

27 Components of the model: Bathymetry Hydrodynamic Modelling Review João Lencart e Silva

28 Some Like It Hot!

29 Before a model is deployed in EcoWin, multicriteria analysis and stakeholder consultation is used to define simulation areas. EcoWin model boxes Division of Lough Foyle into simulation areas 1. WFD water bodies 2. physical: bathymetry 3. physical: salinity 4. water quality River 1 River 2 Lough Foyle 5. aquaculture 6. Foyle river: salinity gradient 7. Coastal areas: WFD 8. expert review water bodies + LA area (Loughs Agency)

30 EcoWin.NET model - EASE Water quality validation chlorophyll (Model Year 2) Box 19 Data LF11 Chlorophyll (µg L -1 ) Time (days) EcoWin appears to reproduce the range and general pattern of chlorophyll variation.

31 EcoWin.NET Lower boxes are assigned for shellfish culture Blue mussel Native oyster Pacific oyster Yields vary substantially among boxes. The following slides show outputs for each species, with all objects active.

32 Down on the farm growing Manila clams in the USA

33 EcoWin.NET Lough Foyle Model Blue mussel harvest over time standard model Box 32 Box 38 Box 42 Box Harvest (tonnes y -1 ) Harvest (tonnes y -1 ) Time (Days) Blue mussel spin-up is very fast and harvest stabilizes.

34 EcoWin.NET Blue mussel Average Physical Product (APP) Average Physical Product (APP) > No culture Yields vary substantially among boxes, with best performance in the northern and middle upstream sections of Lough Foyle.

35 Conceptual model of eutrophication Top-down control : the circuit-breaker between primary and secondary symptoms.

36 Chlorophyll (µg L -1 ) EcoWin.NET Lough Foyle Model Phytoplankton drawdown standard model, Year Box 32 Box 35 Box 37 Box 42 Box Time (Days) The strongest drawdown is in the central and upper parts of the lough, where both native oyster (O. edulis) and blue mussel (M. edulis) are grown.

37 Shellfish culture outperforms source-control in controlling eutrophication symptoms, and provides an additional provisioning service. Mean chlorophyll Percentile 90 (µg L -1 ) EcoWin.NET Lough Foyle Model Chlorophyll drawdown with bottom-up and top-down control Standard seed 75% seed Increasing nutrient load Natural 10% load 20% load 50% load 75% load Standard Bottom-up control of eutrophication symptoms through nutrient source reduction, no cultivated shellfish in Lough Foyle 50% seed 20% seed 10% seed Bottom-up Top-down Top-down control of eutrophication symptoms through cultivated shellfish, standard nutrient loading in Lough Foyle Increasing shellfish stock No shellfish

38 Valuation of chlorophyll drawdown (M y -1 ) EcoWin.NET Lough Foyle Model Valuation of shellfish ecosystem services Increasing shellfish stock 20% seed 50% seed 75% seed Standard seed Equivalent N credit value: 8.7 M y -1 N source control cost (M y-1) Provi s ioning (M y-1) Regulatory s ervices (M y-1) 20% to10% load 50% to 10% load 75% to 10% load 100% to 10% load Increasing nutrient source control Value of harvested biomass The value of the regulatory service of chlorophyll removal by shellfish is far higher than the value estimated through direct nutrient removal.

39 What the model framework helped us learn There is no silver bullet: different models work within a framework but have standalone capabilities; Some areas of Lough Foyle have elevated chlorophyll peaks, production is driven by diffuse sources of nutrients; Cost of lowering chlorophyll P 90 by 1 ug L -1 is about 27 M y -1 ; Shellfish culture is key for eutrophication management more effective than nutrient control, but does not exempt agricultural BMP (moral hazard), or resolve microbiological issues; This framework supports the EU Water Framework Directive (WFD), Marine Strategy (MSFD), and Habitats. Applications include licensing, restoration, risk assessment, climate change analysis. This kind of work requires 1-2 years for an experienced team.

