Prediction of topsoil properties at field-scale by using C-band SAR data

Abstract Designing and validating digital soil mapping (DSM) techniques can facilitate precision agriculture implementation. This study generates and validates a technique for the spatial prediction of soil properties based on C-band radar data. To this end, (i) we focused on working at farm-field scale and conditions, a fact scarcely reported; (ii) we validated the usefulness of Random Forest regression (RF) to predict soil properties based on C-band radar data; (iii) we validated the prediction accuracy of C-band radar data according to the coverage condition (for example: crop or fallow); and (iv) we aimed to find spatial relationship between soil apparent electrical conductivity and C-band radar. The experiment was conducted on two agricultural fields in the southern Argentine Pampas. Fifty one Sentinel 1 Level-1 GRD (Grid) products of C-band frequency (5.36 GHz) were processed. VH and VV polarizations and the dual polarization SAR vegetation index (DPSVI) were estimated. Soil information was obtained through regular-grid sample scheme and apparent soil electrical conductivity (ECa) measurements. Soil properties predicted were: texture, effective soil depth, ECa at 0-0.3m depth and ECa at 0-0.9m depth. The effect of water, vegetation and soil on the depolarization from SAR backscattering was analyzed. Complementary, spatial predictions of all soil properties from ordinary cokriging and Conditioned Latin hypercube sampling (cLHS) were evaluated using six different soil sample sizes: 20, 40, 60, 80, 100 and the total of the grid sampling scheme. The results demonstrate that the prediction accuracy of C-band SAR data for most of the soil properties evaluated varies considerably and is closely dependent on the coverage type and weather dynamics. The polarizations with high prediction accuracy of all soil properties showed low values of σVVo and σVHo, while those with low prediction accuracy showed high values of σVVo and low values of σVHo. The spatial patterns among maps of all soil properties using all samples and all sample sizes were similar. In conditions when summer crops demand large amount of water and there is soil water deficit backscattering showed higher prediction accuracy for most soil properties. During the fallow season, the prediction accuracy decreased and the spatial prediction accuracy was closely dependent on the number of validation samples. The findings of this study corroborates that DSM techniques at field scale can be achieved by using C-band SAR data. Extrapolation y applicability of this study to other areas remain to be tested.

[1]  P. Levelt,et al.  ESA's sentinel missions in support of Earth system science , 2012 .

[2]  Victor O. Sadras,et al.  Quantification of grain yield response to soil depth in Soybean, Maize, Sunflower, and Wheat , 2001 .

[3]  Christoph Rüdiger,et al.  Sensitivity of Sentinel-1 Backscatter to Vegetation Dynamics: An Austrian Case Study , 2018, Remote. Sens..

[4]  Á. Cabrera,et al.  Flora de la Provincia de Buenos Aires , 1972 .

[5]  Kenneth A. Sudduth,et al.  Comparison of sensors and techniques for crop yield mapping , 1996 .

[6]  Philippe Lagacherie,et al.  Digital soil mapping : an introductory perspective , 2007 .

[7]  Yang Song,et al.  Mapping Winter Wheat Planting Area and Monitoring Its Phenology Using Sentinel-1 Backscatter Time Series , 2019, Remote. Sens..

[8]  Armando Apan,et al.  Identifying the spatial variability of soil constraints using multi-year remote sensing , 2011 .

[9]  José Luis Costa,et al.  Sampling scheme optimization to map soil depth to petrocalcic horizon at field scale , 2017 .

[10]  P. Diggle,et al.  Model‐based geostatistics , 2007 .

[11]  Gustavo A. Slafer,et al.  Wheat production systems of the Pampas. , 1999 .

[12]  R. Alvarez.  Predicting average regional yield and production of wheat in the Argentine Pampas by an artificial neural network approach , 2009 .

[13]  Dennis L. Corwin,et al.  Characterizing soil spatial variability with apparent soil electrical conductivity , 2005 .

[14]  Rémy Fieuzal,et al.  Estimation of soybean yield from assimilated optical and radar data into a simplified agrometeorological model , 2015, 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS).

[15]  Michael Dixon,et al.  Google Earth Engine: Planetary-scale geospatial analysis for everyone , 2017 .

[16]  Mehrez Zribi,et al.  Retrieval of Both Soil Moisture and Texture Using TerraSAR-X Images , 2015, Remote. Sens..

[17]  Gilbert Wooding Robinson,et al.  A new method for the mechanical analysis of soils and other dispersions , 1922, The Journal of Agricultural Science.

[18]  L. S. Pereira,et al.  Crop evapotranspiration : guidelines for computing crop water requirements , 1998 .

