Jasmeet Judge

Assimilation of SMOS Soil Moisture for Quantifying Drought Impacts on Crop Yield in Agricultural Regions

This study investigates the effects of agricultural drought on crop yields, through integration of crop growth models and remote sensing observations. The soil moisture (SM) product from SM and Ocean Salinity (SMOS) mission obtained at 25 km was downscaled to a spatial resolution of 1 km, compatible with the crop models. The downscaling algorithm is based upon information theoretic learning and uses data-driven probabilistic relationships between high-resolution remotely sensed products that are sensitive to SM and in situ SM. The downscaled SM values are assimilated in the crop model using an Ensemble Kalman filter-based augmented state-vector technique that estimates states and parameters simultaneously. The downscaling and assimilation framework are implemented for predominantly agricultural region of the lower La-Plata Basin (LPB) in Brazil during two growing seasons. This rain-fed region was affected by agricultural drought in the second season, indicated by markedly lower precipitation compared to the first growing season. The downscaled SM was compared with the in situ SM at a validation site and the root mean square difference (RMSD) was 0.045 m3/m3. The crop yields estimated by the downscaling-assimilation framework were compared with those provided by the Companhia Nacional de Asastecimento (CONAB) and Instituto Brasileiro de Geografia e Estatistica (IBGE). The assimilated yields are improved during both seasons with increased improvement during the second season that was affected by agricultural drought. The differences between the assimilated and observed crop yields were 16.8% during the first growing season and 4.37% during the second season.

Applying glue for estimating CERES-Maize genetic and soil parameters for sweet corn production.

Sweet corn (Zea mays L.) is one of the five most valuable vegetable crops in Florida. The application of nitrogen fertilizer is necessary for farmers to reliably produce sweet corn. The use of crop simulation models can facilitate the evaluation of management practices that are profitable with minimal unwanted impacts on the environment. Before using such models in decision making, it is necessary to specify model parameters and understand the uncertainties associated with simulating variables that are needed for decision making. The generalized likelihood uncertainty estimation (GLUE) method was used to estimate genotype and soil parameters of the CERES-Maize model of the Decision Support System for Agrotechnology Transfer (DSSAT). The uncertainties in predictions for sweet corn production in northern Florida were evaluated using the existing field corn genotype coefficient and soil parameter database contained within DSSAT and field data collected during a series of experiments carried out in 2005 and 2006. Genotype coefficients (P1, P5, and PHINT) and soil parameters (SLDR, SLRO, SDUL, SLLL, and SSAT) were generated using a multivariate normal distribution that preserved the correlations between parameters. The soil parameter SLPF was not correlated with other parameters and was generated with a uniform distribution. After parameters were estimated, the CERES-Maize model correctly predicted the dry matter yields, anthesis dates, and harvest dates. The mean values of these variables were close to those measured in the field, with an average relative error of 4.4% and 2.4% for the data sets of 2005 and 2006, respectively. The calibrated CERES-Maize model simulated the temporal trend of leaf TKN concentration accurately during the early stage of the growth season, but underestimated the leaf TKN concentrations during the latter half of the season. The GLUE procedure accurately estimated soil parameters (SLLL, SDUL, and SSAT) when compared to independent measurements made in the laboratory, with an average absolute relative error of about 8.5%. The simulated time series of soil water content adequately simulated the observed soil water changes during both growth seasons for every layer. However, there were some large differences between simulated and observed soil nitrate contents. In a relevant further study, the average absolute relative error between model-predicted and field-estimated amounts of potential nitrogen leaching was 15.3%, which is much better than some reported comparable studies of nitrogen leaching modeling. In the posterior distribution of estimated parameters, the uncertainties in parameters were substantially reduced, with CV values mostly lower than 10%. The average CV value of the parameters was reduced from 27.2% in the prior distribution to 4.6% in the posterior distribution. In general, the results of this study showed that the CERES-Maize model was capable of simulating sweet corn production in northern Florida and the associated soil water content. The model can also simulate potential nitrogen leaching with acceptable accuracy. We suggest that the model can now be used to compare different management practices relative to productivity and potential nitrogen leaching outcomes.

Sensitivity of GNSS-R Spaceborne Observations to Soil Moisture and Vegetation

Global navigation satellite systems-reflectometry (GNSS-R) is an emerging remote sensing technique that makes use of navigation signals as signals of opportunity in a multistatic radar configuration, with as many transmitters as navigation satellites are in view. GNSS-R sensitivity to soil moisture has already been proven from ground-based and airborne experiments, but studies using space-borne data are still preliminary due to the limited amount of data, collocation, footprint heterogeneity, etc. This study presents a sensitivity study of TechDemoSat-1 GNSS-R data to soil moisture over different types of surfaces (i.e., vegetation covers) and for a wide range of soil moisture and normalized difference vegetation index (NDVI) values. Despite the scattering in the data, which can be largely attributed to the delay-Doppler maps peak variance, the temporal and spatial (footprint size) collocation mismatch with the SMOS soil moisture, and MODIS NDVI vegetation data, and land use data, experimental results for low NDVI values show a large sensitivity to soil moisture and a relatively good Pearson correlation coefficient. As the vegetation cover increases (NDVI increases) the reflectivity, the sensitivity to soil moisture and the Pearson correlation coefficient decreases, but it is still significant.

