Elsevier

Geoderma Regional

Volume 25, June 2021, e00389
Geoderma Regional

Mapping of tank silt application using Sentinel-2 images over the Berambadi catchment (India)

https://doi.org/10.1016/j.geodrs.2021.e00389Get rights and content

Highlights

  • We studied an Indian age-old practice: a black-colored tank silt application over red-colored soils.

  • Soil color maps were produced by supervised classification method using Sentinel-2 images.

  • Soil color changes between two S2 images were used as proxy to identify tank silt applications.

  • The proposed approach allows mapping soil colors with correct accuracy.

  • This anthropogenic soil modification was identified over 202 fields.

Abstract

Mapping soil properties is becoming more and more challenging due to the increase in anthropogenic modification of the landscape, calling for new methods to identify these changes. A striking example of anthropogenic modifications of soil properties is the widespread practice in South India of applying large quantities of silt from dry river dams (or “tanks”) to agricultural fields. Whereas several studies have demonstrated the interest of tank silt for soil fertility, no assessment of the actual extent of this age-old traditional practice exists. Over South-Indian pedological context, this practice is characterized by an application of black-colored tank silt to red-colored soils such as Ferralsols. The objective of this work was to evaluate the usefulness of Sentinel-2 images for mapping tank silt applications, hypothesizing that observed changes in soil surface color can be a proxy for tank silt application. We used data collected in a cultivated watershed in South India including 217 soil surface samples characterized in terms of Munsell color. We used two Sentinel-2 images acquired on February and April 2017. The surface soil color over each Sentinel-2 image was classified into two soil types (“Black” and “Red” soils). A change of soil color from “Red” in February 2017 to “Black” in April 2017 was attributed to tank silt application. Soil color changes were analyzed accounting for possible surface soil moisture changes. The proposed methodology was based on a well-balanced Calibration data created from the initial imbalanced Calibration dataset thanks to the Synthetic Minority Over-sampling Technique (SMOTE) methodology, coupled to the Cost-Sensitive Classification And Regression Trees (Cost-Sensitive CART) algorithm. To estimate the uncertainties of i) the two-class classification at each date and ii) the change of soil color from “Red” to “Black”, a bootstrap procedure was used providing fifty two-class classifications for each Sentinel-2 image. The results showed that 1) the CART method allowed to classify the “Red” and “Black” soil with correct overall accuracy from both Sentinel-2 images, 2) the tank silt application was identified over 202 fields and 3) the soil color changes were not related to a surface soil moisture change between both dates. With the actual availability of the Sentinel-2 and the past availability of the LANDSAT satellite imageries, this study may open a way toward a simple and accurate method for delivering tank silt application mapping and so to study and possibly quantify retroactively this farmer practice.

Introduction

The soil has an integral part to play in the global environmental sustainability challenges of food security, water security, climate stability, biodiversity, and ecosystem service delivery (McBratney et al., 2014) and accurate mapping of soil properties is crucial for adequate management both at global and very local levels. For many years, digital soil mapping represents an alternative to conventional soil survey, to produce soil predictions maps and associated uncertainties, exploiting large sets of spatial data (e.g., Digital Elevation Model, optic satellite data) combined to recent or ancillary data (McBratney et al., 2003).

To face an increasing demand of food, water scarcity and soil health degradation, farmers adapt their crop types, crop rotations, water supplies and practices (Hardaker, 2004) and these land use practices linked to soil management may determine frequently changes in soil class (Dazzi and Lo Papa, 2015), as the Solonchaks developed from Cambisols in arid environments because of irrigation with saline water, or Regosols derived from the truncation of Cambisols due to soil erosion.

India is highlighted as one of the most risk-prone countries for water scarcity, declining soil fertility through land degradation and climate change impacts (Roberts, 2001). A traditional “tank system”, composed of cascades of reservoirs along valleys, has been used for centuries not only to harvest water for irrigation but also to trap sediments to limit erosion losses at the catchment scale by restoring tank-trapped sediments to agricultural fields (Gunnell and Krishnamurthy, 2003). While this latter practice used to be mostly limited to fields in the vicinity of tanks, the recent development of motorization (excavators and tractors) and of governments-implemented vast programs of tanks desilting (DHAN Foundation, 2020) have given this age-old practice a new momentum. However, to our knowledge, no estimate of the temporal and spatial evolutions of this is available.

