ferencz_et_al_2024_ERL

Urban morphology and urban water demand: A case study in the land constrained Los Angeles region using urban growth modeling

Stephen B. Ferencz ¹, Jim Yoon ¹, Johanna Capone ², Ryan A. McManamay ³

1. Earth Systems Science Division, Pacific Northwest National Laboratory, Richland, WA, USA
2. Virginia Tech, Blacksburg, VA, USA
3. Department of Environmental Science, Baylor University, Waco, TX, USA

Abstract: The interactions between population growth, urban morphology, and water demand have important implications for water resources and supply in urban regions. Water use for irrigation comprises a significant fraction of urban water demand, and is potentially influenced by long-term changes in urban morphology. To investigate this, we used spatially explicit projections of urban land development intensity (fraction impervious area) generated from a 30-m resolution urban growth model for the Los Angeles region. Recent historical data on water use and high resolution landcover were used to establish relationships between green area, urban development intensity, and outdoor water demand. These relationships were then used to project outdoor and total water demand in 2100 using the urban growth model outputs. We considered two different population scenarios informed by the Shared Socioeconomic Pathway (SSP) projections for the region (SSP3 and SSP5), and three scenarios of urban development intensification. Our analysis is resolved for over 80 water providers in the region, from the urban core to suburban fringe, and highlights diverse demand responses influenced by initial urban form and water demand attributes. Assumptions about outdoor water use factors based on recent water supply data were found to be nearly as influential on future outdoor demand as the urban growth scenario settings. Compared to previous studies, our work is unique in coherently linking high resolution SSP population scenarios, urban land cover evolution, and urban water demand projections, demonstrating the approach for the Los Angeles region – the largest population center in the western United States.

Data Sources

Input Data

1. NLCD historical urban land class data. Multi-Resolution Land Characteristics Consortium. https://www.mrlc.gov/viewer/. Accessed 10/2/23
2. Hi res land cover. Coleman, Red Willow (2020), “Southern California 60-cm Urban Land Cover Classification ”, Mendeley Data, V1, doi: 10.17632/zykyrtg36g.1
3. Monthly water provider data. California State Water Resources Control Board. Water Conservation and Production Reports. Monthly water supply data from 2014 to 2022. https://www.waterboards.ca.gov/water_issues/programs/conservation_portal/conservation_reporting.html accessed 6/2/23
4. Annual sectoral water provider data. California State Water Resources Control Board. DWR Urban Water Use Objective Analyzer Tool. https://lab.data.ca.gov/dataset/dwr-urban-water-use-objective-analyzer-tool accessed 6/2/23
5. Water provider boundaries. California State Water Resources Control Board. Service area boundaries of drinking water service providers, as verified by the Division of Drinking Water. https://gispublic.waterboards.ca.gov/portal/home/item.html?id=fbba842bf134497c9d611ad506ec48cc accessed 6/2/23
6. Population projections. Zoraghein, H., & O'Neill, B. (2020). Data Supplement: U.S. state-level projections of the spatial distribution of population consistent with Shared Socioeconomic Pathways. (v0.1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.3756179
7. Urban growth model projections. McManamay, R., & Vernon, C. (2023). High-resolution (30-m) urban land cover projections for Los Angeles California Urban Area: 2010 to 2100 under SSP5 (Version v1) [Data set]. MSD-LIVE Data Repository. https://doi.org/10.57931/19088234

Output Data

8. Ferencz, S., Capone, J., Yoon, J., & McManamay, R. (2024). Urban morphology and urban water demand evolution in the Los Angeles region (Version v2) [Data set]. MSD-LIVE Data Repository. https://doi.org/10.57931/2482052

Code Reference

9. Code for Executing the processing and analysis steps in "Reproduce my Experiment" provided in the workflow folder on this meta-repository. The scripts are named by their corresponding "Reproduce my Experiment" step. Ferencz, S., Yoon, J., Capone, J., & McManamay, R. (2024). Ferencz et al. 2024 ERL (Version V1.2). Zenodo. https://10.5281/zenodo.11521847

Contributing Modeling Software

Standard Python Packages along with the Rasterio (https://rasterio.readthedocs.io/en/stable/) and Geopandas modules (https://geopandas.org/en/stable/).

Reproduce my Experiment

1a. Process the NLCD Historical Data:

• Download NLCD for study region from MLRC. Input Data [1]. This is NLCD_2019_Land_Cover_L48_20210604_clNjCWtDUmB6F5woFH6g.tiff in Output Data [8].
• Use NLCD_processing.py to derive the urban landcover attributes for each provider boundary defined by Input Data[5]
• Output is NLCD_LC_areas_historic.csv located in Output Data [8]

1b. Process Population Projection Data:

• Download 1 km² urban population projections for Californina from Input Data [6]
• Clip CA data to study region using QGIS.
• Downscale to 30 m² using QGIS built in function. Save downscaled rasters as .tiff files. These are in Output Data [8] with the format LA_SSPX_urban_YYYY.tiff, where X = "3" or "5" and YYYY = year (e.g. 2100).
• The Python script Data_Processing_Urban_growth_projection_rasters.py located in workflow uses the downscaled population rasters and provider boundaries (Input Data[5]) to calculate the projected population within each water provider region. Output is two .csv files, one for each SSP scenario: SSP3_Aggregated_landclass_projection_data.csv and SSP5_Aggregated_landclass_projection_data.csv

1c. Process urban landcover projections:

