Changes in land-use and land cover (LULC) occur in response to economic development and the growing demand for food. However, the impact on water availability for downstream users is often overlooked in land management policies, likely because of the lack of a well-established approach for evaluating the impact of LULC changes on river flows. This study explores the use of long short-term memory (LSTM) networks trained on data from thousands of stream gauges across the United States (US). Three LSTM networks were considered, each using different levels of LULC information: none, static (constant) and dynamic (updated annually). For this study, we created the DROMEDARY US dataset that incorporates the Cropland Data Layer dataset from the US Department of Agriculture, reflecting significant human-related LULC changes, and includes significantly more basins (3,246) than existing datasets commonly used for benchmarking hydrological models. All LSTMs demonstrated good out-of-sample prediction skills across the US. However, the ones using dynamic LULC information outperformed the others by a significant margin in reproducing observed changes in flow following changes in fallowing, an agricultural practice used to let the land rest after intense cropping cycles, or to spare water during droughts. Interestingly, using static LULC information performed worse than using no LULC information, highlighting that use of inaccurate (or outdated) information can degrade model performance at reproducing the effect of change in LULC, while having good streamflow prediction skill. Scenarios involving increased fallowed land highlighted further the benefits of using dynamic, emphasizing the need for frequently updated LULC datasets.

Improvements in irrigation technology are expected to yield water savings. Recent research highlights the need for accompanying institutional conditions (e.g., restricting irrigation expansion). However, estimating the expected quantity of water savings remains uncertain, even under such institutional conditions. This is because estimates of the water savings resulting from improved irrigation technology are subject to several methodological (sometimes arbitrary) choices. Three key choices are: (1) the underlying hydrologic model used to partition irrigation water into consumed (e.g., evapotranspiration) and non-consumed (e.g., runoff) components, (2) the selected hydrologic model parameters, and (3) the convention used to represent non-beneficial losses (e.g., non-crop evaporative losses during channel conveyance, on-farm application, off-farm storage, or unrecoverable seepage). This study is the first to explore the combined implications of these choices as regards predicting water savings. It is also the first to attribute the uncertainty in expected water savings to each of these choices. To explore these implications, we use an ensemble of water savings under all possible combinations of three different conceptual hydrologic model structures (HYMOD, HBV, SAC-SMA), a hundred equifinal parameter sets (for each model), and two conventions for representing non-beneficial losses - a total of 600 scenarios. The results show that parameter selection and alternative conventions of representing non-beneficial losses are the largest sources of uncertainty in water savings, contributing ~49% and ~33% respectively to overall uncertainty. These results provide a quantitative estimate for the minimum range of uncertainty one may expect when considering policy options that depend on quantified estimates of water savings.