SYSTEMS APPROACH TO THE ASSESSMENT OF CLIMATIC CHANGE IN A SMALL RIVER BASIN

SYSTEMS APPROACH TO THE ASSESSMENT OF CLIMATIC CHANGE IN A SMALL RIVER BASIN Predrag Prodanovic1 and Slobodan P Simonovic1 Department of Civil and Environmental Engineering, University of Western Ontario, London, Ontario, Canada, N6A 5B9; email: pprodano@uwo.ca Abstract: The focus of this study is on the mathematical modeling support for implementation of the inverse approach to climatic change impact assessment in small river basins The model used in the assessment of risk and vulnerability to changing climatic conditions is described in this paper The model is a combination of a sophisticated system dynamics model linked to a continuous hydrologic model of a river basin The system dynamics model represents the socio-economic components such as urban and rural population, housing, business activity, land and water use The hydrologic model is a seventeen parameter semi-distributed water balance model The combined model links the socio-economic aspects of a river basin (such as changing land use patterns) to physical processes (such as potential evapotranspiration and direct runoff) The physical processes (through direct runoff and precipitation) are linked to flood damage and drought levels, which then directly influence the population and business sectors of the socio-economic component A number of such key feedback relationships are discussed in the paper The application of the model is illustrated with a case study of a small river basin (3,500 km 2) in southwestern Ontario, Canada Key words: Continuous Hydrologic Modelling, System Dynamics, Systems Thinking, Climate Change, Risk, Vulnerability Introduction Climatic change impact and vulnerability assessments have received much attention in the past two decades because of a growing realization that climate is changing as a result of human activities (IPCC, 2001) Burning fossil fuels and changing land use patterns have lead to an increase in atmospheric concentration of carbon dioxide (and other greenhouse gases), thus resulting in a steadily increasing mean global surface temperature Documented evidence exists that increased global surface temperature intensifies the hydrological cycle, and therefore induces changes in the frequency and magnitude of hydrologic extremes such as floods and droughts Traditional ways of studying the climatic change in river basins is by downscaling output from global circulation models, and using them as inputs in the hydrologic models A number of studies have implemented such methodologies, and thus estimated impacts of changed climate (Coulibaly and Dibike, 2004; Palmer et al., 2004) However, a number of uncertainties are inherent to this approach Firstly, the global models have temporal scales that are sometimes incompatible with temporal scales of river basins For example, the rainfall-runoff modelling requires data with relatively short time steps (daily and/or hourly); the global models however, are only able to produce monthly outputs with high degree of confidence (Cunderlik and Simonovic, 2006) Temporal downscaling of monthly global output must therefore be employed, and shorter duration events be estimated, thus compounding uncertainty Secondly, spacial scales of global models can also be incompatible with spacial scales of river basins The global models typically have resolutions between 3-5 deg in latitude and 5-10 deg in the earth’s longitude, and are thus significantly larger than many river basins Such coarse resolution is thus inadequate in the representation of many relevant smaller scale river basin phenomena The traditional approach is plagued by uncertainties in spacio-temporal downscaling from global to local scales Furthermore, in the study by Coulibaly and Dibike (2004), the authors report significant differences between different downscaling methods (i.e., some downscaling techniques would produce an increase, while others a decrease in mean annual flow for the same global input) Therefore the choice of downscaling techniques can drive the outcome of the analysis, and thus mask the true system behaviour under the conditions of altered climate As a result the end-users (water management authorities, government policy makers and stakeholders) became sceptical of results of climate change impact analysis The inverse approach, originally developed by Cunderlik and Simonovic (2004b, 2006) takes an alternate (or bottom-up) route to climate change impact assessment Its main strength is that it focuses on users of water resources systems, and keeps them involved in the process The inverse analysis includes the following steps (modified from Cunderlik and Simonovic, 2006): The critical hydrologic exposures (such as floods and droughts) leading to failures of the water resources systems are first identified, together with their risk levels The end-users are involved in this stage, as they are most familiar with particulars of the water resources system in question The critical hydrologic exposures of the previous step are transformed into their corresponding critical meteorological conditions (such as extreme precipitation, sudden warming, sustained dry spells) with a hydrologic model Relationships between critical hydrologic exposures and their corresponding meteorological conditions for each location of interest are formulated They capture specific management practices of the end-users (such as flood/water control, environmental and watershed planning, source water protection, etc.) A weather generator is used next to simulate critical meteorological conditions of present and future climates Meteorological data is synthetically generated for an arbitrary long period that is statistically similar to the observed historical record The generated data is of high spacio-temporal resolution, as required by the hydrologic model Altered climate scenarios are then generated that are conditioned upon the historical data (such as increased or decreased precipitation, warmer or wetter springs, etc) and linked to large scale global circulation model outputs Using an ensemble of weather generator scenarios, relationships similar to those of step are obtained for each scenario considered Such information can provide an estimate of change in the frequency of critical exposures The last step in the inverse approach requires an application of the integrated assessment model that captures behaviour of both physical and socio-economic systems, and thus evaluates risks and vulnerability to changing climatic conditions within the study area The focus of this paper is on the development of a model for use in the last step of the inverse approach In order to adequately describe the physical and socio-economic processes operating within a basin, various links between the two must be established The links sought are of the feedback type, where a state of the system is used to formulate a decision rule, which then alters the state of the system itself Feedback links between hydrologic and socio-economic processes of a region are relatively straightforward For example, an adequate water supply of an area depends on the availability of water within a basin (via aquifers or surface storage) As an area expands, the total demand for water increases, while the availability of water decreases (as lake and ground water levels are lowered) As lake and ground water levels change, so does the regional hydrology, thus completing a feedback loop Although relatively straightforward, this kind of (feedback) thinking is seldom used in practice, where linear or sequential reasoning is most often employed In Canada, cities and municipalities in preparing their master plans take exogenous population projections, compute system demand based on growth rates, and compare it against known sources of supply This kind of reasoning misses key feedback relationships that in reality govern the behaviour of the entire system A paradigm that uses feedback loops as its fundamental building block is referred as systems thinking (Richardson, 1991; Senge, 1990), and is adopted in this study In its application, the boundaries of classical modelling (either hydrologic or socio-economic) are extended, thus forcing the user to postulate and test hypothesis regarding how a physical process affects, and is affected by, a social process In doing this, the user will be able to see implications of policies and decisions that otherwise wouldn’t be considered Use of systems thinking is not only instrumental in the assessment of risk and vulnerability of anticipated future conditions, but brings to the user useful knowledge upon which decisions can be made A number of research papers have recently appeared that use systems thinking and the concept of feedback to water resources management problems The research by Li and Simonovic (2002) demonstrates one of the first applications of system dynamics simulation and systems thinking to studying hydrologic processes in North American prairie watersheds The work of Saysel et al (2002) develops a system dynamics model for the purpose of studying agricultural development in Turkey The model consists of sectors representing water resources, land use, demography and agricultural pollution Stave (2003) on the other hand, uses a very simple model to bring about public understanding of water management options in Las Vegas, Nevada Problem definition similar to Saysel et al (2002) is adopted by Fernández and Selma (2004), who studied water scarcity of irrigated landscapes in Spain Sehlke and Jacobson (2005) proposed a model to investigate the utility of system dynamics approach in modelling large complex hydrologic systems and their interaction between the surface and ground water in the western United States Even though the above modelling efforts have been prepared for different purposes, all share one common feature: they study water resources management problems using the framework of systems thinking and feedback Nearly all employ some sort of interactions between physical and socio-economic realms However, all of the models share another commonality: they are not widely accepted by end-users, practitioners and policy makers As a result, the models rarely get used (other than by the