• Download 30 m urban landcover projections from Input Data [7].
• Python script Data_Processing_Urban_growth_projection_rasters.py aggregates the urban land projections for each water provider. The script generates four .csv files for each SSP and zoning scenario ("low," "medium," "high"). These are located in Output Data [8]. The Python script then aggregates the SSP-specific outputs into single .csv files with the names: SSP3_Aggregated_landclass_projection_data.csv and `SSP5_Aggregated_landclass_projection_data.csv'

1d. Generate landcover rasters for each urban land class (21, 22, 23, 24) within each water provider boundary:

• Use the QGIS model builder function Urban_landclass_landcover_metrics.model3 to process the hi-resolution 60 cm landcover data Input Data [2]. To access the model builder GUI open QGIS and navigate to Processing -> Model Designer -> Model -> Open Model. Inputs are the NLCD recent historical land classification raster (Input Data [1]), the hi-resolution landcover raster from Input Data [2], and the provider boundaries (Input Data [5]. Outputs are four .tiff rasters for each water provider region (or sub-region), one for each NLCD urban land class. This processs is very time consuming so we provide the output rasters in the Folder "Clipped Provider High Res Landcover" (Output Data [8]). The naming convention is PROVIDER_NAME_LC##.tiff where ## denotes the NLCD land classification (21, 22, 23, 24).

1e Calculate recent average monthly (2017-2021), minimum (2014 to 2021), maximum (2014-2021) monthly water demand factors for each water provider:

• Download monthly water data by water provider for all of California (Input Data [3]).
• Filter out providers that are outside of the study area (done manually).
• Python script Demand_data_processing_2014_2021_data.py converts monthly demand to common units (acre-feet, 1 acre-foot = 1,233 m³ and then calculates average monthly demands for each water provider. The output is Provider_historical_demands.csv. These outputs are used for Step 4 and Figure 1 of the paper.
• Python script Demand_data_processing_2017_2021_data.py converts monthly demand to common units (acre-feet, 1 acre-foot = 1,233 m³ and then calculates minimum and maximum monthly demands for each water provider over the 2014 to 2021 period. The outputs are Provider_historical_demands_min.csv and Provider_historical_demands_max.csv.
• These outputs are used for Step 4.

2. Derive NLCD urban land classification -> land cover relationships for each water provider region:

• Python scripts processes all of the clipped landcover data produced in Step 1d located in workflow: Data_Processing_Urban_LC_green_fraction_by_service_region. The outputs are three sets of .csv files: landclass_area_providers.csv, landcover_area_providers.csv, and landcover_fraction_providers.csv. These outputs are used in Step 4 and for Figure 2 of the paper.

3. Analyze pixel-level urban intensification and extensification

• Use Python script Urban_growth_change_mapping.py. Inputs are initial and final urban morphology rasters. Set SSP and scenario to process on Lines 32-33. Outputs are two tiff files. Urban_intensification_SSP_scenario.tiff shows which pixels had an increase in urban LC and the amount of the increase (1, 2, or 3 levels). Urban_growth_SSP_scenario.tiff shows what pixels were converted to urban land. SSP = (SSP3 or SSP5) and scenario = "low", "med", or "hi" zoning. These outputs are used for Figure 3 of the paper.

4 Generate future demand projections:

• Python script Future_demand_landcover_evolution.py generates future decadal demand projections using unit dempand factors based on average monthly water supply over 2017-2021 from Step 1e. Set SSP (3 or 5) and zoning scenario (low, med, hi) in Lines 10-11 of script. Script will output indoor and outdoor demands for each provider and also landcover and irrigation depth estimates. Run for SSP5 low, SSP5 med, SSP5 hi, and SSP3 med. Outputs are used by the plotting scripts associated with Figures 4 through 8 of the paper. To generate outputs used for Figire 8, uncomment lines associated with demand_sensitivity_2100_attributes or master_demand_sensitivity_2100_attributes. Modify Line 340-341 to export decadal monthly indoor and outdoor projections used for Figure 5.
• Python script Future_demand_landcover_evolution_min_and_max_scenarios.py generates future decadal demand projections used on unit demand factors based on minimum and maximum monthly water supply factors over 2014-2021 from Step 1e. Script will output minimum and maximum indoor and outdoor demand projections for each provider and also landcover and irrigation depth estimates. Run for SSP5 low, SSP5 med, SSP5 hi, and SSP3 med by modifying Lines 10-11 of script and specific whether to use the minimum or maximum demand factors by modifying Line 42 to set min or max. Outputs are used by the plotting scripts associated with Figures 4 and 5. Modify Line 346-347 to export decadal monthly indoor and outdoor projections used for Figure 5.

Reproduce my Figures

• Figure 1. Subplots a, b, d, e, f made in QGIS using Input Data [1][2][5] and touched up using Inkscape. Subplot c plotted from Figure_1.py.
• Figure 2. Wire diagram made in Powerpoint. Subplots b and c Figure_2.py. Subplot d visualized in QGIS using Input Data [7].
• Figure 3. Output tiff files from Step 3 visualized in QGIS.
• Figure 4. Outputs from Step 4 and Step 1c. Figure_4_and_5_urban_LC_areas.py & Figure_4_and_5_water_demand_plotting.py. Bar plot outputs formatted in Powerpoint. Outputs from the plotting scripts formatted in Powerpoint.
• Figure 5. Outputs from Step 4 and Step 1c. Figure_4_and_5_urban_LC_areas.py & Figure_4_and_5_water_demand_plotting.py. Bar plot outputs formatted in Powerpoint. Outputs from the plotting scripts formatted in Powerpoint.
• Figure 6. Outputs from Step 4. Figure_6.py. Outputs from the plotting scripts formatted in Powerpoint.
• Figure 7. Outputs from Step 4. Figure_7.py.
• Figure 8. Outputs from Step 4. Figure_8.py.