Raghavendra B. Jana

Upscaling Soil Hydraulic Parameters in the Picacho Mountain Region Using Bayesian Neural Networks

A multiscale Bayesian neural network (BNN) based algorithm was applied to obtain soil hydraulic parameters at multiple scales in the Rio Grande basin (near Picacho Mountain, approximately 11 km northwest of Las Cruces, New Mexico). Point-scale measurements were upscaled to 30 m and 1 km resolutions. These scaled parameters were used in a physically based hydrologic model as inputs to obtain soil moisture states across the study area. The test sites were chosen to provide variety in terrain, land use characteristics, vegetation, soil types, and soil distribution patterns. In order to validate the effectiveness of the upscaled soil water retention parameters, and thus the soil hydraulic parameters, hydrologic simulations were conducted using the HYDRUS-3D hydrologic simulation software. Outputs from the hydrologic simulations using the scaled parameters were compared with those using data from SSURGO and STATSGO soil maps. The BNN-based upscaling algorithm for soil retention parameters from point-scale measurements to 30 m and 1 km, resolutions performed reasonably well (Pearsons R > 0.6) at both scales. High correlations (>0.6) between the simulated soil moisture values based on the upscaled and the soil map-derived soil hydraulic parameters show that the methodology is applicable to semi-arid regions to obtain effective soil hydraulic parameter values at coarse scales from fine-scale measurements of soil texture, structure, and retention data.

A Framework for Assessing Soil Moisture Deficit and Crop Water Stress at Multiple Space and Time Scales Under Climate Change Scenarios Using Model Platform, Satellite Remote Sensing, and Decision Support System

Better understanding of water cycle at different space–time scales would be a key for sustainable water resources, agricultural production, and ecosystems health in the twenty-first century. Efficient agricultural water management is necessary for sustainability of the growing global population. This warrants better predictive tools for aridity (based on precipitation, temperature, land use, and land cover), root zone (~top 1 m) soil moisture deficit, and crop water stress at farm, county, state, region, and national level, where decisions are made to allocate and manage the water resources. It will provide useful strategies for not only efficient water use but also for reducing potential risk of crop failure due to agricultural drought. Leveraging heavily on ongoing multiscale hydrologic modeling, data assimilation, soil moisture dynamics, and inverse model development research activities, and ongoing Land Data Assimilation (LDAS) and National Climate Assessment (NCA) indexing efforts we are developing a drought assessment framework. The drought assessment platform includes: (1) developing disaggregation methods for extracting various field-scale (1-km or less) climate indicators from the (SMOS, VIIRS, SMAP, AMSR-2) satellite / LDAS-based soil moisture in conjunction with a multimodel simulation–optimization approach using ensemble of Soil Vegetation Atmosphere Transfer, SVAT (Noah, CLM, VIC, Mosaic in LIS) models; (2) predicting farm/field-scale long-term root zone soil moisture status under various land management and climate scenarios for the past decades in hindcast mode and for the next decades in forecast mode across the USA using effective land surface parameters and meteorological input from Global Circulation Model (GCM) outputs; (3) assessing the potential risk of agricultural drought at different space–time scales across the USA based on predicted root zone soil moisture; and (4) evaluating various water management and cropping practices (e.g., crop rotation, soil modification, irrigation scheduling, better irrigation method/efficiency, water allocation, etc.) for risk reduction at field, county, state, region, and national scale using a web-based Decision Support System. This ongoing research provides a unifying global platform for forecasting several lagging indices for root zone soil moisture status as aridity index (AI), soil moisture deficit index (SMDI), and crop water stress index (CWSI) at the field, county, state, and regional scale on weekly, biweekly, monthly, and seasonal time scales by using various satellite and LDAS simulated data. Using available historical data, our approach is tested in various hydroclimatic regions (Great Plains, Midwest, West, Northeast, Southeast, and Southwest) across the USA. These indices form the basis for developing efficient management Decision Support Systems (DSS) for agricultural drought risk reduction and mitigation/adaption under the evolving climatic scenarios.

