Location: Agroclimate and Hydraulics Research Unit
2023 Annual Report
Objectives
1. Fingerprint sediment sources and develop spatially distributed soil erosion data using fallout radionuclides for calibrating the Water Erosion Prediction Project (WEPP) model.
2. Enhance existing tools to improve simulation of storm intensification and assessment of climatic impact for long-term strategic conservation planning, and develop new seasonal analogue forecast tool for climate-smart decision support.
Sub-objective 2.A: Improve methods of simulating storm intensification based on GCM/RCM-projected changes in high percentiles (tail) of daily precipitation for better downscaling of GCM/RCM climate projections to a target station for site specific impact assessment and conservation planning.
Sub-objective 2.B. Develop a seasonal analogue forecast tool based on big data mining using an Artificial Intelligence (AI)-driven KNN algorithm for climate-smart decision support in managing winter wheat-livestock production in the Southern Great Plains (SGP).
3. Data collection, model calibration, and model simulation for long-term strategic planning and short-term tactical decision support for crop production and soil and water conservation.
Sub-objective 3.A. Compile weather, wheat yield, cattle grazing data, and crop management information to calibrate and fine tune an existing wheat grazing model.
Sub-Objective 3.B. Simulate wheat and beef production with the wheat-grazing model
using monthly updated, seasonal analogue climate forecast data obtained in Sub-Objective 2.B for tactical within-season decision making in managing the wheat-livestock enterprise for select stations in the SGP.
Sub-Objective 3.C. Simulate runoff, soil water balance, soil loss, and crop yield with the WEPP model to assess the impacts of storm intensification due to climate change on erosion and crop production under various cropping and tillage systems using downscaled GCM/RCM projections in Objective 2.A for strategic conservation planning at decadal scales.
Approach
The Food and Agriculture Organization has projected that food production needs to increase by 70% to feed the world population of 9.3 billion by 2050. However, agricultural production is being adversely impacted by global warming due to increasing extreme weather events and climate variability. Thus, adapting agriculture production to climate change and variation or developing climate-smart decision support information is imperative to feed the world by taking advantage of favorable changes while mitigating adverse impacts. This research seeks to refine climate downscaling tools to improve modeling of extreme precipitation events and their impacts on soil and water conservation measures, develop seasonal analogue climate forecasts and dual-purpose wheat decision support tools, and derive spatially distributed soil erosion data. The two weather generator-based downscaling tools will be further refined to simulate extreme precipitation events by explicitly manipulating the top percentiles of daily precipitation based on projected climate change signals or historical trends. A seasonal climate analogue tool will be developed using a K Nearest Neighbor approach driven by an Artificial Intelligence (AI)-based data mining algorithm. A wheat grazing model will be used along with seasonal forecasts to develop a tactical within season decision support tool for managing the wheat-livestock enterprise in central Oklahoma. In addition, improved simulation of extreme precipitation will afford great opportunities for more accurate assessments of climatic impacts on soil erosion and crop production and for development of better strategic conservation planning at decadal scales. The seasonal climate forecast and decision support tools are expected to have great impacts on the wheat-livestock enterprise, a major economic pillar, in the Southern Great Plains (SGP). Distributed erosion data, derived using the Cs-137 tracking technique, will be used to validate and improve process-based soil erosion models, which in turn will better assist in strategic planning of long-term soil and water conservation.
Progress Report
Objective 1: Evaluate WEPP model.
Research continued on calibrating and validating the Water Erosion Prediction Project (WEPP) model using measured surface runoff and soil loss data between 1954 and 2015, as well as the spatially distributed soil erosion data derived for two small watersheds at Coshocton, Ohio, using the Cs-137 erosion conversion models. For the model calibration, the collected weather, soil properties, tillage, and crop management data since 1954, and digital elevation model (DEM) were used to compile the four WEPP input files: climate, soil, crop management, and topography. The measured input files were used to run the hillslope version of the WEPP model. The model-simulated surface runoff volumes and soil erosion rates were compared to those of the measured data in the two watersheds. The parameter of soil hydraulic conductivity which describes soil’s ability to infiltrate or conduct water was manually varied or adjusted to match the simulated and measured surface runoff volumes during 1954-2015. After the runoff volumes were well matched, soil erodibility parameters, which depict soil’s ability to resist erosion by water, were manually adjusted to match the simulated and measured total soil losses at the watershed outlet during 1954-2015. The calibrated parameters were then used to run the hillslope version of the WEPP model along each downslope transect as a representative hillslope. More than 10 representative hillslopes were run for each watershed, and the simulated soil loss rates at a 10 m interval along each representative hillslope were recorded. The averaged soil loss rates were compared to those derived from the Cs-137 technique for the same slope positions. The WEPP-simulated soil erosion rates were generally comparable to those estimated with the newly improved Cs-137 model as well as a few other empirical Cs-137 conversion models for most slope positions. However, the Cs-137 method predicted soil deposition at upper-middle slopes and near the watershed outlet, while the WEPP model predicted net erosion at these locations. The discrepancy could be caused by the coarse spatial resolution used in developing the representative hillslope. The work is underway to evaluate a flow path version of the WEPP model, which is a research version and provides soil erosion estimates at higher spatial resolutions.
