| Autor: |
Veronica Barraza, Esteban Roitberg, Francisco Grings |
| Rok vydání: |
2020 |
| Předmět: |
|
| Zdroj: |
2020 IEEE Congreso Bienal de Argentina (ARGENCON). |
| DOI: |
10.1109/argencon49523.2020.9505491 |
| Popis: |
The Dry Chaco region (DCF) has the highest absolute deforestation rates in Argentina. In this study, we evaluated the performance of the Extreme Gradient Boosting (XGboost) algorithm to detect breakpoint in MODIS EVI time series. The model takes as input 2-year long "time series segments". This process is usually seen as a breakpoint in the MODIS EVI time series, associated with the change from a forest phenology to something else (e.g. pasture, cropland). The model was validated using in situ data derived from high resolution images (Landsat 7 and 8). We also compared the performance of XGBoost to traditional machine learning models using classification. Finally, we evaluated feature importance using SHAP. This analysis provided evidence of how different sections of the segment would help to identify deforested pixels. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