Continuous data assimilation for downscaling large-footprint soil moisture retrievals

Soil moisture is a key component of the hydrologic cycle, influencing processes leading to runoff generation, infiltration and groundwater recharge, evaporation and transpiration. Generally, the measurement scale for soil moisture is found to be different from the modeling scales for these processes. Reducing this mismatch between observation and model scales in necessary for improved hydrological modeling. An innovative approach to downscaling coarse resolution soil moisture data by combining continuous data assimilation and physically based modeling is presented. In this approach, we exploit the features of Continuous Data Assimilation (CDA) which was initially designed for general dissipative dynamical systems and later tested numerically on the incompressible Navier-Stokes equation, and the Benard equation. A nudging term, estimated as the misfit between interpolants of the assimilated coarse grid measurements and the fine grid model solution, is added to the model equations to constrain the model’s large scale variability by available measurements. Soil moisture fields generated at a fine resolution by a physically-based vadose zone model (HYDRUS) are subjected to data assimilation conditioned upon coarse resolution observations. This enables nudging of the model outputs towards values that honor the coarse resolution dynamics while still being generated at the fine scale. Results show that the approach is feasible to generate fine scale soil moisture fields across large extents, based on coarse scale observations. Application of this approach is likely in generating fine and intermediate resolution soil moisture fields conditioned on the radiometerbased, coarse resolution products from remote sensing satellites.

Soil moisture variability across different scales in an Indian watershed for satellite soil moisture product validation

Strategic ground-based sampling of soil moisture across multiple scales is necessary to validate remotely sensed quantities such as NASA’s Soil Moisture Active Passive (SMAP) product. In the present study, in-situ soil moisture data were collected at two nested scale extents (0.5 km and 3 km) to understand the trend of soil moisture variability across these scales. This ground-based soil moisture sampling was conducted in the 500 km2 Rana watershed situated in eastern India. The study area is characterized as sub-humid, sub-tropical climate with average annual rainfall of about 1456 mm. Three 3x3 km square grids were sampled intensively once a day at 49 locations each, at a spacing of 0.5 km. These intensive sampling locations were selected on the basis of different topography, soil properties and vegetation characteristics. In addition, measurements were also made at 9 locations around each intensive sampling grid at 3 km spacing to cover a 9x9 km square grid. Intensive fine scale soil moisture sampling as well as coarser scale samplings were made using both impedance probes and gravimetric analyses in the study watershed. The ground-based soil moisture samplings were conducted during the day, concurrent with the SMAP descending overpass. Analysis of soil moisture spatial variability in terms of areal mean soil moisture and the statistics of higher-order moments, i.e., the standard deviation, and the coefficient of variation are presented. Results showed that the standard deviation and coefficient of variation of measured soil moisture decreased with extent scale by increasing mean soil moisture.

Development of regional-scale pedotransfer functions based on Bayesian Neural Networks in the Hetao Irrigation District of China

In order to study determination the soil hydraulic parameters in the distributed hydrological models on farmland environmental effects resulted from water-saving practices of large scale irrigation district, the Bayesian Neural Networks and BP ANN model were applied to establish regional pedotransfer functions models based on the relationship of measured soil characteristic contents (saturated water content θs, residual water content θr and field water content θr), soil particle percentage, organic matter and bulk density and fitted VG model parameters of different soil texture classes from 22 soil water and salt monitoring points 110 soil samples in the Hetao Irrigation District. Then, the adaptability of two kinds of ANN models were evaluated by simulated and predicted results through the statistical results and SWRC figures. The several conclusions were reached: the ANN and BNN are both feasible PTFs methods. But, the training simulated accuracy of traditional BP model is better than that of BNN; however, the predicted accuracy of BNN model generally is better than the BP model. Furthermore, the number of input factors groups has significantly influenced the predictive accuracy of BP model. But there are little influences on the different inputs factors of BNN model. So, the BNN showed good robustness for the simple inputs. Second, the predicted SWRC has better fitness with measured and VG fitted curve than that of ANN. So, the BNN model is better than the traditional artificial neural network model has better adaptability in the peodotransfer function establishment when it uses only soil particle distribution. The BNN method is a practical method for regional pedotransfer function establishment.