Impact of Diurnal Variation in Vegetation Water Content on Radar Backscatter From Maize During Water Stress

Microwave backscatter from vegetated surfaces is influenced by vegetation structure and vegetation water content (VWC), which varies with meteorological conditions and moisture in the root zone. Radar backscatter observations are used for many vegetation and soil moisture monitoring applications under the assumption that VWC is constant on short timescales. This research aims to understand how backscatter over agricultural canopies changes in response to diurnal differences in VWC due to water stress. A standard water-cloud model and a two-layer water-cloud model for maize were used to simulate the influence of the observed variations in bulk/leaf/stalk VWC and soil moisture on the various contributions to total backscatter at a range of frequencies, polarizations, and incidence angles. The bulk VWC and leaf VWC were found to change up to 30% and 40%, respectively, on a diurnal basis during water stress and may have a significant effect on radar backscatter. Total backscatter time series are presented to illustrate the simulated diurnal difference in backscatter for different radar frequencies, polarizations, and incidence angles. Results show that backscatter is very sensitive to variations in VWC during water stress, particularly at large incidence angles and higher frequencies. The diurnal variation in total backscatter was dominated by variations in leaf water content, with simulated diurnal differences of up to 4 dB in X- through Ku-bands (8.6-35 GHz) . This study highlights a potential source of error in current vegetation and soil monitoring applications and provides insights into the potential use for radar to detect variations in VWC due to water stress.

Numerical validation of the land surface process component of an LSP/R model

Abstract The University of Michigan’s land surface process/radiobrightness (LSP/R) model was developed as a step toward linking a traditional SVAT model to satellite microwave observations. The LSP model simulates land–air interactions and estimates surface fluxes, temperature and moisture profiles in soil and vegetation when forced with observed weather. These estimates are used by a microwave emission model, called the R model, that predicts terrain brightness temperatures. In this paper, we evaluate accuracy of the numerical methods used in the LSP model. Such rigorous tests were not conducted during the early development of the model. We describe three test-scenarios that included comparing the numerical solution with an analytic solution, evaluating coupled energy and moisture transport for a simple case, and calculating errors in energy and mass balance in the model for a realistic case using field observations. The original version of the model was modified to make it more applicable to the field conditions for the third test-case. Results from these tests demonstrate the physical self-consistency of the model and its successful implementation for the simple scenarios, and argue for its extendibility to more realistically complex cases.

Downscaling Satellite-Based Soil Moisture in Heterogeneous Regions Using High-Resolution Remote Sensing Products and Information Theory: A Synthetic Study

In this study, a novel methodology based upon the information-theoretic measures of entropy and mutual information was implemented to downscale soil moisture (SM) observations from 10 km to 1 km. It included a transformation function that related auxiliary remotely sensed (RS) products at high resolution to in situ SM observations to obtain first estimates of SM at 1 km and merging this estimate with SM at coarse resolutions through Principle of Relevant Information (PRI). The PRI-based estimates were evaluated using synthetic observations in NC Florida for heterogeneous agricultural land covers (LC), with two growing seasons of sweet corn and one of cotton, annually. The cumulative density function showed an overall error in SM of <; 0.03 cubic meter/cubic meter in the region, with a confidence interval of 95% during the simulation period. The PRI estimates at 1 km were also compared with those from the method based upon Universal Triangle (UT). The spatially averaged root mean square error (RMSE) aggregated over the vegetative LC were 0.01 cubic meter/cubic meter and 0.15 cubic meter/cubic meter using the PRI and UT methods, respectively. The RMSE for downscaled estimates using the UT method increased to 0.28 cubic meter/cubic meter when Laplacian errors are used, while the corresponding RMSE for the PRI remains the same for both Laplacian or Gaussian errors. The Kullback-Liebler divergence (KLD) for estimates using PRI is about 50% lower than those using the method based upon UT indicating that the probability density function (PDF) of the PRI estimate is closer to PDF of the true SM, than the UT method.

