Minimizing Impact of the Urbanization on the Physical Habitat Suitability of Downstream River by a Multi-objective Optimization


 Urbanization might considerably change outflow of catchment that might affect physical habitat suitability at downstream river ecosystem. Present study proposes and evaluates an applicable method to minimize impact of urbanization on the suitability of physical habitats in which habitat loss and area of the urban region are optimized. coupled particle swarm optimization- adaptive neuro fuzzy inference system is used to simulate runoff in the structure of a multi-objective metaheuristic optimization. Fuzzy physical habitat simulation was applied to simulate suitability of physical habitats. Different measurement indices including the Nash–Sutcliffe model efficiency coefficient, root means square error and vulnerability index were utilized to measure performance of the simulation-optimization system. Based on the results in the case study, the proposed system is able to mitigate physical habitat impacts by optimizing area of the urban region. Regional government had planned to urbanize 90% of the catchment area. However, it damages physical habitats considerably. The optimal plan reduced the urban area to 56% and minimized physical habitat loss. This method is able to reduce negotiations between regional governments and environmental advocators for development of the new urban areas in terms of minimizing physical habitat loss in river ecosystems.


1-Introduction 23
Stormwater is mainly defined as runoff that originates from the rain including snow and ice melt. The 24 most generated runoff is directly conveyed to the rivers or other water bodies (Jefferson et.al, 2017). One 25 of the main tasks for the civil engineers in the urban management is appropriate management of the 26 stormwater. In other words, stormwater management might be defined as controlling the surface runoff to 27 reduce water pollution and restore ecosystem integrity (Shishegar et.al, 2018). Urbanization is one of the 28 main challenges in the stormwater management. Because, the primary effect of the urbanization is 29 increasing impervious surfaces compared with previous surfaces (Kong et.al, 2017). In other words, 30 urbanization increases surface runoff to the river ecosystems. Owing to importance of the stormwater 31 management in the urban areas, different hydrodynamic and hydrologic models have been developed to 32 improve management of the urban areas. 33 As a review on the used models in the urban water management, two models including SWMM and 34 MUSIC are briefly reviewed. The EPA Storm Water Management Model (SWMM) is one of the 35 applicable models for simulating urban water quality and quantity. This model is mainly applied in the 36 post-development runoff, surface drainage hydraulics, detention pond design, low impact development, 37 runoff water quality, runoff treatment, dual drainage systems, combined sewer overflows and continuous 38 simulations (Gironás et.al, 2010).Moreover, MUSIC is the Model for Urban Stormwater Improvement 39 Conceptualization. In fact, MUSIC is a decision-making system that enables engineers for evaluating 40 conceptual designs of the stormwater management systems. MUSIC applies a risk-based approach in 41 which three main examinations are carried out as follows (eWater, 2011).  were applied to measure predictive skills of the runoff data driven model as displayed in the following 136 equations. NSE is originally developed for the hydrologic models (Gupta et.al, 2009). Thus, it is 137 applicable in the present study. 138 (1) 139 where OBSt is observed or recorded data in the time step t, SIMt is the simulated data by the model and T 141 is total number of time steps. 142

2-2-Optimization system 143
The main purpose of the optimization system is to minimize physical habitat loss in which the area of the 144 urban region is optimized. On the one hand, the regional government might aim maximum development 145 of the urban areas. On the other hand, there is a serious concern regarding impacts of the increasing 146 surface runoff on the physical habitat loss in the stream ecosystem. It should be noted that based on the 147 initial survey in the case study, it was demonstrated that current physical habitat suitability is close to the 148 natural flow. The main component of each optimization system is the objective function. Two objective 149 functions were developed including physical habitat loss objective function and urban area objective 150 function as displayed in the following equations. In fact, first objective function is responsible to reduce 151 concerns of the environmental advocators. Conversely, the second objective function would maximize the 152 developing urban area that is aimed by the regional government. 153 where NWUACt is normalized weighted useable area in the current condition, NWUAFt is normalized 156 weighted useable area in the future condition or initial urbanization plan, MUR is maximum percentage of 157 the urban area based on the initial plan in the catchment and OURt is optimized percentage of the 158 urbanized area. Figure 3 displays fuzzy physical habitat simulation method that is used in the present 159 study. More details have been addressed in the literature (Noack et..al, 2013).. Owing to development of two objective functions in the optimization system, a multi-objective 163 optimization algorithm is required. We selected one of the known multi objective metaheuristic 164 algorithms to optimize surface runoff considering physical habitat impacts. Multi-objective particle 165 swarm optimization (MOPSO) has been applied in the optimization problems successfully.