authors themselves) If such models are constructed with the participation of the end-users, utilizing tools they are already familiar with, the outlook can be more promising This is what the current study is set out to The rest of the paper is organized as follows: Section provides a basic sketch of the modelling methodology It describes the model’s system structure, the data used, and the key feedback loops The application of the methodology is outlined in Section 3, together with outputs of simulated system behaviour Concluding remarks are given in Section Methodology The model used in the assessment of risk and vulnerability to climatic change is a combination of two different models, linked with a number of feedback loops The combined model consists of a continuous hydrologic model, coupled with a system dynamics model of the study area Each is described next Continuous hydrologic modelling provides a way to simulate the long term behaviour of a basin All such models found in practice (Bennett, 1998; Johanson et al., 1980; Leavesley et al., 1983) are based on the principles of water balance; they divide an area studied into a number of conceptual buckets (canopy, surface, soil and ground water) and estimate the flows between them (infiltration, percolation, runoff) The input to such models usually includes precipitation, temperature and sometimes solar radiation, and their outputs are flows (surface excess, baseflow, direct runoff, evapotranspiration, ground water recharge, etc.) In essence, continuous the hydrologic models are rainfall-runoff models that take into account the long term soil moisture balance In contrast, event based models are those with simplified account of the moisture balance (sometimes representing losses by a simple function or a coefficient), and are therefore used for determination of basin response from single storm events (Bedient and Huber, 1988, p 313) System dynamics is a methodology grounded in feedback control theory (Forrester, 1961) Its main premise is that of interconnectedness of system components; in other words, it seeks to uncover interconnections between different parts of a larger whole The methodology emphasizes a long-term aggregate continuous view point, meaning that actions resulting from major decisions not take place instantly, but rather gradually adjust over time The strength of system dynamics lies in using mathematical models to formulate, test, and re-formulate dynamic hypothesis of a real world problem, especially when social and physical interactions are present The system dynamics modelling can help the user gain insight into situations of dynamic complexity, and possibly reveal causes of policy resistance (Sterman, 2000) 2.1 Model Description The model developed as part of this research is a combination of a continuous hydrologic model and a system dynamics model The continuous hydrologic framework used in the study was one developed by the United States Army Corps of Engineers (Bennett, 1998; USACE, 2000)—Hydrologic Modelling Center Hydrologic Modelling System (HECHMS) The model was previously applied, callibrated and verified for the study area in the work by Cunderlik and Simonovic (2004a, 2005) The input data used in the model was produced by a weather generator developed by Burn and Sharif (2004) The weather generator model was used to create an ensemble of climate scenarios based on statistical properties of historical data, and selected data from the global circulation model outputs Time series of precipitation, maximum and minimum temperatures were synthetically generated for each scenario, and used as exogenous input into the hydrologic component of the model The system dynamics model consists of several sectors representing the most relevant socio-economic activities within the basin The model sectors include urban and rural demographics, housing, business activity, land and water use, and is an extension of the work of Alfeld and Graham (1976) The level of detail of only one such sector (urban business activity) is shown as a stock and flow map in Figure 1; note that italicized variables represent information obtained from other model sectors, not shown in the map Figure 1: Stock and flow map of the urban business activity sector The urban business activity sector consists of four feedback loops: availability of business land, labour, and (surface and ground) water Each of these imposes a limit to growth of business activity, meaning that if one limit is reached, growth stops The first loop is the land availability loop, and represents the dynamics of land occupancy by businesses Each business unit occupies an average amount of land, which aggregates (when all business units are taken into account) to a total business land occupied The availability of business land is a ratio of business land occupied to business land available (determined from the land use sector, not shown here), which then determines an investment rate in future business For example, if initial land occupancy is low, and if conditions are favourable (in terms of plentiful labour force, and availability of resources such as