Rapid prediction of soil available sulphur using visible near-infrared reflectance spectroscopy

BHABANI PRASAD MONDAL; RABI NARAYAN SAHOO; NAYAN AHMED; RAJIV KUMAR SINGH; BAPPA DAS; NILIMESH MRIDHA; SHALINI GAKHAR

doi:10.56093/ijas.v91i9.116080

Authors

BHABANI PRASAD MONDAL ICAR-Indian Agricultural Research Institute, New Delhi 110 012, India
RABI NARAYAN SAHOO ICAR-Indian Agricultural Research Institute, New Delhi 110 012, India
NAYAN AHMED ICAR-Indian Agricultural Research Institute, New Delhi 110 012, India
RAJIV KUMAR SINGH ICAR-Indian Agricultural Research Institute, New Delhi 110 012, India
BAPPA DAS ICAR-Central Coastal Agricultural Research Institute, Goa
NILIMESH MRIDHA ICAR-National Institute of Natural Fibre Engineering and Technology, Kolkata
SHALINI GAKHAR ICAR-Indian Agricultural Research Institute, New Delhi 110 012, India

https://doi.org/10.56093/ijas.v91i9.116080

Keywords:

Available sulphur, Multivariate models, PLSR, Reflectance spectroscopy, RF

Abstract

Rapid and accurate prediction of soil available S, an important secondary nutrient, is crucial for its site-specific management in a cultivated region. Although traditional chemical analysis of any nutrient is an accurate method, but often costly, time-consuming and destructive in nature. Recently visible near-infrared (VIS-NIR) reflectance spectroscopic technique has gained its popularity for rapid, non-destructive and cost-effective assessment of soil nutrients. Hence, a study was carried out in an intensively cultivated region of Katol block of Nagpur, Maharashtra, during 2018-20 for rapid prediction of soil available S using spectroscopic technique. Both spectroscopic and chemical analyses were carried out using 132 georeferenced surface soil samples (0-15 cm depth). The descriptive statistical analysis showed that the available S content varied from 1.09 to 47.88 mg/kg. Multivariate models namely partial least square regression (PLSR) and random forest (RF) were applied to develop spectral models for S prediction from spectral dataset. Several statistical diagnostics like coefficient of determination (R2), root mean square error (RMSE), ratio of performance deviation (RPD) and ratio of performance to interquartile distance (RPIQ) were used to evaluate the performances of two models. The best prediction of S was achieved from nonlinear RF model (R2 = 0.71, RMSE = 8.86, RPD =1.18, RPIQ = 1.69) as compared to linear PLSR model (R2 = 0.53, RMSE = 9.04, RPD = 1.16, RPIQ = 1.66) datasets. Therefore, the result suggested applying non-linear multivariate model (RF) for obtaining best predictability for S from spectroscopic technique.

Downloads

Download data is not yet available.

References

Breiman L. 2001. Random forests. Machine Learning 45: 5–32. DOI: https://doi.org/10.1023/A:1010933404324

Chang C W, Laird D A, Mausbach M J and Hurburg C R J. 2001. Near-infrared reflectance spectroscopy – principal component regression analysis of soil properties. Soil Science Society of America Journal 65: 480–90. DOI: https://doi.org/10.2136/sssaj2001.652480x

Chodak M, Ludwig B, Khanna P and Beese F. 2001. Use of near infrared spectroscopy to determine biological and chemical characteristics of organic layers under spruce and beech stands. Journal of Plant Nutrition and Soil Science 165: 27–33. DOI: https://doi.org/10.1002/1522-2624(200202)165:1<27::AID-JPLN27>3.0.CO;2-A

Chong I G and Jun C H. 2005. Performance of some variable selection methods when multicollinearity is present. Chemometrics and Intelligent Laboratory Systems 78: 103–12. DOI: https://doi.org/10.1016/j.chemolab.2004.12.011

Das B, Sahoo R N, Pargal S, Krishna G, Verma R, Chinnusamy V, Sehgal V K and Gupta V K. 2020a. Comparative analysis of index and chemometric techniques-based assessment of leaf area index (LAI) in wheat through field spectroradiometer, Landsat-8, Sentinel-2 and Hyperion bands. Geocarto International 35: 1415–32. DOI: https://doi.org/10.1080/10106049.2019.1581271