40 Screening models Distilling information Used for broad comparison and assessment Relate pressure, state and response May be ecosystem scale or other scales, e.g. regional, fish farm Are highly aggregated and easy to apply Can be data-driven or use inputs from more complex models Are easily understood and interpreted by managers Screening models synthesise information, and are quick and easy to apply

41 Screening models Examples Model name Objective Usage OSPAR-COMPP Eutrophication assessment EU, mainly NE Atlantic ASSETS Eutrophication assessment US, Worldwide, EU HEAT Eutrophication assessment EU mainly Baltic DEPOMOD Local effects of finfish aquaculture Scotland, elsewhere MOM Local effects of salmon aquaculture Norway, elsewhere ORGANIX Local effects of suspended culture Tests ongoing FARM Open water shellfish culture, IMTA EU, US, China FARM Pond-scale culture EU, Asia FISHNETS Disease modelling Tests ongoing Useful for diagnosis, scoping, and comparisons

42 Screening models case studies Two examples Screening models (farm-scale): integrated multi-trophic aquaculture Screening models (system-scale): eutrophication assessment A.M. Cubillo, J.G. Ferreira, S.M.C. Robinson, C.M. Pearce, R.A. Corner, and J. Johansen Role of deposit feeders in integrated multi-trophic aquaculture - a model analysis. Aquaculture, 453, Bricker, S.B., J.G. Ferreira, T. Simas, An Integrated Methodology for Assessment of Estuarine Trophic Status. Ecological Modelling, 169(1),

43 Role of Deposit Feeders in Integrated Multi-Trophic Aquaculture Case study Integrated Multi-Trophic Aquaculture (IMTA) Supply of organic matter to the benthos Individual model for deposit feeders FARM model for population in monoculture and IMTA

44 Conceptual diagram for IMTA

45 Allochtonous supply of organic material to benthic deposit-feeders below a fish cage Background organics A f Finfish cage z f Sea cucumber pens z S b S w S f S b : Background loading (g d -1 ) S w : Waste feed loading (g d -1 ) S f : Faecal loading (g d -1 ) A f : Area of polar cage (m 2 ) A d : Area of benthic footprint (m 2 ) z: Water column depth (m) Z f : fish cage depth A d The simplest model with no advection or dispersion considers A d = A f

46 Allochtonous supply of organic material to deposit-feeders under a fish cage Polar cage z A d Longitudinal (main) current axis Advection shifts the dispersion footprint as a function of the residual current.

47 Mass balance for an Atlantic salmon growth cycle Matched FCR and end-point weight.

48 Feed Conversion Ratio (FCR) and mass apportionment Example for 1kg of fish, FCR = 1.12 FW to DW conversion Consider a moisture content of 73.65% for Salmo salar muscle (Atanasoff et al., 2013): 1.00 kg wet weight = kg DW. Feed 1120 g DW Fish intake 1033 g DW FCR 1.12 Fish production 1000 g WW Total loss 87 g DW Assimilation 83% Feed used 1033 g DW = Fish faeces 177 g DW + Metabolism + Equiv g DW Fish mass g DW FCR is the result of input/output. Input-Output = Total Loss

49 Organic Sedimentation Model - ORGANIX ORGANIX predicts the benthic loading footprint. Many other models (Gowen, Silvert, Cromey, Corner, and respective co-workers) do this; Dispersion in 2 dimensions is based on Gaussian distribution functions; Advection is based on residual circulation; Model algorithm determines time to settle based on fall velocity. Probability distribution (dispersion) and advective shift is determined at each timestep until the plume reaches the bottom; Loading from culture structures is distributed over the modelled surface; Calibration for Atlantic Salmon, experimental data from DFO and literature. feed pellets fall faster than faeces; ORGANIX does not account for physiological variation. Calculation of bottom loading and spatial distribution under different culture and environmental conditions is essential for deposit feeder model.

50 ORGANIX ORGANIC Sedimentation model Multiple deposition plumes of waste feed and faeces for 14 salmon cages

51 Sea cucumber a delicacy in China Sea cucumber is the most expensive seafood product in China.

52 Simulation of sea cucumber growth in integrated culture under salmon farms Live Weight (g) Days 23 gpom m-2 d-1 9 gpom m-2 d gpom m-2 d-1

53 Mass balance for a four year sea cucumber growth cycle Parastichopus californicus weight data - large animals: g WW (Hannah et al, 2013), g WW (Hannah et al., 2012).

54 FARM model IMTA layout Fallow Kelp Salmon Oysters Water flow 200 m 50 m Farm (zoomed view) 200 m Deposit feeders cover the whole bottom (40,000 m 2 per section) Water flow Farm (full view) FARM simulates changes to individual weight, harvest, environment, and income.