water), the investment rate will increase, and the number of businesses will grow However, as more and more land gets occupied, it becomes difficult for business units to acquire land and move to the area This is because “convenient access by truck or railroad is essential to merchandisers and manufacturers, and a variety of other businesses who must consider availability of customer and office parking”, among other things (Alfeld and Graham, 1976, p 179) Similar reasoning is employed in the formulation of the labour availability loop In terms of water availability loops, use to availability ratios are found by computing business water use, and dividing them by the available water (either from ground or surface sources) If the use to availability ratios are low (meaning businesses are using much less water than is available), business investment can proceed provided other conditions are favourable However, as the use to availability ratios grow, less and less water is available for new businesses, thus slowly imposing a limit to growth In cases like this, the local governments usually cease to issue additional permits for industrial and commercial uses of water, which in turn halts additional business investment 2.2 Combined Model Schematic and Key Feedback Loops A number of feedback loops are used to link the hydrologic component of the model (left side in Figure 2) with the socio-economic component (right side in Figure 2) For example, the hydrologic model computes a value of the ground water recharge, which the population and business sectors use The activities of these sectors generate the increases in largely impervious urban land Increasing urban land contributes to surface excess and direct runoff, thus reducing ground water recharge In order to assemble the feedback links between the socio-economic component (i.e., urban land) and the hydrologic component, a number of assumptions are introduced One such assumption is the relationship between urban land and the time of concentration, a part of the routing component in the hydrologic model Another link is between the urban land and the level of imperviousness The feedback relationships are shown in Figure Figure 2: Upper Thames System Model Structure Figure 3: Assumed feedback relationships The rationale for the shape of assumed relationships is the following As more and more of basins’s land becomes urbanized, the time of concentration decreases This is because increasing urbanization alters the path of runoff to basin’s outlet The urban land also has a direct effect on the level of imperviousness (a measure of how much rainfall becomes surface runoff, and does not infiltrate into the soil) A simple linear relationship is formulated for the purposes of this study Of course, other shapes could easily be tested using a sensitivity analysis Similar logic is used to formulate links between the amount of vegetation cover (forest and agricultural land) and potential evapotranspiration For example, as more and more of the basin’s area become urbanized, its vegetation cover decreases due to rezoning of forested and agricultural lands As the vegetation cover decreases, so the rates of potential evapotranspiration, which over time tend to increase the surface runoff An additional link exists between the drought level and water use The level of drought is computed according to local drought definitions (OLWR, 2003) Generally, when an area experiences conditions of drought, the local water management agency announces this information to the public, and asks for a reduction of non-essential water uses (such as washing cars, or watering lawns) As drought conditions worsen, additional water use restrictions may take place Case Study The model developed has been applied to the Upper Thames River basin, located in southwestern Ontario The basin’s total drainage area is about 3,500 km 2, and its population is slightly less than half a million 80% of the basin’s land area is classified as agricultural, 10% as urban, and 10% as forest The major urban center in the basin is the City of London (population of 350,000), which has experienced much growth in recent times Majority of urban growth has been taking place along the Thames River and its smaller tributaries Since the watershed is located in a highly developed part of the province of Ontario, the region constantly experiences increasing pressures from both urban and rural land uses The hydrologic model developed by Cunderlik and Simonovic (2004a, 2005) consists of a thirty two sub catchments (or spacial units), twenty one river reaches, and three major reservoirs The precipitation and temperature input data is obtained from the weather generator, developed by Burn and Sharif (2004) for a number of climate scenarios Sub catchment losses are computed via the Soil Moisture Accounting algorithm (Bennett, 1998), while hydrologic routing is employed for the river reaches and reservoirs via the Modified Puls method (USACE, 2000) The hydrological model operates on a time step of six hours The structure of the system dynamics model follows the description given in Section Only three spacial