2-3-Case study 177
Tajan River is one of the important rivers in the southern Caspian Sea basin in Iran. This river basin is 178 one of the popular regions for living in the country due to appropriate weather condition. Hence, 179 urbanization is a challenge in this river basin. In other words, increasing population raises needs for more 180 urbanizing areas that mean impervious areas will be increased.

3-Results and Discussion 203
In the first step, it is necessary to present results of the data driven model to simulate runoff in the scale of 204 the catchment and sub-catchments. Two sub-catchments were selected to demonstrate ability of ANFIS 205 based model to simulate outflow in the simulated period. It should be noted that these sub-catchments 206 were different in terms of land use. In other words, one of them mainly includes non-urban areas. 207 Conversely, second sub-catchment mainly includes urban areas. Figures 6 and 7  when NSE is more than 0.5, predictive skill of the model is highly robust. Minimum NSE in the sub-211 catchments is more than 0.9. Thus, model is robust in terms of the NSE. However, using one index might 212 14 not be sufficient to authenticate abilities of the data driven model for further applications. RMSEs in the 213 non-urban sub-catchment and urban sub-catchment are 45.2 and 11.7 L/s respectively that demonstrate 214 model might be reliable in the most of time steps. However, model might not reliable to predict outflow 215 in very low rate of flows. It should be noted that stormwater management is the main purpose of the 216 proposed framework that means higher outflow might be noticed. Thus, using the developed model is 217 reliable. Area of the simulated sub-catchments is different. Hence, outflow in the non-urban sub-218 catchment is much higher than other one. In the next step, it is necessary to present and discuss on the results of the fuzzy physical habitat 228 simulation at the downstream river ecosystem of the simulated catchment. Table 3 displays developed  229 fuzzy rules for the physical habitat simulation based on the expert opinions and field observations in the 230 river habitats of the studied basin for the target species that is the Brown trout.    optimization algorithm presents different areas for the urban region in each time step to achieve the 259 optimal solution. However, it is not useable practically. Because, area of the urban region could not be 260 changed in different months. Hence, it is essential to apply a statistical index to assess the area of the 261 urban region. Arithmetic mean was considered as an index in this regard as displayed by the dash line in 262 the figure 11. Based on this figure, urban areas should be considered as 56% of the total area of the 263 simulated catchment to minimize physical habitat impacts considering physical habitat loss modeling. In 264 the next step, it is necessary to evaluate optimal plan for the urbanization by the multi-objective 265 optimization. 266 Figure 12 displays the normalized physical habitat loss in three scenarios including current condition, 267 initial plan for the urbanization and optimal plan for the urbanization proposed by the MOPSO. As 268 presented in the previous section, urban area had been planned up to 90% of the total area in the initial 269 plan. A significant difference between the current condition and the initial plan of the urbanization in 270 some time steps is a serious threat for the physical habitats. For example, difference between the current 271 condition and the initial plan for the urbanization in the third time step is 20% approximately that 272 increases physical habitat loss for the fish considerably. In contrast, the difference between optimal plan 273 and the current condition is limited. In fact, the performance of the optimization algorithm seems robust 274 to minimize physical habitat loss. However, more discussion needs using measurement indices. Figure 13  275 displays outflows of the catchment in three scenarios including the current condition, initial plan for the 276 urbanization and optimal plan urbanization. This figure demonstrates that performance of the 277 optimization algorithm is robust as well. Optimal plan reduced surface runoff of the catchment 278 remarkably compared with the initial plan of the urbanization in the case study. 279 In the next step, it is essential to discuss on the different aspects of the results and the proposed 280 framework. Table 3 displays computed measurement indices regarding the performance of the 281 optimization model. RMSEs for the initial plan and the optimal plan are 4.98 and 0.66 respectively. It 282 seems that 4.98 L/S is not a significant discharge. However, it should be noted that RMSE might not be a 283 good index in our case study. Because, outflow is low in many time steps. Hence, vulnerability index 284 should be noticed in the cases such as our case study. It might be possible RMSE is an appropriate index 285 in other case studies. Thus, we do not recommend excluding RMSE as a measurement index in the future 286 studies. Vulnerability indices for the initial plan and the optimal plan are 71.66% and 5.75% respectively. 287 In fact, vulnerability index demonstrates how the proposed optimization model is able to mitigate 288 physical habitat impacts of the stormwater in the simulated catchment. Vulnerability index in the initial 289 plan indicates that it might be very harmful for the aquatic habitats. Because, maximum physical habitat 290 loss has increased significantly. In fact, it might increase required energy for the fish to swim to the 291 upstream of the river. It might be problematic for the fish to reproduce and search for food when physical 292 habitat loss is high. Conversely, optimal plan is able to decrease physical habitat loss significantly. 293 Because, vulnerability index is 6% approximately that indicates increasing of the physical habitat loss is 294 not remarkable. A point should be noted regarding the optimal plan. It is able to reduce physical habitat 295 loss and increase urban area simultaneously. In fact, results indicate that 56% for the urbanizing area is 296 more than half of the initial plan that was 90% of the total area. In other words, it might reduce 297 negotiation between regional government and environmental advocators regarding the development of the 298 urban area in the simulated catchment. Hence, we recommend utilizing the proposed framework in the 299 future studies to optimize area of the developing urban region in the catchments considering physical 300 habitat impacts by applying physical habitat loss modeling as an advanced method in this regard 301 It is necessary to discuss on the different aspects of the proposed framework. First, we should discuss on 302 why the proposed method or mechanism might be appropriate for the practical projects. Using an ANFIS 303 based model to simulate runoff in the catchment and sub-catchment scale is possible to apply the runoff 304 model in the structure of the optimization algorithms. It should be noted available hydrodynamic and 305 hydrologic models to simulate surface runoff in the urban and non-urban catchments are not useable in 306 the structure of the optimization algorithm directly. In fact, in the conventional urban water management, 307 different scenarios might be simulated. Then, they will be analyzed in the structure of a decision-making 308 system. However, the proposed method provides a flexible environment to optimize urban area. 309 Moreover, our method considered the physical habitat impacts by application of one of the most advanced 310 methods in the assessment of physical habitats. It should be noted that the previous models lack the 311 physical habitat component in their structure that might be a significant drawback. Because, not only 312 would the physical habitat loss be important in the streams, but it might also be more sensitive compared 313 with water quality factors in some cases. Thus, the proposed method is able to cover this weakness point 314 for the future studies. Furthermore, using MOPSO is an advantage for the proposed framework. In fact, 315 the developed multi-objective optimization system is able to reduce concerns for the regional 316 governments and environmental advocators simultaneously. MOPSO is a robust algorithm that is able to 317 provide optimal solution for the problem properly. It should be noted limited number of the multi-318 objective algorithms have been developed in the literature. MOPSO is one of the robust algorithms that 319 has been applied in many engineering branches. It seems that simulation frameworks are not able to cover 320 complex needs for environmental engineering. In fact, there is a serious conflict between environment and 321 development that might be intensified in the future years due to increasing population. Hence, it is 322 required to use robust simulation-optimization frameworks. Using optimization systems might reduce

