Remote sensing-based classification of floodplain vegetation along the Naryn River, Kyrgyzstan

The remote sensing model ‘Google4Habitat’ (Egger et al. 2024) enables large-scale vegetation classification based on freely available satellite imagery in combination with ground truth data. In a current project, it was further developed to predict hydrogeomorphology-driven vegetation succession phases at the Naryn River in Kyrgyzstan. This river has high hydrologic and morphologic dynamics over its entire length of 600 km (Betz et al. 2023). Additionally, drought stress due to the continental climate of Central Asia causes specific succession pathways (Lauermann et al. 2024). This makes the Naryn an interesting case study for demonstrating the capabilities of the ‘Google4Habitat’ model for analyzing the spatial pattern of vegetation communities and their biogeomorphic implications.

The ‘Google4Habitat’ model runs on the Google Earth Engine platform. By using globally available satellite data together with recorded reference points for each classified habitat type, it performs a spatial and temporal analysis for each habitat. Specifically, Sentinel-2 scenes from one year are selected and monthly spectral values and vegetation indices are calculated. Orthophotos are integrated and texture parameters are added. Pixel values at training points are used to create the classification model (358 collected in the field, supplemented by 987 points from aerial image interpretation), which is then validated, evaluated, and used to create the classification map of the study area. To obtain precisely delineated habitat maps, a segmentation of habitat structures is carried out based on a high resolution orthophoto. Then, the classification map is intersected with the segments and final habitat types are assigned using zonal statistics. The habitat map is intersected with the three succession series delineated on the orthophoto and area balances are created.

On the Naryn River, nine different habitat types were obtained using this supervised classification approach. The freshly to moderately dry sites of the sedimentation series are characterized by pioneer vegetation located within the active channel, Salix-Myricaria shrub, and softwood riparian forests of varying ages with Populus talassica, as well as extensive Hippophae shrubs. The shallow sites of the aggradation series are, however, much more sparsely vegetated than the sedimentation series and are characterized by the presence of Tamarix ramosissima. Both series end with an Artemisia steppe. Finally, the siltation series is characteristic of the former side arms that are in the process of silting up and are particularly dominated by Typha minima and Phragmites australis.

References

Betz, F., Lauermann, M., & Egger, G. (2023). Biogeomorphology from space: Analyzing the dynamic interactions between hydromorphology and vegetation along the Naryn River in Kyrgyzstan based on dense satellite time series. Remote Sensing of Environment, 299, 113890.

Egger, G., Preinstorfer, S., Kollmann, M., Becker, I., Izquierdo-Verdiguier, E. & Paul, P. (2024). Google4Habitat - a novel method for remote sensing-based habitat classification using Google Earth Engine. Carinthia II, 3/1, 8–28.

Lauermann, M., Heckmann, T., Poeppl, R., Egger, G., Eichel, J. & Betz, F. (2024). How does Hydrological Connectivity Influence Fluvial Biogeomorphic Succession in Semiarid Regions? A Case Study from the Naryn River in Kyrgyzstan, Central Asia. AGU Fall Meeting 2024, Poster No. 1302, EP43C-1302.