units are considered, one for each county in the basin—Middlesex, Oxford and Perth Three spacial units are selected partly because the socio-economic data for smaller spacial units were not available Each spacial unit of the model has an identical structure, but is populated with different data The system dynamics model operates on a monthly time step The two models are linked via feedback as described in the previous section Monthly average flows are used as inputs into the socio-economic component, which after processing provides the relevant information to the hydrologic component once per month (i.e., the parameters of the hydrologic model change on a monthly basis) This means that hydrologic parameters such as potential evapotranspiration, percent of imperviousness and time of concentration are updated once per month, even though the hydrologic model component operates on a six hour time step 3.1 Simulation Output The combined model is simulated for a 300 year period, for four different weather generator inputs: (i) historically identical scenario (or base case), (ii) an increasing and (iii) decreasing precipitation and (iv) increasing temperature scenario Daily annual maximum flows are selected and statistically analysed as indicators of vulnerability to flooding Seven day annual minimum flows are chosen as indicators of droughts The resultant statistical analysis plots for one of the stream gauges are shown in Figures 4-5 Figure 4: Frequency of daily annual maximum flows at Byron Figure 5: Frequency of day annual minimum flows at Byron In terms of floods, we can observe that a flow of, say 600 m 3/s has a return period of about 100 yrs for the historically identical case, and an 18 yr return period for the increasing precipitation case This result is somewhat expected -an increasing precipitation tends to decrease the return period of flows However, we also observe an unexpected result: peak flows past a return period of about 30 yrs tend to increase for a decrease in precipitation scenario when compared to the historical or base case This is indeed counterintuitive, and it implies that it is possible for the peak flow to increase, even if precipitation decreases This result can be partially explained by continuously increasing urbanization, which gradually increases peak flows The numerous feedback processes operating in the combined model are responsible for producing counter-intuitive behaviour such as this It is worthwhile mentioning that classical studies which consider steady state hydrologic models (i.e., ones that not change as social and physical conditions of the basin change) are unlikely to produce behaviour such as this Regarding low flows, we see more of an orderly result Under the increased precipitation scenario, low flows generally increase, while for the increased temperature and reduced precipitation scenario, low flows mostly worsen The outputs of the system dynamics model are shown as time series plots The variables selected are ground water recharge, business activities (in terms of business units), and business ground water use, shown in Figures 6-8 Figure 6: Simulated ground water recharge Figure 7: Simulated Business Units Figure 8: Simulated business ground water use Rapid development for the first fifty years of simulation tends to reduce ground water recharge significantly However, it is interesting to note that after the intensive development, ground water recharge springs back up (although short of its pre-development level) The increasing precipitation scenario shows the highest recharge values, while the reduced precipitation scenario the lowest Area’s business activity (shown in Figure 7) tend to grow for the first forty years of simulation, although at a declining rate In essence, the business activities are limited by water availability (see Figure 8) some twenty years after the beginning of the simulation, and are therefore unable to increase further We see that nearly all scenarios equilibrate to roughly the same final value; we also see that decreased precipitation scenario shows the lowest peak value when compared to other scenarios considered With the difference between the historic and the changed climate scenarios, we can observe the impact of climatic change on area’s business activities Conclusions This paper presents a framework for assessment of climatic change impacts, risk and vulnerability for small watersheds The inverse approach (an alternative to downscaling of global circulation model outputs) is applied A continuous hydrologic model making use of the latest available knowledge of hydrologic processes is coupled with a system dynamics model representing socio-economic processes of the study area The benefits of the selected approach are mainly in the improved knowledge and understanding of the major drivers of change in the basin A number of insights are identified, especially those regarding the magnitude of vulnerability and risk that a changed climate could potentially impose Acknowledgements The research described in this paper was supported by grants from the Canadian Foundation for Climatic and