4-Conclusions 347
Present study proposed a novel framework to optimize the area of the developing urban area in a 348 catchment scale considering physical habitat impacts. We applied PSO-ANFIS framework to simulate 349 runoff. Testing process of the model was carried out in the sub-catchment and catchment scale. Moreover, 350 fuzzy physical habitat modeling as one of the novel methods was utilized to model the physical habitat 351 impacts in the structure of a multi-objective optimization. MOPSO was used to optimize the area of the 352 developing urban region. Based on the results in the case study, initial plan for the urbanization might 353 damage physical habitat at downstream river drastically. However, the optimal plan proposed by the new 354 method is able to minimize impacts of the physical habitat loss. Initial plan considered to increase the 355 urban area to 90% of the total area. However, the optimal plan proposed 56% as the optimal area for the 356 urbanization considering environmental impacts in the physical habitats. 357

Conflicts of interest/Competing interests 361
None 362

Availability of data and material 363
Some or all data used are available from the corresponding author by request. 364

Code availability 365
Code is available. However, it is not free of charge 366

Authors' contributions 367
Methodology, calculations and draft version by the first author. Field studies and review by the  Land use, location of the Rajaei reservoir and river network map of Tajan basin Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors. NWUA curve at the downstream river ecosystem of the simulated catchment based on the output of the physical habitat simulation  Direct response by the MOPSO Figure 12 NWUA in the current condition, initial plan of the urbanization and the optimal plan of the urbanization Figure 13 Out ows in the current condition, initial plan of the urbanization and the optimal plan of the urbanization