Abstract Shallow surface runoff is a primary transport agent for interrill sediment delivery. Runoff, rainfall intensity, and slope interactively affect interrill erosion. We hypothesized that the inclusion of a runoff factor in an interrill erosion model can reduce the dependence of the interrill soil erodibility ( K i ) on soil infiltration characteristics as well as improve model predictability. A complete factorial rainfall simulation experiment with two soils (Cecil sandy loam, a clayey, kaolinitic, thermic Typic Kanhapludult, and Dyke clay, a clayey, mixed, mesic Typic Rhodudult), four rainfall intensities, four slopes, and two replicates was conducted under prewetted conditions to measure runoff and sediment delivery rates. Tap water with electrical conductivity <0.2 dS m −1 was used in all the runs. Rainfall intensity I , unit discharge q , slope S , soil type, and their interactions significantly affected sediment delivery per unit area ( D i ). Sediment delivery had the greatest correlation ( r = 0.68) with unit discharge; however, neither discharge nor rainfall alone adequately predicted sediment delivery. The equation D i = K i Iq 1/2 S 2/3 was proposed. The linear intensity term ( I ) represents detachment of soil by raindrop impact and enhancement of transport capacity of sheet flow, while the product of q 1/2 S 2/3 describes sediment transport by sheet flow. Validation with independent data showed that the model predicted soil erodibilities well. The mean r 2 for four validation soils was 0.93 when the proposed model was fitted to validation data to predict interrill erodibility ( K i ). The better estimation of K i indicates that interrill erosion processes were adequately described by the model.
Core Ideas Rare‐earth element oxide tracers are preferably bound to silt and clay particles. Rare‐earth element tracers are conservative during erosion and transport in each size class. A simple correction factor reduced erosion estimation error to greater than 4.5%. The correction method does not require rare‐earth element measurement in different classes. The rare‐earth element technique is capable of producing high‐quality erosion data. Spatially distributed soil erosion data are needed to better understand soil erosion processes and validate distributed erosion models. Rare‐earth element (REE) oxides have been extensively tested and used to trace soil movement at plot or small watershed scales to generate spatial erosion data. However, a general concern regarding the accuracy of the technique arose due to selective binding of REE to fine soil particles and their preferential transport during soil erosion by water. Our objective was to develop a simple enrichment correction factor to improve the accuracy of the technique without the need for measuring REE for multiple size classes. A coarse Lithic Ustipsamment soil (2% clay, 20% silt), deliberately tagged with eight REE oxides, was packed into eight zones in a physical model of a small watershed to trace erosion from each zone. Three 30‐min rains were simulated at 1, 1.5, and 2 mm min −1 , and runoff was collected every 2 min. Sediment was separated into nine size classes, and REE oxides in each class were extracted and analyzed by inductively coupled plasma–mass spectrometry. Both REE and the silt + clay fraction (<0.05 mm) were enriched in the sediment, resulting in a consistent overestimation of soil loss. An enrichment correction factor, based on enrichment of the <0.05‐mm fraction, effectively corrected the overestimation and reduced the event mean estimation errors from 18 to 40% to <4.5%. The correction circumvents the need for measuring REE for multiple size classes, which is costly and time consuming. Moreover, this study found that REE tracers were conservative within each size class during erosion, and this scientifically underpins the use of an enrichment correction factor and erosion tracing by multiple size classes. The correction factor should be further tested with different soils.
Precipitation across Oklahoma exhibits a high degree of spatial and temporal variability and creates numerous water resources management challenges. Risk-based decision making in agricultural and water resources management may find value in knowing the extent to which historical precipitation can provide guidance for inferring expectations of future monthly precipitation. A 123-year long monthly precipitation record from the central Oklahoma climate division was evaluated, in a proof-of-concept analysis, to establish whether a simple monthly precipitation decomposition into repeatable (expected) and random (unexpected) components could identify the odds of encountering repeatable or random monthly precipitation in the future. The metric for identifying precipitation kind was based on it falling within or outside of a predetermined range, defined by the relative difference between random and a percentage of repeatable precipitation. For climate conditions in central Oklahoma, the odds of future precipitation being repeatable were between 18% and 33%, depending on the month of the year. The corresponding odds of precipitation being random were 67–82%. Thus, management decisions that rely solely on historical precipitation records to anticipate future monthly precipitation have a low probability of success in central Oklahoma.
Abstract This study proposes a new hybrid model for monthly streamflow predictions by coupling a physically based distributed hydrological model with a deep learning (DL) model. Specifically, a simplified hydrological model is first developed by optimally selecting grid cells from a distributed hydrological model according to their soil moisture characteristics. It is then driven by bias corrected general circulation model (GCM) predictions to generate soil moistures for the forecasting months. Finally, model‐simulated soil moisture along with other predictors from multiple sources are used as inputs of the DL model to predict future monthly streamflows. The proposed hybrid model, using the simplified Variable Infiltration Capacity (VIC) as the hydrological model and the combination of Convolutional Neural Network and Gated Recurrent Unit (CNN‐GRU) as the DL model, is applied to predict 1‐, 3‐, and 6‐month ahead reservoir inflows for the Danjiangkou Reservoir in China. The results show that the hybrid model consistently performs better than VIC and CNN‐GRU models with great improvement in Kling‐Gupta efficiency (KGE) values for lead times up to 6 months. Additional tests indicate that hybrid models based on CNN‐GRU outperform those based on LASSO, XGBoost, CNN, and GRU models. Moreover, compared with the distributed hydrological model, the hybrid model greatly reduces the computation burden of rolling prediction. It also saves decision‐makers the time and effort of trying different combinations of predictors, which is indispensable when building DL models. Overall, the new hybrid model demonstrates great potential for monthly streamflow prediction where training data are limited.
Core Ideas We used 34 years of measured plot erosion data for 137 Cs model validation. For the first time, deposition prediction of three widely used models was evaluated. A simple proportional model appeared to perform better than mass balance models. More rigorous validation is needed in conditions including 137 Cs peak fallout period. Although the 137 Cs technique has been widely used to estimate soil redistribution in past decades, most 137 Cs‐conversion models have not been rigorously validated. The objective of this study was to explicitly evaluate the sediment deposition components of three widely used 137 Cs models using 34 yr of soil loss data from a plot (200 × 80 m). The average slope of the plot is approximately 4% in the upper section and 1% in the lower section. The primary soil (fine, mixed, thermic, Udertic or Pachic Paleustoll) is silt loam with 23% sand and 56% silt. Winter wheat ( Triticum aestivum L.) was raised primarily under conventional tillage. Sediment load was measured with a pump sampler at the outlet. Bulk soil cores were taken in a 10‐m grid to estimate 137 Cs inventory. The 137 Cs depth profiles were measured at eight locations to determine net deposition depth. The proportional model (PM) and two mass balance models (mass balance model 1 [MBM1] and mass balance model 2 [MBM2]) were evaluated. The measured average deposition depth in the depositional area of the lower section was 5.83 cm, and the predicted deposition depths in the area were 4.12, 2.02, and 1.64 cm for PM, MBM1, and MBM2. The results indicated that the simple PM appeared to predict deposition depths better than the two sophisticated mass balance models under the study conditions. However, the true capability of the two mass balance models needs to be further evaluated under more complex conditions that include the critical period of 137 Cs peak fallout in the 1950s and 1960s.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
