This repository consists of the code base for the web tool developed under the publication:
The webtool can be found free of charge at: https://water-reuse-map.herokuapp.com/
If you wish to use the tool for your own analysis please provide proper attribution by citing the work: Kavvada et al
The core and integral criterion for sustainable non-potable reuse (NPR) implantation is the issue of system scale. Larger facilities can benefit from the increased "economies of scale" in treatment but in the context of water reuse a tradeoff occurs as in a centralized system the water would be further away from the point of demand which exacerbates the upgradient conveyance and distribution impacts. This project defines a modeling method for decentralized water reuse systems to estimate their economic and environmental impacts under real topographical and demographic conditions. The developed tool estimates both the financial cost and environmental impacts of NPR as a function of treatment scale and optimized conveyance networks. It is based on a heuristic modeling approach using geospatial algorithms to determine the optimal degree of decentralization. The webtool is developed to assess and visualize alternative NPR system designs considering topography, economies of scale and building size.
The developed tool is currently based on the data of the city of San Francisco. However, it is a generalizable tool as long as the required input data is given. See input data section for more details.
The main input data required for the tool to run is a csv with building data of the area of interest. An example dataset can be found under input_data\combined_buildings_2.csv.
The required attributes in the csv file are:
- the lat, lon coordinates of each buildings (x_lon, y_lat),
- a combined coordinates field in the form of
(lat, lon), - the building footprint area (Area_m2),
- the number of floors (num_floor),
- the building base elevation (ELEV_treat),
- the residential population of the building (SUM_pop_residential) and
- the commercial population of the building (SUM_pop_commercial)
The residential and commercial population attributes are shown how they can be calculated in the example jupyter notebooks found in jupyter_notebooks folder.
Other inputs in the model that can be customized are located in the Parameters.py. This file includes default parameters that are used in the analysis but can be overridden if better data of the specific location are known.
The entire algorithmic process is included in the optimization2.py file. The basic functions used in calculations are located in the functions.py file. To account for the spatially sensitive conditions and for the model to be able to assess multiple levels of decentralization, the developed algorithmic model expands its size through an iterative process with an input of the existing building location and characteristics. The model is capable of assessing different metrics of water reuse, namely economic cost, energy intensity and GHG emissions as defined by the user. As the model is capable of assessing different metrics each of the metric in question will be called "impact" from now on.
The algorithm involves the following steps:
- All the input buildings are loaded and a KD tree is constructed from their coordinate locations. The KD tree is used for optimal searching of the 500 closest buildings to the user query point and listed in ascending order based on the euclidean distance from the user query point. The 500 is an arbitrarely chosen value to provide a cutoff point for the analysis. It could be altered as see fit.
- The building located the closest to the point of interest is considered as the first building to enter the algorithm.
- Impact calculation: Given this starting point the impact of water reuse is calculated with only this building in mind. This involves estimating the impact of treating the required volume of water (given the population of the building), the embodied impact for treatment (includes construction and transportation), the in-building impact for delivering the recycled water to all the building floors and finally the in-between building impact for sharing recycled water in the case of multiple buildings (in this first case this is zero). These operations are defined under the
metric == impactoperation in the module. - After this first assessment, the second in-line building is considered. This building is added to a Graph along with the previous building. This system is now considered a cluster. Given this cluster the distance between the two buildings is calculated.
- The process for impact calculation is repeated similarly as before but now considering the new cluster of buildings. Given this new cluster a new impact is defined.
- This new impact is compared to the previous impact of the cluster. If the new impact is smaller than the previous then the new building is kept in the cluster as it is beneficial for the overall systems performance. If not, then the building is discarded and the previous system's performance is kept.
- The process continues until all close (k=500) buildings are assessed.
The output of the algorithmic process as described previously is a csv file located at the specified path when calling the model (see Standalone Modeling section below for application). The csv includes the locations of all the buildings in order of assessment along with information of the population served, the total impact of interest, and the breakdown of the total impact of interest to the pumping, treatment operational, treatment embodied and piping. It also outputs a json file with all the building locations and a tag of whether they are selected or not (accept = 'yes' or 'no') along with the total population served and the number of buildings for immediate visualization in the webtool.
The webtool is connected to the python algorithm described previously through Flask. Details on the server part can be found on the server.py file. As it is a javascript based webtool the web development parameters are specified in the static\index.html file. The user can interact with the webpage and create certain events that are logged and passed to the python module. Specifically, the user can:
- Click on the map, to given the signal of the query point location, from where the algorithmic process is going to initiate.
- Click on a metric, "cost", "energy", "GHG" to define the metric of interest.
- Set the treatment equation parameters as floats in the corresponding boxes.
- Set the direct GHG emissions for treatment as a float in the corresponding box.
- Change the visualization by making the background map satellite, road or greyscale.
The outputs of the algorithmic process are going to appear on the screen after the computation is done. The outputs consist of the number of the total number of houses served and the served population and a visualization on the map of the selected and discarded buildings assessed.
The model can be run as a standalone python model for scenario planning and multiple iterations for analysis. This can be done by simply calling the getServiceArea function that is located inside the optimization2.py file. The required inputs for the function to run are:
- the query point, which is the lat, lon location of the point of interest. Input that as a tuple (lat, lon)
- the destination path, which will be used for saving the csv
- a string of the metric of interest, this could be either "cost", "energy" or "GHG"
- four float parameters a, b, c, d, which correspond to the treatment energy equation constant (the equation is modeled as a polynomial) and
- a float for the direct GHG emissions of the treatment. This is zero as a default.
To run a local instance of the webtool:
- Clone the repo
- Navigate your command prompt inside the repo
- Start the server by running
FLASK_DEBUG=1 FLASK_APP=server.py flask run - Navigate to
localhost:5000 - and Done! Easy!