An official website of the United States government
Abstract. Deep learning (DL) rainfall–runoff models outperform conceptual, process-based models in a range of applications. However, it remains unclear whether DL models can produce physically plausible projections of streamflow under climate change. We investigate this question through a sensitivity analysis of modeled responses to increases in temperature and potential evapotranspiration (PET), with other meteorological variables left unchanged. Previous research has shown that temperature-based PET methods overestimate evaporative water loss under warming compared with energy budget-based PET methods. We therefore assume that reliable streamflow responses to warming should exhibit less evaporative water loss when forced with smaller, energy-budget-based PET compared with temperature-based PET. We conduct this assessment using three conceptual, process-based rainfall–runoff models and three DL models, trained and tested across 212 watersheds in the Great Lakes basin. The DL models include a Long Short-Term Memory network (LSTM), a mass-conserving LSTM (MC-LSTM), and a novel variant of the MC-LSTM that also respects the relationship between PET and evaporative water loss (MC-LSTM-PET). After validating models against historical streamflow and actual evapotranspiration, we force all models with scenarios of warming, historical precipitation, and both temperature-based (Hamon) and energy-budget-based (Priestley–Taylor) PET, and compare their responses in long-term mean daily flow, low flows, high flows, and seasonal streamflow timing. We also explore similar responses using a national LSTM fit to 531 watersheds across the United States to assess how the inclusion of a larger and more diverse set of basins influences signals of hydrological response under warming. The main results of this study are as follows: The three Great Lakes DL models substantially outperform all process-based models in streamflow estimation. The MC-LSTM-PET also matches the best process-based models and outperforms the MC-LSTM in estimating actual evapotranspiration. All process-based models show a downward shift in long-term mean daily flows under warming, but median shifts are considerably larger under temperature-based PET (−17 % to −25 %) than energy-budget-based PET (−6 % to −9 %). The MC-LSTM-PET model exhibits similar differences in water loss across the different PET forcings. Conversely, the LSTM exhibits unrealistically large water losses under warming using Priestley–Taylor PET (−20 %), while the MC-LSTM is relatively insensitive to the PET method. DL models exhibit smaller changes in high flows and seasonal timing of flows as compared with the process-based models, while DL estimates of low flows are within the range estimated by the process-based models. Like the Great Lakes LSTM, the national LSTM also shows unrealistically large water losses under warming (−25 %), but it is more stable when many inputs are changed under warming and better aligns with process-based model responses for seasonal timing of flows. Ultimately, the results of this sensitivity analysis suggest that physical considerations regarding model architecture and input variables may be necessary to promote the physical realism of deep-learning-based hydrological projections under climate change.
more »
« less
High resolution spatiotemporal patterns of flow at the landscape scale in montane non‐perennial streams
Abstract Intermittent and ephemeral streams in dryland environments support diverse assemblages of aquatic and terrestrial life. Understanding when and where water flows provide insights into the availability of water, its response to external controlling factors, and potential sensitivity to climate change and a host of human activities. Knowledge regarding the timing of drying/wetting cycles can also be useful to map critical habitats for species and ecosystems that rely on these temporary water sources. However, identifying the locations and monitoring the timing of streamflow and channel sediment moisture remains a challenging endeavor. In this paper, we analyzed daily conductivity from 37 sensors distributed along 10 streams across an arid mountain front in Arizona (United States) to assess spatiotemporal patterns in flow permanence, defined as the timing and extent of water in streams. Conductivity sensors provide information on surface flow and sediment moisture, supporting a stream classification based on seasonal flow dynamics. Our results provide insight into flow responses to seasonal rainfall, highlighting stream reaches very reactive to rainfall versus those demonstrating more stable streamflow. The strength of stream responses to precipitation are explored in the context of surficial geology. In summary, conductivity data can be used to map potential stream habitat for water‐dependent species in both space and time, while also providing the basis upon which sensitivity to ongoing climate change can be evaluated.