Objective 3/Sub-objective 3.A: Complete calibration of wheat grazing model. Research continued on calibrating and evaluating a wheat growth model of the Decision Support Systems for Agrotechnology Transfer (DSSAT) using collected data from Oklahoma State University (OSU) wheat variety trials and fall forage experiments at Stillwater, Oklahoma, and Chickasha sites. Five most popular cultivars of Doublestop CL+, Gallagher, Duster, Smith’s Gold, and Green Hammer and a legacy cultivar of Jagger were selected for calibration and evaluation. The wheat model was calibrated for Chickasha site and then tested and evaluated at the Stillwater and Oklahoma sites. The Five cultivar parameters and several phenological development stage parameters in the Ecotype file were calibrated. The experimental data from 2014 to 2021 were used in the calibration. There were generally 3 to 5 years of experimental data available for each cultivar. First, the wheat development stage parameters were calibrated by ‘matching’ the observed first hollow stem dates and heading dates with those simulated by the model for each cultivar. Second, following the growth stage calibration, we calibrated the fall forage production by varying a radiation use efficiency parameter. Finally, the yield analysis parameters such as seed size and seed number per unit area were used to calibrate grain yields. Given the large variations in the observed data, the calibration was first done manually to confine the parameter values to a proper range, and then an autocalibration program distributed with the DSSAT wheat model was run automatically to fine tune the parameter values. The calibrated parameter values for each cultivar were used to make predictions at the Oklahoma and Stillwater sites using the measured climate, wheat management, and soil data as input. Overall results showed that the simulated plant growth stages, fall forage production, and grain yields were positively correlated with those of the measured values. The calibrated wheat model will be used with the seasonal climate forecasts developed in Sub-objective 2B for a decision support exercise at El Reno, Oklahoma, in the coming years.
Objective 3/Sub-objective 3.C: Complete compilation of cropping and tillage systems and execution of WEPP model for impact assessment. Research continued on compiling cropping and tillage systems for use in simulating the impacts of climate changes on surface runoff, soil erosion, and crop production under various management systems so that effective systems in conserving soil and water under future climates can be selected for the study site. Tillage systems could be roughly categorized into three types: intensive tillage (or conventional tillage), conservation tillage (including reduced tillage in number and intensity), and no-tillage. Conventional tillage (CT) accounted for about 43% of farmlands in Oklahoma. Conventional tillage is a combination of intensive chiseling and disking that includes about four passes per year and leaves the soil surface almost bare. Reduced tillage (RT) or conservation tillage includes disking immediately after harvest and field cultivations afterward and leaves < 30% residue cover on the soil surface after planting. No-till (NT) leaves soil undisturbed and retains all aboveground plant biomasses on the soil surface. Additionally, delayed tillage (DT) was proposed to reflect Oklahoma farmers’ preference for a cleaner seedbed while having the benefit of surface residue cover during a fallow period. Delayed tillage and reduced tillage had similar tillage implements and passes but differed in sequence. The main difference is the timing of the primary disking tillage, which occurred a few days before planting of the next crop in delayed tillage, while it occurred immediately following the harvest of the previous crop in reduced tillage. Together, CT, RT, and NT represented the major tillage types in Oklahoma, and thus were selected for use as potential tillage systems in future under climate change. The top representative crops in Oklahoma included winter wheat, soybean, corn (maize), cotton, sorghum, and canola (oilseed rape). Among the top regional representative crops, winter wheat accounted for approximate 66% of the planted acreage. Since corn and sorghum often share similar tillage practices and management schedules and have a similar impact on soil erosion, sorghum, along with winter wheat, soybean, cotton, and canola, were used in devising cropping systems. The continuous and rotational cropping systems were evaluated. Crops using a continuous monocultural cropping system include winter wheat, soybean, cotton, sorghum, and canola in four tillage systems of CT, DT, NT, and RT. These crops were also grown under the conventional tillage management in three-year rotations with two-year alfalfa. Additionally, a continuous winter wheat-summer soybean was simulated for CT, DT, NT, and RT. A total of 29 combinations of cropping and tillage management were simulated by WEPP. The typical crop management practices in Oklahoma such as tillage implements, sequences and timing as well as planting and harvest dates were used to compile the WEPP crop management input files for use in simulating the climatic impacts in the next phase of Sub-objective 3.C under the climate change scenarios developed in Sub-objective 2.A.
