Skip to main content

Table 4 Examples of application of regression methods towards SDGs using EO data

From: Monitoring sustainable development by means of earth observation data and machine learning: a review

SDGs

Field

Main finding

References

SDG 2 (Zero Hunger)

Agriculture

Results increased the potential of using Sentinel-2 to obtain cotton Leaf Area Index and comparison of methods showed that the Gradient Boosting RT was the best

[100]

Estimate the crop yield, at a pixel level, using ML proved to be an accurate approach

[73]

The results obtained from the comparison of methods showed that Boosted RT was the best to predict maize yield

[101]

SDG 3 (Good Health and Well-Being)

Spread of diseases

By mapping the relationship between EO variables and vector population, the proposed RF Regression methodology was able to predict the temporal distribution of yellow fever mosquito populations

[102]

SDG 6 (Clean Water and Sanitation)

Water quality

Landsat 7 images are a solid option for assessing water quality characteristics

[55]

SDG 7 (Affordable and Clean Energy)

Renewable energy sources

During Spring and Autumn is harder to predict the hourly solar irradiation compared to Winter and Summer

[103]

SDG 9 (Industry, Innovation and Infrastructure)

Pollution

RT effectively estimates carbon dynamics and allowed the analysis of its impacts on meteorology and vegetation

[49]

The improved GPR had a high accuracy compared to the original GPR and other methods predicting the CO2 emissions

[104]

SDG 11 (Sustainable Cities and Communities)

Land cover

RF Regression was very accurate (96%) in delineating house-attached, semi-public and public green spaces

[105]

SDG 13 (Climate Action)

Drought

The use of ML to acquire the Normalised Microwave Reflection Index is an effective way to monitor the variation of vegetation water content to predict droughts

[106]

SDG 14 (Life Below Water)

Freshwater habitat

Geographically Weighted Regression technique was accurate in the estimation of stream bathymetry of water with a depth less than 1 m

[107]

SGD 15 (Life on Land)

Terrestrial ecosystem

The best performance, to obtain the latent heat evaporation using a small dataset, was achieved by Kernel Ridge Regression, and using a large dataset, was achieved by Bagging RT

[108]

Grassland

Vegetation indices acquired from Sentinel 2 have high potential concerning grasslands productivity, management, monitoring and conservation

[109]

Landslide

Catchment map units and model selection are crucial for the performance of landslide susceptibility maps

[110]