Summer school | Mysite

GEOBIA SUMMER SCHOOL 2020

Project: Spatio-Temporal Dynamics of Urban Green Space and Health Assessment a case of Vienna City

GEOBIA Summer School was planned took place in early July, embedded in the GEOBIA 2020 event during the Salzburg GI Week, jointly organized by UNIGIS Salzburg and within the public outreach and education activities of the CopHub.AC project.

As part of the summer school I participated in a project in Assessment of the Spatio-Temporal Dynamics of Urban Green Space and Health in Vienna City.

INTERNSHIP

I'm a paragraph. Click here to add your own text and edit me. It's easy.

Introduction

Urban areas facing pressure due to urbanization leading to limited resources and growing impacts of climate change. Urban living limits the people’s access to nature. This could cause exposure to hazards such as air and noise pollution. Green spaces and other nature-based solutions offer innovative approaches to increase the quality of urban areas and enhance local resilience. This project aims to answer the research questions as follows:

Research questions

•What is the trend of the green spaces over the last five years with increasing urbanization - are they increasing or decreasing?

•What is the trend of vegetation condition during this period?

•Are green spaces equally spatially distributed across Vienna?

Data

The datasets used for this project include:

1.Sentinel 1 data for the year 2015 and 2020

GRD product, IW acquisition mode, DESCENDING

Corresponding with time of S2 acquisition (+/- few days)

2.Sentinel 2 data for the year 2015 and 2020

Methodology

Figure 1. Workflow

Multi-resolution segmentation

In order to perform object-based classification, multi resolution segmentation was performed by use of e-cognition software. The segmentation performed on Sentinel 1 using the vh_db & vv_db polarization for Sentinel 2 bands 2, 3, 4, and 8 were used. The parameters used for segmentation include:

Scale parameter: 2, Shape: 0,1 and Compactness: 0,5.

Classification

Random Forest(RF) and Support Vector Machine(SVM) algorithms were used to classify three classes that is water, urban areas and green areas. The green area maps for the years 2015 and 2020 were used to estimate the change detection as shown in figure 2. The classification maps for 2020 were validated by use of land cover maps for the year 2018. The accuracy for SVM and RF are shown in figure 2. Figure 3 shows the changes of of green areas for Vienna city, it was estimated that 24.56% of urban green space reduced between 2015 and 2020 in Vienna city.

Figure 3. Classification results for Random Forest and Support Vector Machine

Figure 4. Green areas that changed in Vienna

Pre-processing

The sentinel 1 data was pre-processed by use of

Apply-Orbit-File, Subset, Calibration, ThermalNoiseRemoval, Speckle-Filter, and Terrain-Correction by use of SNAP software. For the case of sentinel 2 data the processing steps are such as

Atmospheric correction for older image by use of (Sen2Cor)

Multi-size mosaicking-Subsetting-Resampling in (SNAP) and finally

Conversion from .img raster format to .tif format (QGIS)

Figure 2. Districts served with green areas

The areas served by green areas was estimates per district level in Vienna. The map shows that on the left side of the map districts had high proportion of green areas compared to the center.

Figure 5. Vegetation health monitoring of green areas in Vienna

To check the trend of vegetation health in Vienna google earth engine was used to monitor the trend Of NDVI figure 5 shows the methodology used. An interactive platform was created in google earth engine to view the NDVI changes for different months over the years.

Discussion

•Data acquisition: suitable satellite imagery to avoid preprocessing (e.g. atmospheric corrections) respectively different input data

•Validation/Training of data/analyses through ground truthing

•Validation of 2020 land cover classification on 2018 land use data

•Difference in time of the year satellite imagery was acquired for the two years à has an influence on NDVI and thus potentially on identification of urban green areas

•No information on plant species à has influence on NDVI

•Variations of climate from year to year not taken into account when assessing vegetation health

Conclusion

The case study shows an efficient method for monitoring cities in order to provide a healthy and sustainable environment and thus improve the well-being of urban residents