55 FARM model Application to Integrated Multi-Trophic Aquaculture (IMTA) FARM model for finfish, shellfish, seaweed, and deposit feeders. A.M. Cubillo, J.G. Ferreira, S.M.C. Robinson, C.M. Pearce, R.A. Corner, and J. Johansen Role of deposit feeders in integrated multi-trophic aquaculture - a model analysis. Aquaculture, in press.

56 FARM model IMTA of Atlantic salmon and sea cucumber Model setup: Area of 600 m (3 X 200 m sections) by 200 m; sea cucumber density for standard model: 5 ind. m -2 ; culture period for tests: 400 days; drivers as in WinFish. FARM simulates changes to individual weight, harvest, and income.

57 Synthesis of FARM outputs for deposit feeders Scenario Mono IMTA 1 5 fish m -2 IMTA 2 20 fish m -2 IMTA 3 Oysters Individual weight (g) IMTA 4 IMTA 2 + IMTA 3 IMTA 5 IMTA4 + seaweeds Length (cm) Harvest (t cycle -1 ) APP Profit (k ) as EBITDA POM removal ( gc m -2 y -1 ) Net POM loading (g C m -2 y -1 ) Populationequivalents (y -1 ) Scenarios for monoculture (20 ind. m -2 ), different finfish densities in IMTA, shellfish longline culture (100 ind. m -2 ), shellfish + finfish, and seaweeds (50 ind. m -2 ). IMTA6 (not shown) increases deposit feeders to 80 ind. m -2.

58 Kelp monoculture: final individual weight of 134 g Increases to 175 g in IMTA5 Two key questions Role of seaweed (winged kelp Alaria esculenta) culture 22% increase in total physical product (TPP) for plants of harvestable size from 153 to 214 t cycle -1 No significant effect on DIN concentration (P 90 decreases by 0.4 µm) Role of suspended shellfish (Pacific oyster C. gigas) culture Oyster individual weight increases from g to g TPP from to t cycle -1 Increase of ratio of suspended particles to 80% makes little difference (end points are 65.7 g and t) Shellfish suspended culture is not enhanced by salmon culture; seaweeds do not reduce DIN significantly. This is basin-scale IMTA.

59 The Eutrophication Process From: Bricker et al National Estuarine Eutrophication Assessment Update

60 Eutrophication Stages of environmental degradation From: Bricker et al National Estuarine Eutrophication Assessment Update

61 Method MSFD guidance synthesis Eutrophication assessment models Biological Indicators Physico - Chemical Indicators Load related to WQ Integrated final rating TRIX Chlorophyll (Chl) DO, DIN, TP No Yes EPA NCA WQ Index ASSETS LWQI/TWQI OSPAR COMPP UK WFD HEAT IFREMER Chl Water clarity, DO, DIN, DIP No Yes Chl, macroalgae, seagrass, HAB Chl, macroalgae, seagrass Chl, macroalgae, seagrass, PP indicator spp. Primary production, Chl, macroalgae, benthic invertegrates, seagrass Chl, macroalgae, benthic invertegrates, seagrass, HAB Chl, seagrass, macrobenthos, HAB DO Yes Yes DO, DIN, DIP No Yes DO, DIN, DIP, TP, TN, Yes Yes Water clarity, DO, DIN, DIP, TN, TP Water clarity, DO, DIN, DIP, TN, TP, C Water clarity, DO, DIN, SRP, TN, TP, sediment organic matter, sediment TN, TP Some methods do not consider pressure-state relationships Ferreira et al. 2011, Estuarine Coastal and Shelf Science, 93, No No No Yes Yes Yes

62 Indicators used by various assessment methods Indicators ( 评价指标 ) Nutrient Index I* Nutrient Index II* EPA NCA OSPAR COMPP ASSETS Nutrient (N,P) load, conc. X X X X X Chemical oxygen demand X X Chlorophyll a X X X X Dissolved oxygen X X X X X Water clarity HABs (nuisance/toxic) X X Phytoplankton indicator sp. Macroalgal abundance X X Seagrass loss X X Zoobenthos-fish kills Temporal focus Unspecified Unspecified Summer Spring/winter Full year Integration Additive Ratio Ratio Integration PSR * Commonly applied in China Methods with red crosses fall short of a full eutrophication assessment Adapted from: Xiao et al. 2007, Estuaries. and Coasts 30: X X X