Soil moisture mapping using ESTAR under dry conditions from the Southern Great Plains Experiment (SGP99)

The electronically scanned thin array radiometer (ESTAR) was utilized for soil moisture mapping during the Southern Great Plains Experiment (SGP99). A retrieval algorithm was applied to obtain soil moisture from passive microwave measurements at 1.4 GHz. The algorithm was verified using ground data collected during SGP99. The results indicate a good correlation between observed and predicted soil moisture values and are consistent with results obtained from the same instrument in previous experiments. The present results demonstrate the validity of the retrieval algorithm for moderately to extremely dry soils. The ESTAR measurements along with ancillary data were used to create soil moisture maps of the entire region.

Radar Remote Sensing of Agricultural Canopies: A Review

Observations from spaceborne radar contain considerable information about vegetation dynamics. The ability to extract this information could lead to improved soil moisture retrievals and the increased capacity to monitor vegetation phenology and water stress using radar data. The purpose of this review paper is to provide an overview of the current state of knowledge with respect to backscatter from vegetated (agricultural) landscapes and to identify opportunities and challenges in this domain. Much of our understanding of vegetation backscatter from agricultural canopies stems from SAR studies to perform field-scale classification and monitoring. Hence, SAR applications, theory, and applications are considered here too. An overview will be provided of the knowledge generated from ground-based and airborne experimental campaigns that contributed to the development of crop classification, crop monitoring, and soil moisture monitoring applications. A description of the current vegetation modeling approaches will be given. A review of current applications of spaceborne radar will be used to illustrate the current state of the art in terms of data utilization. Finally, emerging applications, opportunities and challenges will be identified and discussed. Improved representation of vegetation phenology and water dynamics will be identified as essential to improve soil moisture retrievals, crop monitoring, and for the development of emerging drought/water stress applications.

CALIBRATION OF THE CERES-MAIZE MODEL FOR LINKAGE WITH A MICROWAVE REMOTE SENSING MODEL

Stored water, i.e., soil moisture in the root zone, is the most important factor governing energy and moisture fluxes at the land surface. Crop models are typically used to estimate these fluxes and simulate crop growth and development. Remotely sensed microwave observations can be used to improve estimates of these fluxes, biomass, and yield. This research aims to calibrate a crop growth model, CERES-Maize, for a growing season of corn in north-central Florida. The CERES-Maize model was extended to weather and soil conditions of the region and calibrated using data from our second Microwave Water and Energy Balance Experiment (MicroWEX-2). The calibrated model was linked to a microwave brightness (MB) model to estimate brightness signatures of the growing corn canopy. Overall, the CERES-Maize model estimated realistic total biomass with a root mean square error (RMSE) of 1.1 Mg/ha and a Willmott d-index of 0.98. However, the partitioning of total biomass into stem and leaf biomasses were under- and overestimated, respectively. LAI matched well with the MicroWEX-2 observations with an RMSE of 0.10 and a Willmott d-index of 0.99. The model estimated realistic daily latent heat flux with an RMSE of 42 W/m2. The soil moisture and temperature profiles of deeper soil layers matched reasonably well with observations, with RMSE of 1% to 3.5% and 1.4 to 3.7 K, respectively. Near-surface (0-5 cm) soil moisture and temperatures were less realistic because the hydrological processes near the surface need to be modeled on a much shorter timestep than is allowed by the crop model. The microwave emission model was run using observed canopy and soil inputs, as well as with the modeled canopy and soil inputs (linked crop-MB). The two methods produced similar seasonal trends in brightness temperatures with an RMS difference of 18.50 K. However, the linked model could not capture diurnal variations in brightness temperatures due to its daily timestep. Such integrated crop-MB models can be used for assimilation of remotely sensed microwave brightness in future studies to improve estimates of land surface fluxes and crop growth and development.

Comparison of Calibration Techniques for Ground-Based C-Band Radiometers

We quantify the performance of three commonly used techniques to calibrate ground-based microwave radiometers for soil moisture studies, external (EC), tipping-curve (TC), and internal (IC). We describe two ground-based C-band radiometer systems with similar design and the calibration experiments conducted in Florida and Alaska using these two systems. We compare the consistency of the calibration curves during the experiments among the three techniques and evaluate our calibration by comparing the measured brightness temperatures (TBs) to those estimated from a lake emission model (LEM). The mean absolute difference among the TBs calibrated using the three techniques over the observed range of output voltages during the experiments was 1.14 K. Even though IC produced the most consistent calibration curves, the differences among the three calibration techniques were not significant. The mean absolute errors (MAEs) between the observed and LEM TB s were about 2-4 K. As expected, the utility of TC at C-band was significantly reduced due to transparency of the atmosphere at these frequencies. Because IC was found to have a MAE of about 2 K that is suitable for soil moisture applications and was consistent during our experiments under different environmental conditions, it could augment less frequent calibrations obtained using the EC or TC techniques