Atmospheric Sciences (CFCAS), and Natural Sciences and Engineering Research Council (NSERC) to whom the authors are grateful The authors would also like to thank Dr Juraj Čunderlík, Dr Donald H Burn, and Mr Mark Helsten for their assistance during the course of the research References Alfeld, L E and Graham, A K (1976) Introduction to Urban Dynamics Wright–Allen Press, Inc., Cambridge, Massachusetts Bedient, P B and Huber, W C (1988) Hydrology and floodplain analysis Addison-Wesley Publishing, Upper Saddle River, NJ, second edition Bennett, T (1998) “Development and application of a continuous soil moisture accounting algorithm for the Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS) Master’s thesis, Department of Civil and Environmental Engineering, University of California, Davis, CA Burn, D H and Sharif, M (2004) “Assessment of water resources risk and vulnerability to changing climatic conditions: Development and application of a K-NN weather generating model.” Report No III, The University of Western Ontario, London, Ontario, Canada Coulibaly, P and Dibike, Y B (2004) Downscaling of Global Climate Model Outputs for Flood Frequency Analysis in the Saguenay River System Final Project Report prepared for the Canadian Climate Change Action Fund, Environment Canada, Hamilton, Ontario, Canada Cunderlik, J M and Simonovic, S P (2004a) “Assessment of water resources risk and vulnerability to changing climatic conditions: Calibration, verification and sensitivity analysis of the HEC-HMS hydrologic model.” Report No IV, The University of Western Ontario, London, Ontario, Canada Cunderlik, J M and Simonovic, S P (2004b) “Inverse modeling of water resources risk and vulnerability to changing climatic conditions.” 57th Canadian Water Resources Association Annual Congress, Montreal, Quebec, Canada Cunderlik, J M and Simonovic, S P (2005) “Hydrological extremes in a southwestern Ontario river basin under future climate conditions.” Hydrological Sciences, 50(4), 631–654 Cunderlik, J M and Simonovic, S P (2006) “Inverse flood risk modeling under changing climatic conditions.” Hydrological Processes, (in print) Fernández, J M and Selma, M A E (2004) “The dynamics of water scarcity on irrigated landscapes: Mazarrón and Aguilas in south-eastern Spain.” System dynamics Review, 20(2), 117–137 Forrester, J W (1961) Industrial Dynamics The MIT Press, Cambridge, MA IPCC (2001) Climate Change 2001: Scientific Basis Contribution of the Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change Cambridge University Press, Cambridge, UK Johanson, R., Imhoff, J., and Davis, H (1980) “User’s manual for Hydrological Simulation Program-FORTRAN (HSPF).” Report No EPA-600/9-80-015, United States Environmental Protection Agency, Athens, GA Leavesley, G., Lichty, R., Troutman, B., and Saindon, L (1983) “Precipitation-Runoff Modeling System (PRMS), User’s manual.” Report No 83-4238, U.S Geological Survey Water-Resources Investigations, Denver, CO Li, L and Simonovic, S (2002) “System dynamics model for predicting floods from snowmelt in North American prairie watersheds.” Hydrological Processes, 16, 2645–2666 OLWR (2003) Ontario Low Water Responce Manual Ontario Ministry of Natural Resources, Queen’s Printer for Ontario Palmer, R N., Clancy, E., VanRheenen, N T., and Wiley, M W (2004) The Impacts of Climate Change on The Tualatin River Basin Water Supply: An Investigation into Projected Hydrologic and Management Impacts Department of Civil and Environmental Engineering, University of Washington, Seattle, WA Richardson, G P (1991) Feedback Thought in Social Science and Systems Theory University of Pennsylvania Press, Philadelphia, PA Saysel, A K., Barlas, Y., and Yenigün, O (2002) “Environmental sustainability in an agricultural development project: a System dynamics approach.” Journal of Environmental Management, 64, 247–260 Sehlke, G and Jacobson, J (2005) “System dynamics modeling of transboundary systems: The Bear River Basin model.” Ground Water, 43(5), 722–730 Senge, P M (1990) The Fifth Discipline: The Art & Practice of the Learning Organization Doubleday Currency Press, New York, NY Stave, K A (2003) “A system dynamics model to facilitate public understanding of water management options in Las Vegas, Nevada.” Journal of Environmental Management, 67, 303–313 Sterman, J D (2000) Business Dynamics: Systems Thinking and Modeling for a Complex World Irwin McGraw-Hill, Boston, MA USACE (2000) Hydrologic Modelling System HEC–HMS, Technical reference manual United States Army Corps of Engineers, Davis, CA  ... business units are taken into account) to a total business land occupied The availability of business land is a ratio of business land occupied to business land available (determined from the land... study area The benefits of the selected approach are mainly in the improved knowledge and understanding of the major drivers of change in the basin A number of insights are identified, especially... following As more and more of basins’s land becomes urbanized, the time of concentration decreases This is because increasing urbanization alters the path of runoff to basin? ??s outlet The urban land