Santiago.jl

versionBuild StatusCoverage Status

Santiago (SANitation sysTem Alternative GeneratOr) is a Julia package to generate appropriate sanitation system options. It is able to

  • find all possible systems given a set of sanitation technologies;
  • assess the appropriateness of a technology in a given context;
  • assess the overall appropriateness of a sanitation system in a given context;
  • calculate (optionally with uncertainly quantification) the massflows for each system for total phosphor, total nitrogen, totalsolids, and water;
  • select a desired number of diverse but appropriate systems.

Installation

  1. Install Julia version >= 1.4.

  2. Install the Santiago package from the Julia prompt:

] add Santiago
  1. To edit Julia files you may also want to install Visual Studio Code and its Julia Extension. Alternatively, see the Julia home page for support for other editors.

Usage

The example below demonstrates the typical steps needed to identify sanitation systems appropriate for a given case. See the references below for a clarification of the terminology and the recommended embedding in the strategic planning process.

Most functions have a documentation string attached that can be accessed with ?functionname on the Julia prompt.

For reproducibility it is a good idea to create a separate Julia project (similar like virtualenv in Python) for every analysis, see here.

Minimal Example

using Santiago

# -----------
# 1) Import technologies

# we use the test data that come with the package
input_tech_file = joinpath(pkgdir(Santiago), "test/example_techs.json")
input_case_file = joinpath(pkgdir(Santiago), "test/example_case.json")

sources, additional_sources, techs = import_technologies(input_tech_file, input_case_file)

# -----------
# 2) Build all systems

allSys = build_systems(sources, techs);

# number of found systems
length(allSys)

# -----------
# 3) Calculate (or update) system properties

sysappscore!.(allSys)
connectivity!.(allSys)
ntechs!.(allSys)
template!.(allSys)

# see all properties of the first system
allSys[1].properties

# -----------
# 4) Mass flows

# Inputs for different sources in kg/year/person equivalent.
# See references below.
input_masses = Dict("Dry.toilet" => Dict("phosphor" => 0.548,
                                         "nitrogen" => 4.550,
                                         "totalsolids" => 32.12,
                                         "water" => 547.1),
                    "Pour.flush" => Dict("phosphor" => 0.548,
                                         "nitrogen" => 4.55,
                                         "totalsolids" => 32.12,
                                         "water" => 1277.1),
                    "Cistern.flush" => Dict("phosphor" => 0.548,
                                            "nitrogen" => 4.55,
                                            "totalsolids" => 32.12,
                                            "water" => 22447.1),
                    # Urine diversion dry toilet
                    "Uddt" => Dict("phosphor" => 0.548,
                                   "nitrogen" => 4.55,
                                   "totalsolids" => 32.12,
                                   "water" => 547.1)
                    )


# Calculate massflows with 20 Mont Carlo iterations (probably not enough)
# for all systems and save to system properties
massflow_summary_parallel!(allSys, input_masses, n=20);

# Alternatively, the non-parallelized version can be used:
# massflow_summary!.(allSys, Ref(input_masses), n=20);

# Examples how to extract results
allSys[2].properties["massflow_stats"]["entered"]
allSys[2].properties["massflow_stats"]["recovery_ratio"]
allSys[2].properties["massflow_stats"]["recovered"]

allSys[2].properties["massflow_stats"]["lost"][:,"air loss",:]
allSys[2].properties["massflow_stats"]["lost"][:,:,"mean"]
allSys[2].properties["massflow_stats"]["lost"][:,:,"q_0.5"]

# -----------
# 5) select a subset of systems

# For example, select eight systems for further investigation
selectedSys = select_systems(allSys, 8)

# -----------
# 6) write some properties in a DataFrame for further analysis

df = properties_dataframe(selectedSys,
                          massflow_selection = ["recovered | water | mean",
                                                "recovered | water | sd",
                                                "lost | water | air loss| q_0.5",
                                                "entered | water"])

size(df)
names(df)

# export as csv
import CSV  # the package 'CSV' needs to be installed separately
CSV.write("mysystems.csv", df)


# -----------
# 7) create a visualization of a system as pdf

# First write a dot file
dot_file(selectedSys[1], "system.dot")

# Then, convert it to pdf (The program `graphviz` must be installed on the system)
run(`dot -Tpdf system.dot -o system.pdf`)


# -----------
# 8) export to JSON

# Note, the JSON export is designed to interface other applications,
# but not for serialization.

open("system_export.json", "w") do f
    JSON3.write(f, selectedSys)
end

Input format

Typically the information on the case specification and the available technologies are provided via files. Santiago can only import JSON files. The structure must match these examples:

Many tools are available to browse and edit JSON files. For example, Firefox renders JSON files nicely, or Visual Studio allows for editing.

Logging

By default, Santiago prints only few information. This can be adapted by the logging level. With the package LoggingExtras.jl (needs to be installed extra) different logging levels can be used for the console output and the log file:

using Logging
using LoggingExtras

# - on console show only infos and errors, write everything in the logfile 'mylogfile.log'
mylogger = TeeLogger(
    MinLevelLogger(FileLogger("mylogfile.log"), Logging.Debug),  # logs to file
    MinLevelLogger(ConsoleLogger(), Logging.Info)                # logs to console
)
global_logger(mylogger)

... use Santiago functions ...

Update systems for a new case profile

The generation of all systems is computationally intense. The code below demonstrates how to first generate all systems without case information and later update the system scores with case data.

using Serialization

## 1) build systems without case information and cache result

sources, additional_sources, techs = import_technologies(tech_file)

if isfile("mycachfile.jls")
    allSys = deserialize("mycachfile.jls")
else
    allSys = build_systems(sources, techs)
    ...
    massflow_summary!.(allSys, Ref(input_masses), n=100);
    ...
    serialize("mycachfile.jls", allSys)
end

sysappscore!.(allSys) # all are '-1.0' because no case profile was defined yet

## 2) read case file and update sysappscore

tas, tas_components = appropriateness(tech_file, case_file);
update_appropriateness!(sources, tas)
update_appropriateness!(additional_sources, tas)
update_appropriateness!(techs, tas)

sysappscore!.(allSys)  # now we have the updated SAS

## 3) select systems

fewSys = select_systems(allSys, 6)

## 4) scale massflows for 100 people

fewSys = scale_massflows.(fewSys, 100)

The slowest parts are build_systems and massflow_summary!. Therefore we could cache the output as shown in this example. Steps 2 and 4 are fast and can be quickly adapted to new cases.

Multi-threading

The functions build_systems and especially massflow_summary_parallel! benefit from multi-threading. As this may involves some overhead, benchmarking is recommended. See the official documentation how to control the number of threads.

References

Spuhler, D., Scheidegger, A., Maurer, M., 2018. Generation of sanitation system options for urban planning considering novel technologies. Water Research 145, 259–278. https://doi.org/10.1016/j.watres.2018.08.021

Spuhler, D., Scheidegger, A., Maurer, M., submitted. Ex-ante quantification of nutrient, total solids, and water flows in sanitation systems. Journal of Environmental Management.

Spuhler, D., Scheidegger, A., Maurer, M., 2020. Comparative analysis of sanitation systems for resource recovery: influence of configurations and single technology components. Water Research 116281. https://doi.org/10.1016/j.watres.2020.116281

License

The Santiago.jl package is free software: you can redistribute it and/or modify it under the terms of the GNU Affero General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Affero General Public License for more details.

You should have received a copy of the GNU Affero General Public License along with this program. If not, see https://www.gnu.org/licenses/.

Copyright 2020, Eawag. Contact: Dorothee Spuhler, Dorothee.Spuhler@eawag.ch