[19]  J. Campbell,et al.  Remote sensing of crop residue and tillage practices: Present capabilities and future prospects , 2014 .

[20]  B. Brisco,et al.  The effect of soil and crop residue characteristics on polarimetric radar response , 2002 .

[21]  U. Grömping.  Dependence of Variable Importance in Random Forests on the Shape of the Regressor Space , 2009 .

[22]  D. Corwin,et al.  Application of Soil Electrical Conductivity to Precision Agriculture , 2003 .

[23]  R. D. Ramsey,et al.  Landsat Spectral Data for Digital Soil Mapping , 2008 .

[24]  Alexandre Bouvet,et al.  Understanding the temporal behavior of crops using Sentinel-1 and Sentinel-2-like data for agricultural applications , 2017 .

[25]  Alfred E. Hartemink,et al.  Digital Soil Mapping with Limited Data , 2008 .

[26]  B. Huwe,et al.  Uncertainty in the spatial prediction of soil texture: Comparison of regression tree and Random Forest models , 2012 .

[27]  Farshid Vahedifard,et al.  Investigating the correlation between radar backscatter and in situ soil property measurements , 2017, Int. J. Appl. Earth Obs. Geoinformation.

[28]  Alfred E. Hartemink,et al.  Digital Soil Mapping: Bridging Research, Production, and Environmental Application , 2010 .

[29]  Andy Liaw,et al.  Classification and Regression by randomForest , 2007 .

[30]  M. Barral,et al.  Land-use planning based on ecosystem service assessment: A case study in the Southeast Pampas of Argentina , 2012 .

[31]  G. Gee,et al.  Particle-size Analysis , 2018, SSSA Book Series.

[32]  J. Wolf,et al.  Yield gap analysis with local to global relevance—A review , 2013 .

[33]  Shoba Periasamy,et al.  Significance of dual polarimetric synthetic aperture radar in biomass retrieval: An attempt on Sentinel-1 , 2018, Remote Sensing of Environment.

[34]  Mehrez Zribi,et al.  Soil Texture Estimation Over a Semiarid Area Using TerraSAR-X Radar Data , 2012, IEEE Geoscience and Remote Sensing Letters.

[35]  Budiman Minasny,et al.  A conditioned Latin hypercube method for sampling in the presence of ancillary information , 2006, Comput. Geosci..

[36]  Claude R. Duguay,et al.  Defining the Sensitivity of Multi-Frequency and Multi-Polarized Radar Backscatter to Post-Harvest Crop Residue , 2001 .

[37]  Heather McNairn,et al.  A Review of Multitemporal Synthetic Aperture Radar (SAR) for Crop Monitoring , 2016 .

[38]  Nahuel Raúl Peralta,et al.  Prediction of Soil Properties at Farm Scale Using a Model-Based Soil Sampling Scheme and Random Forest , 2015 .

[39]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[40]  Shaun Quegan,et al.  High-resolution measurements of scattering in wheat canopies-implications for crop parameter retrieval , 2003, IEEE Trans. Geosci. Remote. Sens..

[41]  Thuy Le Toan,et al.  Multitemporal C-band radar measurements on wheat fields , 2003, IEEE Trans. Geosci. Remote. Sens..

[42]  Pierre Roudier,et al.  A conditioned Latin hypercube sampling algorithm incorporating operational constraints , 2012 .

[43]  Heather McNairn,et al.  Early season monitoring of corn and soybeans with TerraSAR-X and RADARSAT-2 , 2014, Int. J. Appl. Earth Obs. Geoinformation.

[44]  Andrej Ceglar,et al.  Detecting flowering phenology in oil seed rape parcels with Sentinel-1 and -2 time series , 2020, Remote sensing of environment.

[45]  D. Lobell,et al.  A scalable satellite-based crop yield mapper , 2015 .

[46]  José Luis Costa,et al.  A spatial dataset of topsoil texture for the southern Argentine Pampas , 2018 .

[47]  David Clifford,et al.  The Australian three-dimensional soil grid: Australia’s contribution to the GlobalSoilMap project , 2015 .

[48]  R. Lavado,et al.  Climate, organic matter and clay content relationships in the Pampa and Chaco soils, Argentina , 1998 .

[49]  Jean-Michel Poggi,et al.  Variable selection using random forests , 2010, Pattern Recognit. Lett..

[50]  F. J. Kriegler,et al.  Preprocessing Transformations and Their Effects on Multispectral Recognition , 1969 .

[51]  Fernando R. Momo,et al.  Changes in Average Annual Precipitation in Argentina’s Pampa Region and Their Possible Causes , 2015 .