63 Key aspects of the ASSETS approach Three stages... The ASSETS approach may be divided into three parts: Division of coastal systems into homogeneous areas Evaluation of data completeness and reliability Application of indices Tidal freshwater (<0.5 psu) Mixing zone ( psu) Seawater zone (>25 psu) Spatial and temporal quality of datasets: completeness Confidence in results: sampling and analytical reliability Influencing Factors (IF) index Eutrophic Condition (EC) index Future Outlook (FO) index Pressure State Response S.B. Bricker, J.G. Ferreira, T. Simas, An integrated methodology for assessment of estuarine trophic status. Ecol. Modelling 169:

64 ASSETS Approach: Pressure - State - Response Influencing Factors (IF) Eutrophic Condition (EC) Future Outlook (FO) Susceptibility Low Moderate High Moderate Moderate Low Low Moderate High Moderate Low High Moderate High Moderate Low Primary Symptoms Low Moderate High Moderate Moderate Low Low Moderate High Moderate Moderate Low High High Moderate High Susceptibility High Moderate Low Improve High Improve Low Improve Low No Change No Change No Change Worsen Low Worsen Low Worsen High Low Moderate High Nutrient Pressures Susceptibility dilution & flushing + Nutrient Inputs land based or oceanic Influencing Factors Low Moderate High Secondary Symptoms Primary Symptoms Chl and Macroalgae Average of ratings Secondary Symptoms D.O., HABs, SAV change Worst case Decrease No Change Increase Future Nutrient Pressures Susceptibility Nutrient pressure changes population, management, watershed use (particularly agricultural) IF + EC + FO = ASSETS Full accounting of eutrophication symptoms, including time and space Adapted from: Bricker et al. 2003, Ecological Modelling, 169(1), 39-60

65 ASSETS Strangford Lough, N. Ireland Indices ASSETS: HIGH Methods Parameters Rating Expression Index Influencing Factors (IF) ASSETS: 5 Susceptibility Dilution potential High Low Flushing potential Moderate susceptibility LOW Nutrient inputs Low Chlorophyll a Moderate Eutrophic Condition (EC) ASSETS: 5 Primary Secondary Macroalgae Dissolved Oxygen Problems observed No problems Moderate Submerged Aquatic Losses Vegetation observed Low LOW Nuisance and Toxic Blooms No Future Outlook (FO) ASSETS: 4 Future nutrient pressures Future nutrient pressures decrease Improve Low High status system, classified as an SAC under UK law.

66 ASSETS Combination of research and screening models Eutrophic Condition (OEC) ASSETS OEC: 4 Methods PSM SSM Parameters Value Level of expression Chlorophyll a 0.25 Epiphytes Macroalgae 0.96 Moderate Dissolved Oxygen 0 Submerged Aquatic Vegetation Low Nuisance and Toxic 0 Blooms Index MODERATE LOW Eutrophic Condition (OEC) ASSETS OEC: 4 PSM SSM Chlorophyll a 0.25 Epiphytes Macroalgae 1.00 Moderate Dissolved Oxygen 0 Submerged Aquatic Vegetation Low Nuisance and Toxic 0 28% lower Blooms MODERATE LOW Eutrophic Condition (OEC) PSM ASSETS OEC: 4(5) SSM Chlorophyll a 0.25 Epiphytes Macroalgae 0.50 Moderate Dissolved Oxygen 0 Submerged Aquatic Vegetation Low Nuisance and Toxic 0 Blooms MODERATE LOW

67 ASSETS Pressure-State-Response 压力状态反馈 ASSETS 结果 Influencing Factors Eutrophic Condition Future Outlook ASSETS Top-down control shellfish aquaculture 贝类养殖的下行控制效应 Huang He Sanggou Jiaozhou Chiangjiang Huangdun Sanmen Xiamen Daya Zhujiang High Moderate High Moderate Moderate Low Low or No Problem Unknown Worsen High Worsen Low No Change Improve Low Improve High Unknown Bad Poor (WFD) Moderate Good High Unknown or Not Applicable

68 Summary No question, no model. What is your question? No model can predict the weather. The weather affects circulation (wind, freshwater flow), salinity (rainfall), food (chlorophyll depends on e.g. clouds, temperature). Ecosystem models show general patterns; Many different models exist. Models are simplifications of reality, but can be very useful. No model does everything; Models can (and often should) be combined, which often adds huge value to the end product. All slides