While the primary objective of tank desilting is to increase the water storage capacity of these tanks, several studies have demonstrated the interest of restoring tank silt to agricultural fields in India, as it improves soil fertility (DHAN Foundation, 2020; Karanam et al., 2008; Osman et al., 2009; Obi Reddy et al., 2017), specifically increasing soil water holding capacity (Deshmukh et al., 2019), organic carbon and available nutrient status (Patil et al., 2017). Moreover, these benefits of tank sediments for the soil quality has been also demonstrated in others parts of the word such as in Peru (Walter et al., 2012) and Poland (Baran et al., 2019).

The potential of this technique is large in India: The number of irrigation tanks is estimated around 0.3 million all over India (Reddy et al., 2018), with 35% located in south India (Tamil Nadu, Karnataka and Andhra Pradesh) (Narayanamoorthy, 2007). From the 1990s, tank irrigation declined sharply at the expense of groundwater irrigation, which accounted for almost 60% of the irrigated area after only 10 years (Shah et al., 2003), thanks to the development of submersible pump technology. However, tank irrigation is still one of the major strategies for adapting to rainfall variability (CWC, 2010; Reddy et al., 2018), and has a great potential provided sustainable solution are found to resolve the problem of silting. In three of the southern Indian states (Andhra Pradesh, Karnataka and Tamil Nadu), about 140,000 tanks are silted up and their rehabilitation could both increase irrigation potential and improve soil health (DHAN Foundation, 2020).

Over pedological contexts characterized by Vertisols, Ferralsols and Chromic Luvisols like over the Deccan Plateau, tank silt is characterized by a black color and is applied to red-colored soils such as Ferralsols and Chromic Luvisols. Therefore, tank silt applications are easily recognizable over landscapes as they change topsoil color. On the field, soil color is commonly and widely measured using a Munsell soil color chart (Munsell Color Company, 1975) which is a system for categorical qualifications of soil color. This chart was designed to reflect our perception of color and its variations, and is widely used by pedologist as a determinant of soil type. The topsoil color can also be studied by Visible-Near Infrared and Short Wave Infrared (VNIR-SWIR, 400 to 2500 nm) remote sensing as various soil components exhibit spectral response in the visible range of the electromagnetic spectrum, between wavelengths 400 and 700 nm. The topsoil color study by VNIR-SWIR remote sensing may allow to discriminate eroded and non-eroded soils (e.g., Pickup and Nelson, 1984), identify surface efflorescence and salt crust (e.g., De Jong, 1992; Mougenot and Pouget, 1993) and estimate soil organic carbon (e.g., Viscarra Rossel et al., 2006).

With the successful launch of the Sentinel-2 satellites, the Copernicus program has provided global coverage of terrestrial surfaces with multispectral images, with a revisit time of ten days from 2015 to 2017 and five days since 2017 (European Space Agency (ESA), 2015; https://sentinel.esa.int/web/sentinel/missions/sentinel-2). Sentinel-2 images are acquired with a spatial resolution of 10 m to 60 m on 13 spectral bands in the VNIR-SWIR spectral domain. Several studies have proven the relevance of individual Sentinel-2 images for soil studies, including primary soil properties mapping such as soil organic mapping (e.g., Gholizadeh et al., 2018; Vaudour et al., 2019) and soil texture (e.g., Gomez et al., 2019), and soil salinity retrieval based on an Electrical Conductivity mapping (e.g., Taghadosi et al., 2019; Wang et al., 2019). Some works have studied the multitemporal dimension of Sentinel-2 images for soil studies, and showed that it may allow to i) increase the probability of image acquisition in clear sky conditions during periods with consistent bare soil coverage over cultivated areas (e.g., Vaudour et al., 2019), ii) elaborate mosaic images of bare soil area (e.g., Loiseau et al., 2019) and iii) estimate uncertainties of permanent soil properties predictions (Gomez et al., 2019).

The objective of this work is to evaluate the utility of Sentinel-2 images for mapping tank silt application over red soils. This work is based on a changes analysis of soil color from a Sentinel-2 image to another one and was carried out in an Indian cultivated region (Berambadi catchment, Karnataka state). It used two Sentinel-2 images acquired on February and April 2017 and 217 soil surface samples, collected over the study area and characterized in terms of Munsell color.