Accomplishments
1. Soil redistribution estimation using Cs-137. Process-based erosion prediction models were developed to predict spatial and temporal changes of soil erosion. However, most soil loss data measured with traditional methods at the outlets of runoff plots and watersheds are spatially lumped, and thus cannot be used to validate erosion models. Sediment tracers like radionuclide Cs-137 have been used to develop spatially distributed erosion data. However, this technique has not been rigorously tested due to the rarity of long-term measured soil loss data. ARS researchers at El Reno, Oklahoma, thoroughly evaluated the Cs-137 technique using the soil loss data measured during 1954-2015 in two small watersheds at Coshocton, Ohio. Measured Cs-137 activity (or concentration) for each sample was converted to a soil loss or deposition rate for each sampling location using widely available Cs-137 erosion conversion models. The estimated mean annual soil erosion rates for each watershed were compared to the overall mean annual erosion rates measured at the outlet. The results showed that all theoretical conversion models significantly overestimated the mean soil erosion rates. A close examination showed that a key assumption upon which the theoretical models rest is invalid because the enriched loss of Cs-137 in surface runoff water and with sediment occurred during the Cs-137 fallout periods when Cs-137 was being transferred from rainwater to soil particles. A new model was developed to overcome the invalid assumption by including a Cs-137 loss flux during the transfer process. The new model improved soil redistribution estimation considerably. A spatially distributed soil erosion dataset was then developed with the new model. The new model would be useful to soil conservationists and erosion scientists for estimating spatial variation of soil erosion using the Cs-137 technique and for further validating erosion models. The derived spatial erosion data is critical for implementing precision soil conservation practices.
2. Developed a local seasonal precipitation forecast tool. Seasonal forecasts of rain at specific locations would improve the ability of farmers to make climate smart and money smart decisions for the coming season. Such a tool would allow for better use of farmland by allowing for good decisions about present water resources with knowledge of the future water state. However, to this day, a local forecast tool for seasonal rainfall is not available, owing to the difficulty in predicting local rainfall. To address this issue, ARS researchers at El Reno, Oklahoma, along with outside research collaborators, have created a forecast tool that uses local historical weather data to inform and predict future periods of precipitation over monthly and seasonal periods. To do this, a computer program was developed by matching the temporal precipitation and temperature patterns of historical weather data with those of the current forecast year to identify analogue years for use as potential forecasts. Temporal climate patterns at every n clustered day (n=1, 2, 3, 4, …) were examined and compared to find analogue years. This clustered pattern recognition approach, compared with an existing method that only considers daily climate patterns (n=1), improved the ability of the system to predict seasonal rainfall correctly. This new forecast system will provide farmers with future rainfall predictions in areas with available weather data. This is more useful than the currently available Climate Prediction Center seasonal forecasts, which is difficult to adapt for use for farmers due to the overlapping nature of the forecasts. Given the importance of local seasonal totals of rain, farmers need better predictions of rain at their location to make useful and climate smart decisions related to their farming needs. This new tool will be used to give farmers the rainfall data they need to succeed and as such we will begin testing the tool at El Reno, Oklahoma, before making the tool available to farmers online.
3. Quantifying and generating storm intensification under climate change. Frequent and heavier rainfall storms dubbed storm intensification are projected to increase under climate change; however, it is quite challenging to statistically simulate storm intensification for a location based on large-scale global climate model (GCM) projections. Computer programs called weather generators are often used as a bridging/downscaling tool to generate future local weather data based on the GCM-projected climate change trends. Proper adjustment of key parameters of weather generators is critical for successful generation of future daily precipitation sequences and magnitudes, including extreme events. ARS researchers at El Reno, Oklahoma, tested the SYNthetic weather generaTOR (SYNTOR) and the Generator for Point Climate Change (GPCC) for simulating future storm intensification. The statistical parameters are first developed using the present climate records and then are adjusted based on GCM-projected climate change signals/trends. Rainfall occurrence probabilities which determine when rain occurs are adjusted based on changes in total monthly precipitation amounts projected by the GCM. Probability distribution parameters for precipitation amounts for GPCC, which determine the amount of rain, are adjusted using GCM-projected monthly precipitation statistics as well as percent changes in extreme events. For SYNTOR, the projected percent changes in extreme events are directly applied to scale the generated daily precipitation time series. Results showed that GCM projections can provide change signals for parameter adjustment for generating proper sequence and magnitudes of daily precipitation for locations, including extreme events. This information will be useful to scientists and engineers who are interested in assessing the impacts of climate changes on hydrology, crop productivity, and soil erosion at local scales and in developing strategic plans for conserving soil and water under climate change.