more »
« less
- PAR ID:
- 10395806
- Publisher / Repository:
- Wiley Blackwell (John Wiley & Sons)
- Date Published:
- Journal Name:
- River Research and Applications
- Volume:
- 39
- Issue:
- 2
- ISSN:
- 1535-1459
- Page Range / eLocation ID:
- p. 225-240
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Tropical regions are experiencing high rates of forest cover loss coupled with changes in the volume and timing of rainfall. These shifts can compromise streamflow and water provision, highlighting the need to identify how forest cover influences streamflow generation under variable rainfall conditions. Although rainfall is the key driver of streamflow regimes, the role of forests is less clear, particularly in tropical regions where forest loss is an ongoing risk. Forest cover loss alters evapotranspiration, rainfall infiltration and storage, and may increase stream ecosystem vulnerability to rainfall extremes. Puerto Rico, an island with spatially heterogenous forest cover and a marked geographic rainfall gradient, is projected to experience more frequent droughts and flash flooding. Using 15‐min streamflow data collected between 2005 and 2016 from 20 US Geological Survey stream gages and 3‐hourly Multi‐Source Weighted‐Ensemble Precipitation rainfall estimates, we utilized flow‐duration curves and linear mixed regression models to examine the role of forest cover in regulating the timing and volume of streamflow. The mixed model approach helps to account for differences in watershed characteristics. We determined the effects of rainfall and forest cover on low and peak flows in Puerto Rican streams, then evaluated changes in these relationships under dry and wet antecedent rainfall conditions. Watersheds with high forest cover had consistently greater low and peak streamflow than deforested ones under all rainfall conditions, although the effect was more marked during wet antecedent conditions, suggesting that peak flow is largely the result of saturation excess overland flow. During dry antecedent rainfall conditions, highly forested watersheds had higher streamflow than deforested ones, suggesting greater hillslope storage and release may also be at play. Our results demonstrate that forest cover generated a net increase in hillslope infiltration and storage and may lessen drought impacts on streamflow in Puerto Rico. Resilience to prolonged drought may be limited by finite water storage potential in this steep, mountainous setting, highlighting maintenance of forest cover as an important water management strategy to increase infiltration.more » « less
-
Abstract Water‐limited ecosystems are highly sensitive to not only precipitation amount, but also precipitation pattern, particularly variability in the size and timing of growing season rainfall events. Both rainfall event size and timing are expected to be altered by climate change, but the relative responses of dryland ecosystems to changes in rainfall event size versus timing have not been resolved. Here, we disentangle the effects of these different aspects of precipitation pattern on ecosystem dynamics.We experimentally assessed how these two aspects of rainfall variability impacted a semi‐arid grassland ecosystem by altering an ambient precipitation pattern to eliminate variability in (a) rainfall event size (all events were made the same size), (b) rainfall event timing (all events were uniformly spaced in time) and (c) both. Total precipitation amount was constant for all treatments. We measured responses of soil moisture, ecosystem carbon flux (e.g. net primary production and soil CO2flux), plant community composition and physiological responses of the dominant C4grass,Bouteloua gracilis.Removing variability in rainfall event size altered ecosystem dynamics more than a pattern of uniform event timing, but the largest impact occurred when variability in both were removed. Notably, eliminating variability in both event size and timing increased above‐ground net primary productivity by 23%, consistent with reduced water stress in the dominant C4grass, while also reducing seasonal variability in soil CO2flux by 35%, reflecting lower seasonal variability in soil moisture.Synthesis. Unique responses to different aspects of precipitation variability highlight the complexity of predicting how dryland ecosystems will be affected by climate change‐induced shifts in rainfall patterns. Our results provide novel support for the key roles of rainfall event size and timing, in addition to total precipitation amount, as determinants of ecosystem function.more » « less
-
Abstract Water temperatures in mountain streams are likely to rise under future climate change, with negative impacts on ecosystems and water quality. However, it is difficult to predict which streams are most vulnerable due to sparse historical records of mountain stream temperatures as well as complex interactions between snowpack, groundwater, streamflow and water temperature. Minimum flow volumes are a potentially useful proxy for stream temperature, since daily streamflow records are much more common. We confirmed that there is a strong inverse relationship between annual low flows and peak water temperature using observed data from unimpaired streams throughout the montane regions of the United States' west coast. We then used linear models to explore the relationships between snowpack, potential evapotranspiration and other climate‐related variables with annual low flow volumes and peak water temperatures. We also incorporated previous years' flow volumes into these models to account for groundwater carryover from year to year. We found that annual peak snowpack water storage is a strong predictor of summer low flows in the more arid watersheds studied. This relationship is mediated by atmospheric water demand and carryover subsurface water storage from previous years, such that multi‐year droughts with high evapotranspiration lead to especially low flow volumes. We conclude that watershed management to help retain snow and increase baseflows may help counteract some of the streamflow temperature rises expected from a warming climate, especially in arid watersheds.more » « less
-
Abstract Riparian zones and the streams they border provide vital habitat for organisms, water quality protection, and other important ecosystem services. These areas are under pressure from local (land use/land cover change) to global (climate change) processes. Woody vegetation is expanding in grassland riparian zones worldwide. Here we report on a decade‐long watershed‐scale mechanical removal of woody riparian vegetation along 4.5 km of stream channel in a before–after control impact experiment. Prior to this removal, woody plants had expanded into grassy riparian areas, associated with a decline in streamflow, loss of grassy plant species, and other ecosystem‐scale impacts. We confirmed some expected responses, including rapid increases in stream nutrients and sediments, disappearance of stream mosses, and decreased organic inputs to streams via riparian leaves. We were surprised that nutrient and sediment increases were transient for 3 years, that there was no recovery of stream discharge, and that areas with woody removal did not shift back to a grassland state, even when reseeded with grassland species. Rapid expansion of shrubs (Cornus drummondii,Prunus americana) in the areas where trees were removed allowed woody vegetation to remain dominant despite repeating the cutting every 2 years. Our results suggest woody expansion can fundamentally alter terrestrial and aquatic habitat connections in grasslands, resulting in inexorable movement toward a new ecosystem state. Human pressures, such as climate change, atmospheric CO2increases, and elevated atmospheric nitrogen deposition, could continue to push the ecosystem along a trajectory that is difficult to change. Our results suggest that predicting relationships between riparian zones and the streams they border could be difficult in the face of global change in all biomes, even in well‐studied sites.more » « less
