Item No. 1 of 1

INVESTIGATOR: Mishra, A.

PERFORMING INSTITUTION:
CLEMSON UNIVERSITY
CLEMSON, SOUTH CAROLINA 29634

TOWARDS A NEAR REAL-TIME AGRICULTURAL DROUGHT MONITORING AND FORECASTING

NON-TECHNICAL SUMMARY: Our research objectives, which align with the NIFA objectives is quite relevant due to recent drought situations nationwide in that it will help to improve agricultural water management under drought scenarios. The funding will be used to improve technologies to provide near real-time drought forecast information for farmers, ranchers, forest owners and managers, public policy experts, public and private managers and citizens to improve water resource quantity. Currently two major limitations that exists in agricultural drought monitoring/forecasting are: (i) soil moisture derived from hydrologic models or remote sensing products provide aggregated information at coarse resolution and often witness larger uncertainty, however, the agricultural drought relies on finer/local scale information; (ii) the agricultural drought is monitored and forecasted using available soil moisture at a uniform (constant) depth, which may not be suitable in real world scenarios.

OBJECTIVES: Agricultural drought, usually, refers to a period with declining soil moisture and consequent crop failure without any reference to surface water resources. A decline of soil moisture depends on several climate-catchment variables; therefore by incorporating high resolution real time soil moisture into drought monitor will improve predicting agricultural drought at near real-time conditions. This is important as farmers/growers require real-time information on status of soil moisture availability to decide 'when to irrigate and how much to irrigate'. Additionally, the drought forecasted information will indicate whether the ongoing drought will be progressive or recessive in nature. So far, advancement is made in agriculture drought monitoring, however limited success was observed for developing near realtime forecasting capabilities. The models, interactive analysis and results will be disseminated for use by different stakeholders. These tools/models will be designed for application and deployment anywhere in the U.S.

APPROACH: Technological advancements (e.g., remote sensing, weather/climate forecasts), has greatly improved drought identification, monitoring, and with reasonable accuracy in forecasting at a regional scale. We propose to forecast agricultural drought based on surface and sub-surface drought information generated by using a crop model-data assimilation framework using a combination of multiple data sets.

PROGRESS: 2016/03 TO 2017/03
Target Audience:We investigated the performance of two different type models to improve rootzone soil moisture (agricultural drought) prediction, which includes: (a) A combination of support vector machines (SVMs) and data assimilation technique is applied to predict soil moisture as well as to forecast the drought index at the subsequent one week time period, and (b) we developed a crop modeling system (DSSAT-CSM model) over the study area, Barnwell County, South Carolina, to simulate the crop yield and assess the influence of water availability on the crop yield. The remote sensing data is assimilated to update the DSSAT-CSM modeling results with Ensemble Kalman Filter (EnKF) technique. The above two products will be useful to agricultural extension specialist, farmers/growers to improve agricultural water management under near real-time scenarios. These tools/models are designed for use anywhere in the U.S. The project is expected to improve operational drought management by providing improved near real-time information for agricultural water management. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?One postdoctoral researcher was trainednew robust data-driven tools, data assimilation techniques, how to solve scientific problems andwritingscientific papers. These training activities would benefit for the participants to improve their professional skills and attain greater proficiency. How have the results been disseminated to communities of interest?The results during this period study: (a) published in the top journals in the hydrology and water resources field, (b)the models, interactive analysis and results aredisseminated for use by users identified in the objectives. These tools/models will be designed for use anywhere in the U.S. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

IMPACT: 2016/03 TO 2017/03
What was accomplished under these goals? Two major findings includes: 1. Weexaminedthe combination of support vector machines (SVM) and the dual ensemble Kalman filter (EnKF) technique to estimate root zone soil moisture at different soil layers up to 100cm depth. Multiple experiments are conducted in a data rich environment to construct and validate the SVM model and to explore the effectiveness and robustness of the EnKF technique. It was observed that the performance of SVM relies more on the initial length of training set than other factors (e.g., cost function, regularization parameter, and kernel parameters). The dual EnKF technique proved to be efficient to improve SVM with observed data either at each time step or at a flexible time steps. The EnKF technique can reach its maximum efficiency when the updating ensemble size approaches a certain threshold. It was observed that the SVM model performance for the multi-layer soil moisture estimation can be influenced by the rainfall magnitude (e.g., dry and wet spells). 2.we evaluated the performance of community land surface model (CLM4.5) to simulate the hydrologic fluxes, such as, soil moisture (SM), evapotranspiration (ET) and runoff with (without) remote sensing data assimilation. The Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR E) daily SM (both ascending and descending) are incorporated into the CLM4.5 model using data assimilation (DA) technique. The GLDAS data is used to validate the AMSR E SM data and evaluate the performance of CLM4.5 simulations. The AMSR E SM data are rescaled to meet the resolution of CLM4.5 model. By assimilating the AMSR E SM data into the CLM4.5 model can improve the SM simulations, especially over the climate transition zones in Africa, East Australia, South South America, Southeast Asia, and East North America in summer season. The Local Ensemble Kalman Filter (LEnKF) technique improves the performance of CLM4.5 model compared to the directly substituted method. The improvement in ET and surface runoff simulations from CLM4.5 model assimilated with AMSR E SM data shares similar spatial patterns with SM.

