Drivers and extent of surface water occurrence in the Selenga River Delta, Russia

Abstract

Study region: Selenga River Delta (SRD), Russia.

Study focus: How is water occurrence changing in the SRD, and what are the hydroclimatic drivers behind these changes? The presence of water on the surface in river deltas is governed by land use, geomorphology, and the flux of water to and from the Delta. We trained an accurate image classification of the Landsat satellite imagery during the last 33 years to quantify surface water occurrence and its changes in the SRD. After comparing our estimations with global-scale datasets, we determined the hydrological drivers of these changes.

New hydrological insights for the region: We find mild decreases in water occurrence in 51% of the SRD's surface area from 1987-2002 to 2003-2020. Water occurrence in the most affected areas decreased by 20% and in the most water-gaining areas increased by 10%. We find a significant relationship between water occurrence and runoff (R² = 0.56) that does not exist between water occurrence and suspended sediment concentration (SSC), Lake Baikal's water level, and potential evapotranspiration. The time series of water occurrence follows the peaks in the runoff but not its long-term trend. However, the extremes in SSC do not influence surface water occurrence (R² < 0.1), although their long-term trends are similar. Contrary to expected, we find that the Delta has a relatively stable long-term water availability for the time being.

Keywords: Selenga River Delta; Supervised classification; Surface water occurrence.

Fig. 1.

The Selenga River Delta. (a) The Selenga River Delta location in Russia over a false-color Landsat 8 image acquired on 23 June 2017, path 132, frame 24. Water (blue), vegetation is shown by values in Near-Infrared (NIR) and urban areas are in Short-Wave Infrared (SWIR2) range; (b) Discharge measuring stations (red triangles) and extent of the high-resolution land cover classification (solid black-line border) available by Berhane et al. (2018) used here to validate the results of water occurrence. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2.

Five-year moving averages of (a) monthly runoff (R), precipitation (P), and (b) temperature (T) in the SRD (For the runoff calculations, we divided discharge data at station Mostovoi by the upstream hydrological basin of 440,200 km², of which 67% falls in Mongolia and 33% in Russia.) See Section 2.5 for sources.

Fig. 3.

Land cover (CGLS-LC100) and three 10-km-wide (at their center) focus regions in the Delta (D1, D2, and D3). Point A is near the bifurcation of the river that we consider the head of the Delta here.

Fig. 4.

Mean surface water occurrence $(\overline{w})$ in 1987–2020 with pixels not holding water in any class images (0%; yellow) and holding water in all class images (100%; dark blue). The arrows in zoom panel m2 show river bends and flood-prone areas with high water occurrence, yet less than permanent water bodies. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5.

Change in surface water occurrence $(\overline{\Delta w})$ between 1987–2002 and 2003–2020. Red areas (−1) show losses of water surface, and blue areas (+1). The arrow in zoom panel m1 show a change in flow direction from the NE (red, indicating change) to the NW (blue, indicating the increase in surface water occurrence). We masked out pixels with no change between two periods $(\overline{\Delta w}=0)$ .. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 6.

Pixel distributions of (a) mean surface water occurrence $\overline{w}$, (b) and its change $\overline{\Delta w}$ by land cover and (c) $\overline{\Delta w}$ by general categories (i.e. permanent water, seasonal water, and land). Crosses are the outliers. Abbreviations: Permanent Water bodies (PW), Bare Land and Sediments (Sand Bar), Open Forest Not matched with any classes (OFN), Closed Forest Not matched with any classes (CFN), Herbaceous vegetation (Herb), Wetland herbaceous (Wetland), Closed Forest Deciduous Needle Leaf (CFDNL), Open Forest Deciduous Needle Leaf (OFDNL).

Fig. 7.

Change from water to land and vice versa for this study, Pekel et al. (2016), and Donchyts et al. (2016) in D2 region (mid-delta) shown in Fig. 3.

Fig. 8.

Change from water to land and vice versa in m1 and m2 regions shown on Fig. 4; (a–c) this study, Pekel et al. (2016), and Donchyts et al. (2016) respectively in m1 and; (d–f) in m2.

Fig. 9.

The monthly change in water occurrence $\left(\Delta w_m\right)$ for non-zero values and runoff in station Kabansk (ΔR) between the periods 1987–2002 and 2003–2020. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 10.

The relationship between ${\overline{\mathit{w}}}_{\mathit{s}}$ (left vertical axis) and (a) runoff, (b) suspended sediment concentration (SSC), (c) water level (altimetry and Babushkin gauging station), and (d) potential evapotranspiration (PET) in the mid-Delta. A Loess filter smoothes all data. The surface of reference for the gauge stations is the sea level and for the altimetric water levels is geoid GGMO2C (Tapley et al., 2005).