Changes in floodplain hydrology following serial damming of the Tocantins River in the eastern A...

1. Study site

1.1. Our study site is an approximately 145-km stretch of the Tocantins River south (i.e., upstream) of its confluence with the Araguaia River. The study site extends north of the city of Miracema do Tocantins (−9.5591, −48.3798) to south of the city of Tupiratins (−8.3917, −48.1114, WGS 84; Fig. 1A). This area was chosen because it has five dams upstream so that cumulative impacts of dams could be studied. Also, continuous flow data collected by the Agência Nacional das Águas (ANA) during the study period was available for this section of the river but not others. The section of river has a floodplain of approximately 750 km2 as delineated by the GFPLAIN250m floodplain boundary product (Nardi et al., 2019), which is based on the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM).

2. Materials and methods

2.1. To quantify dam-induced changes to floodplain extent, hydroperiod, and inundation timing across the study region, we created daily floodplain inundation maps for the area.We then compared the flooding regime before and after each dam's operation,

2.1.1. it divides

2.1.2. Floodplain inundation estimation

2.1.2.1. Water level data came from two Agência Nacional das Águas (ANA) flow gauge stations located near Miracema do Tocantins (station ID #22500000) and Tupiratins (station ID #23100000; F. These two stationswere selected from117 candidate stations on the Tocantins River because theywere adjacent to each other on the river and had the most complete data from1985 to 2019. The Miracema do Tocantins station had 214 missing daily values over the 34-year period (<1.7%)while Tupiratins had 49 missing days (0.4%).

2.1.3. Field validation

2.1.3.1. software-provided relationships between pressure, temperature, and water level. Total pressure data were converted to water pressure using barometric pressure data collected by the Pedro Afonso InMET station (Instituto Nacional de Meteorologia; #82863) 23 km away from the HOBO sensor location (Fig. 1A). To determinewhether a flooding event occurred at our field point on any given day, we temporally aggregated the 15-minute water level data. Because information on how the ANA water level data was measured was not available, we considered three criteria (minimum, mean, andmaximumrecordedwater level) to determine if the data logger location was flooded on a given day and compare it with floodplain inundation values derived from ANA and the SRTM DEM data at the same location.

2.1.4. Climate and land cover change analysis

2.1.4.1. software-provided relationships between pressure, temperature, and water level. Total pressure data were converted to water pressure using barometric pressure data collected by the Pedro Afonso InMET station (Instituto Nacional de Meteorologia; #82863) 23 km away from the HOBO sensor location (Fig. 1A). To determinewhether a flooding event occurred at our field point on any given day, we temporally aggregated the 15-minute water level data. Because information on how the ANA water level data was measured was not available, we considered three criteria (minimum, mean, andmaximumrecordedwater level) to determine if the data logger location was flooded on a given day and compare it with floodplain inundation values derived from ANA and the SRTM DEM data at the same location.

3. results

3.1. it divides

3.2. Field validation

3.2.1. Modeled floodplain inundation maps predicted observed site inundation with an accuracy of 0.88 (95% confidence interval (CI): 0.84–0.91) when using mean daily data from the field sensor (Table A1). Using minimum daily water level increased accuracy to 0.93 (CI: 0.90–0.95), while aggregating by maximum daily water level had the lowest accuracy (0.83, CI: 0.79–0.87).

3.3. Flooded extent

3.3.1. Average flooded extent decreased from 132.1 km2 to 48.7 km2 after the first dam, Serra daMesa,was installed . During the Lajeado/ Cana Brava period (2003–2006), mean flooded extent increased to 52.5 km2 before decreasing to 37.0 km2 after the installation of Peixe Angical and 31.8 km2 after all dams were installed. Annual maximum flooded extent was significantly related to the number of dams on the river (p < 0.001) as well as the total annual rainfall (p < 0.01). There was also significantly lower variance in flooded extent pre-dam versus post-dam (p < 0.05; ). Specifically, before any dams were installed, maximum annual flooded extent ranged from 30.7 km2 to 438.9 km2 . After all damswere installed, this rangewas reduced to 8.0 km2 to 70.0 km2 .

3.4. Hydroperiod

3.4.1. Hydroperiod of the core inundated area (pixels thatwere inundated at least one day per year during the whole study period) declined as more dams were added to the river, after accounting for total annual rainfall and spatial dependency between pixels

3.5. Flood timing

3.5.1. For the core inundated area, the first day of inundation was significantly related to number of dams on the river after accounting for onset of rainfall,whichwas also significantly related to first day of inundation

3.6. Cumulative impacts of dams

3.6.1. The Serra da Mesa dam, the first installed upstream of the study area, had the largest impact on floodplain hydrology. Our models suggest that the addition ofmore dams on the river continues to decrease hydroperiod of the core inundated area and flooded extent but at decreasing rates.

3.7. Climate and land cover change

3.7.1. There was no significant linear trend in mean annual precipitation over the study period p = 0.30, however there was a strong and significant decrease in mean annual flow p < 0.001. Pasture in the upstream contributing area increased from 19% to 30% over the study period , during which time the river was also being dammed, making it difficult to disentangle the individual contributions of climate, damming, and land cover change on observed changes in floodplain inundation patterns.

4. Damming of the Tocantins

4.1. it divides

4.2. Biophysical, ecological, and social impacts

4.2.1. Disruptions in connectivity between the river and floodplain wetlands, such as those shown here for the Tocantins system, can lead to important geomorphological changes (Hupp et al., 2009).Wefound that 86% of the former Tocantins floodplain no longer floods after damming.

4.3. Cumulative impacts of dams

4.3.1. extent and hydroperiod were similarly affected by the addition of dams on the riverwhile flood timingwas not.Measuring impacts of additional dams on a river is challenging due to short time periods between dam installations and complex interactions between other environmental variables, such as seen in this system.

4.4. mate and land cover change

4.4.1. Mean annual river flow showed a decreasing trend over the study period, however therewas no significant trend in rainfall, indicating that the trend was not primarily driven by climate. We also found a significant negative relationship between precipitation-scaled specific discharge and percent pasture, but land cover change in the region is temporally correlatedwith damming.

4.5. Methodological considerations

4.5.1. Based on the comparison of flooding calculated fromour inundation model versus the field sensor, the models may not capture shortduration flooding events and may thus underestimate flooded area and hydroperiod. However, since themodel captured the temporal pattern in observed flooding well, including showing the pixel to be flooded during thewet season , errors in our estimates of flooded extent are likely limited to pixels that flood only very briefly