Forest Extent
Back to the Monitoring Hub
Problem: Long-term overgrazing by sheep has hindered the regeneration of new trees in this area. Furthermore, the construction of the Dali dam disrupted natural flood patterns essential for floodplain forest species to thrive, resulting in a gradual loss of floodplain forest extent and condition, and reduced rates of natural regeneration.
Action: Fences have been erected to exclude sheep, and controlled water releases from the Dali dam were implemented, to encourage the natural regeneration of the floodplain forest tree species, which require or prefer moist soil to thrive. Lastly, gabions are being built to retain water downstream of the dam when water is released, which should benefit the floodplain forest.
Indicator: An increase in the area of floodplain forest will indicate a reversal of previous trends and that the restoration actions have been sufficient for restoring the floodplain forest extent.
On-site monitoring: assessing tree regeneration in the floodplain forest below the Dali dam. Tape measures are used to measure out plots for collecting on-the-ground data. Photo: Vasil Metreveli.
Methods
This indicator continues monitoring of forest extent established by the Iori River Valley project. Sentinel 2 satellite imagery was obtained for the floodplain forest areas both above and below the dam, corresponding to the control and treatment areas, respectively. These images were analysed to determine the extent of the floodplain forest as a baseline, or pre-intervention, dataset in 2019 (when the restoration measures were initiated). This was compared to images from 2022 and 2023, with the aim of evaluating any short-term changes resulting from the restoration efforts, which began in 2020. Satellite imagery also exists as far back as 1978, to provide a historic baseline comparison. The classification of forested regions was accomplished using random forest models, validated through on-site fieldwork conducted in sample plots. During these field studies, researchers verified the presence or absence of trees and integrated this reference data into the classification model to enhance its accuracy.
Results
A time series analysis revealed that the forest regions both above and below the dam exhibit temporal fluctuations (Fig 1), and it was too early to detect any significant long-term patterns. Observationally it appeared that forest extent decreased both above and below the dam between 1978 and 1990, coinciding with the commencement of operations at the Dali reservoir. While above the dam the control area appeared to show some increase in extent after 2000, treatment areas below the dam did not yet show recovery. This data will be reanalysed annually, and over time we will see if SABUKO’s efforts gradually increase forest extent below the dam.
Figure 1: Changes in forest extent at sites above and below the Dali dam since 1978. Error bars represent OOB (out-of-bag) estimates of error rate for models, a method used to estimate the prediction error of random forest models without requiring a separate validation dataset.
Partners