Mondal B P and Sekhon B S. 2019. Using diffuse reflectance spectroscopy for assessment of soil phosphorus status of an intensively cropped region. Agricultural Research Journal 56: 657–61. DOI: https://doi.org/10.5958/2395-146X.2019.00102.9

Mondal B P, Sekhon B S, Sahoo R N and Paul P. 2019. VIS-NIR reflectance spectroscopy for assessment of soil organic carbon in a rice-wheat field of Ludhiana district of Punjab. International Archives of the Photogrammetry, Remote Sensing & Spatial Information Sciences XLII-3/W6: 417–22. DOI: https://doi.org/10.5194/isprs-archives-XLII-3-W6-417-2019

Mondal B P, Sekhon B S, Paul P, Barman A, Chattopadhyay A and Mridha N. 2020. VIS-NIR reflectance spectroscopy as an alternative method for rapid estimation of soil available potassium. Journal of the Indian Society of Soil Science 68: 323–30. DOI: https://doi.org/10.5958/0974-0228.2021.00009.8

Nawar S and Mouazen A M. 2019. On-line vis-NIR spectroscopy prediction of soil organic carbon using machine learning. Soil and Tillage Research 190: 120–27. DOI: https://doi.org/10.1016/j.still.2019.03.006

Peng X, Shi T, Song A, Chen Y and Gao W. 2014. Estimating soil organic carbon using VIS/NIR spectroscopy with SVMR and SPA methods. Remote Sensing 6(4): 2699–2717. DOI: https://doi.org/10.3390/rs6042699

Salazar D F U, Dematte J A M, Vicente L E, Guimaraes C C B, Sayao V M, Cerri C E P, Padilha C de M C and Mendes W D S. 2019. Emissivity of agricultural soil attributes in south eastern Brazil via terrestrial and satellite sensors. Geoderma 114038. DOI: https://doi.org/10.1016/j.geoderma.2019.114038

Savitzky A and Golay M J E. 1964. Smoothing and differentiation of data by simplified least squares procedure. Analytical Chemistry 36: 1627–39. DOI: https://doi.org/10.1021/ac60214a047

Shepherd K D and Walsh M G. 2002. Development of reflectance spectral libraries for characterization of soil properties. Soil Science Society of America Journal 66: 988–98. DOI: https://doi.org/10.2136/sssaj2002.9880

Takele C and Iticha B. 2020. Use of infrared spectroscopy and geospatial techniques for measurement and spatial prediction of soil properties. Heliyon 6(10): e05269. DOI: https://doi.org/10.1016/j.heliyon.2020.e05269

Thissen U, Pepers M, Ustun B, Melssen W J and Buydens L M C. 2004. Comparing support vector machines to PLS for spectral regression applications. Chemometrics and Intelligent Laboratory Systems 73: 169–79. DOI: https://doi.org/10.1016/j.chemolab.2004.01.002

Viscarra Rossel R A and Behrens T. 2010. Using data mining to model and interpret soil diffuse reflectance spectra. Geoderma 158(1-2): 46–54. DOI: https://doi.org/10.1016/j.geoderma.2009.12.025

Williams C H and Steinbergs A. 1969. Soil sulphur fractions as chemical indices of available sulphur in some Australian soils. Australian Journal of Agricultural Research 10: 340–52. DOI: https://doi.org/10.1071/AR9590340

Wold S, Sjöström M and Eriksson L. 2001. PLS-regression: a basic tool of chemometrics. Chemometrics and Intelligent Laboratory Systems 58: 109–30. DOI: https://doi.org/10.1016/S0169-7439(01)00155-1

Xu S, Zhao Y, Wang M and Shi X. 2018. Comparison of multivariate methods for estimating selected soil properties from intact soil cores of paddy fields by Vis–NIR spectroscopy. Geoderma 310: 29–43. DOI: https://doi.org/10.1016/j.geoderma.2017.09.013