PUBLICATIONS (not previously reported): 2016/03 TO 2017/03
1. Type: Journal Articles Status: Published Year Published: 2016 Citation: D. Liu, A.K. Mishra, Z. Yu, 2016. Evaluating Uncertainties in Multi-layer Soil Moisture Prediction with Support Vector Machines and Ensemble Kalman Filtering, Journal of Hydrology, 538:243-255.
2. Type: Journal Articles Status: Published Year Published: 2017 Citation: D. Liu, A.K. Mishra, 2017. Performance of AMSR E Soil Moisture Data Assimilation in CLM4. 5 Model for Monitoring Hydrologic Fluxes at Global Scale. J. Hydro. 547, 67-79.

PROGRESS: 2015/03/15 TO 2016/03/14
Target Audience:The project will addresses agricultural drought monitoring and forecasting under near real-time scenarios, which will improve decision making activity of several stake holders (e.g., agriculture extension specialist, farmers/growers). While there have been advances in agriculture drought monitoring, developing forecasting capabilities has been marginally successful. The models, interactive analysis and results will be disseminated for use by users identified in the objectives. These tools/models will be designed for use anywhere in the U.S. The project is expected to improve operational drought management by providing improved near real-time information for agricultural water management. Rather than providing drought information in relative terms (e.g., a drought is a long, dry period), we will identify the onset, severity, and termination of drought periods. The information will be useful for stakeholders from agriculture, hydrology, water managers and state/federal agencies. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?This period study provides some training activities for the participants to learn the new robust data-driven tools, data assimilation techniques, how to solve scientific problems.These training activities would benefit for the participants to improve their professional skills and attain greater proficiency. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?We will extend our DA modeling framework to be incorporated in Crop models for near real time drought monitoring and forecasting.

IMPACT: 2015/03/15 TO 2016/03/14
What was accomplished under these goals? Technological advancements (e.g., remote sensing, weather/climate forecasts), has greatly improved drought identification, monitoring, and with reasonable accuracy in forecasting at a regional scale. We proposed to forecast agricultural drought based on surface and sub-surface drought information generated by using a crop model-data assimilation framework using a combination of multiple data sets. Previous agricultural drought research considered uniform depth of soil moisture for all types of available crops to quantify agricultural drought scenarios. However, as discussed above, the moisture available in different layers and root zone depth will play an important role for the quantification of agricultural drought. In this period, we developed a multi-layer soil moisture prediction model using a data-driven model (SVM) and a sequential DA method known as EnKF technique. The data used in this study is from the Blackville experiment site located in South Carolina of United States. Blackville experimental site (latitude 33°21'18", longitude 81°19'40", elevation 317ft) is located in South Carolina, United States. In addition to the real time sensor based soil moisture monitoring station, this site also includes a state of the art NOAA U.S. Climate Reference Network station (SC Blackville 3W) with automated measurements of air temperature, humidity, solar radiation, rainfall, soil temperature and moisture at different soil depth from 5-100cm. With a group of sensitivity experiments, we found that the combination of SVM and dual EnKF is a robust statistical tool suitable for the multi-layer soil moisture prediction. Based on the sensitivity experiments, the main findings are listed as follows: (a) The SVM is good at simulating the soil moisture at different layer from surface to root zone at the training process but the performance for predicting lacks confidence especially for the deeper soil layers. (b) The performance of SVM relies most on the initial training ensemble size compared with other factors (e.g., cost function, regularization parameter, and kernel parameters). The ability of SVM can be improved with enough training ensemble members and the deeper the soil layer, the more ensemble members are needed. (c) Compared with the precipitation, the soil moisture predictions have a distinctive relationship with the precipitation pattern, where the soil moisture predictions at each layer tend to overestimate when there is little precipitation while underestimate when there is more precipitation. (d) The dual EnKF technique for the model state and parameter state is better than the DA update at the model state or parameter state only. (e) The EnKF technique can reach its maximum efficiency when the updating ensemble size approaches around 15, which is consistent with the previous findings,and (f) The EnKF technique can improve the SVM even with limited initial training ensemble members. This study proves the prediction of root zone soil moisture with data assimilation technique, which would benefit for the future soil moisture predictions over a large study area.

PUBLICATIONS: 2015/03/15 TO 2016/03/14
Type: Journal Articles Status: Under Review Year Published: 2016 Citation: D. Liu, A. K. Mishra, Z. Yu, 2016. Evaluating Uncertainties in Multi-layer Soil Moisture Prediction with Support Vector Machines and Ensemble Kalman Filtering, submitted to Journal of Hydrology (under review)