|SCUDIERO, ELIA - University Of California|
|LESCH, SCOTT - City Of Riverside|
Submitted to: Journal of Environmental Quality
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/10/2016
Publication Date: 2/22/2016
Citation: Scudiero, E., Lesch, S.M., Corwin, D.L. 2016. Validation of sensor-directed spatial simulated annealing soil sampling strategy. Journal of Environmental Quality. doi: 10.2134/jeq2015.09.0458.
Interpretive Summary: Soil spatial variability has a profound influence on a variety of environmental and agronomical issues, such as site-specific management, vadose zone hydrology and transport, and soil quality, to mention a few. Site-specific management requires field-scale soil property maps that are accurate to within a few meters (or yards). This level of accuracy can be achieved by with mobile sensors, which serve as a proxy for a target soil property (/ies), such as geospatial measurements of apparent soil electrical conductivity (i.e., ECa). To calibrate sensor readings, soil samples are needed. Sensor-directed soil sampling reduces the number of locations needed for calibration by specifying locations that represent the range and variability of the target soil properties. In this study, two ECa-directed sampling strategies are analyzed and compared: spatial simulated annealing (SSA) and response surface sampling design (RSSD). For the first time, SSA is compared with a more established sampling method (i.e., RSSD). A 6.8-ha (16.8 acres) field near San Jacinto, California was surveyed with an intense ECa sensor survey. The geospatial ECa data was used to optimize the soil sampling schemes: 30 locations per sampling strategy, at 0.3 m intervals, down to 1.8m. The ECa readings were calibrated to measured soil salinity. Results showed that the SSA sampling locations produced soil salinity maps of the same quality as those produced using RSSD soil samples, which supports the use of SSA in environmental and agronomical studies. The SSA approach has the added advantage that it defines soil properties that show abrupt change better than the RSSD approach. The use of sensor-directed SSA and RSSD has a beneficial impact on labor costs because it maps soil properties with a significant reduction in the number of soil samples as compared to designed-based sampling strategies such as stratified random sampling or unsupervised classification. This work is relevant to the soil mapping needs of producers, agriculture consultants, extension specialists, and Natural Resource Conservation Service field staff.
Technical Abstract: Soil spatial variability has a profound influence on most agronomical and environmental processes at field- and landscape-scales, including: site-specific management, vadose zone hydrology and transport, and soil quality, to mention a few. Mobile sensors are a practical means of mapping spatial variability, as their measurements serve as a proxy for many soil properties, provided a sensor-soil calibration is conducted. A viable means of calibrating sensor measurements over soil properties is through linear regression modeling of sensor and target property data. In the presented study, two sensor-directed, model-based, sampling scheme delineation methods were compared to validate recent applications of soil apparent electrical conductivity (ECa)–directed Spatial Simulated Annealing (SSA) against the more established ECa-directed Response Surface Sampling Design (RSSD) approach. A 6.8-ha study area near San Jacinto, California, was surveyed for ECa and 30 soil sampling locations per sampling strategy were selected. SSA and RSSD were compared (i) for sensor calibration to a target soil property (i.e., salinity) and (ii) for evenness of spatial coverage of the study area, which is beneficial for mapping non-target soil properties (i.e., those not correlated with ECa). The results indicate that the linear modeling ECa-salinity calibrations obtained from the two sampling schemes provided salinity maps characterized by similar errors. The maps of non-target soil properties show similar errors across sampling strategies. The SSA methodology is, therefore, validated, and its use in agronomical and environmental soil science applications is justified.