Section snippets

Study area

The Berambadi catchment is a subcatchment of the South Gundal located in the Deccan Plateau of Southern India (Fig. 1a) and covers 84 km2. Our study area is located in the eastern part of the Berambadi catchment, which is covered by crop fields and corresponds to 60% of the catchment (Fig. 1b), while the western part is covered by forest. The Berambadi catchment belongs to the Kabini Critical Zone Observatory (AMBHAS, BVET, Sekhar et al., 2016; Tomer et al., 2015), which is part of the OZCAR

Methods

To map the tank silt application from February to April 2017, the proposed approach followed 3 main steps. In a first step, N Calibration and Validation databases were built with a bootstrap approach from both Sentinel-2 image and the corresponding observed Munsell soil colors (Fig. 4b). A classification model was built from each of the N Calibrations dataset and then validated from the associated Validation dataset (Fig. 4b). So N maps of “Black” soils and “Red” soils were provided (Fig. 4a)

Classifications performances

50 classification models were built from the Sentinel-2 image acquired on the 3rd of February 2017 and using balanced data obtained by the SMOTE method and the cost-sensitive CART classification method. These 50 models provided 50 classifications of “Black” and “Red” soils with accurate overall accuracies (mean overall accuracy = 0.81) and moderate agreement (mean kappa = 0.51) (Fig. 5). The standard deviations of overall accuracies and kappa over the 50 models derived from the Sentinel-2 image

Mapping of tank silt applications

Based on the proposed approach, 100 fields corresponding to 16.2 ha and 102 fields corresponding to 25.6 ha were associated to a silt tank application with more than 45 changes over 50 iterations and between 40 changes to 45 iterations, respectively. So this approach is able to identify this old-age practice in our context in spite of the small size of fields (0.27 ha in average, Sharma et al., 2018) and allows highlighting that this practice of tank silt application concerns a minority of

Conclusion

The objective of this study was to evaluate the utility of Sentinel-2 images for mapping tank silt application characterized by a black color as they come from Vertisols soils, over red soils using a couple of Sentinel-2 images encompassing the tank silt applications period. In this study a two-class classification at each Sentinel-2 date was produced with correct accuracies and then the changes of soil color from the first image to the second one were produced. This approach took care of

Declaration of Competing Interest

None.

Acknowledgments

The study was funded by the project ATCHA ANR-16-CE03-0006. The authors are indebted to NBSS&LUP for soil samples collection. The Kabini Critical Zone Observatory (AMBHAS, BVET, Sekhar et al., 2016; Tomer et al., 2015, www.ambhas.com; https://mtropics.obs-mip.fr/) which is part of the OZCAR network (Gaillardet et al., 2018, http://www.ozcar-ri.org/ozcar/), are also acknowledged.

References (63)

  • A. Ali et al.

    Classification with class imbalance problem: a review

    Int. J. Adv. Soft. Compu. Appl

    (2015)
  • AMBHAS Team

    A Manual for Agro-Hydrological Monitoring in Pilot Experimental Watersheds

    (2015)
  • A. Baran et al.

    An assessment of bottom sediment as a source of plant nutrients and an agent for improving soil properties

    Environ. Eng. Manag. J.

    (2019)
  • L. Barbiero et al.

    Ferrolysis induced soil transformation by natural drainage in vertisols of sub-humid South India

    Geoderma

    (2010)
  • R.J. Barnes et al.

    Standard normal variate transformation and de-trending of near-infrared diffuse reflectance spectra

    Appl. Spectrosc.

    (1989)
  • H. Bartholomeus et al.

    Soil organic carbon mapping of partially vegetated agricultural fields with imaging spectroscopy

    Int. J. Appl. Earth Obs. Geoinf.

    (2011)
  • L. Breiman et al.

    Stone Classification and Regression Trees Belmont, Wadsworth, CA

    (1984)
  • L. Brodský et al.

    Uncertainty propagation in VNIR reflectance spectroscopy soil organic carbon mapping

    Geoderma

    (2013)
  • N.V. Chawla et al.

    Smote: synthetic minority over-sampling technique

    J. Artif. Intell. Res.

    (2002)
  • N.V. Chawla et al.

    Editorial: special issue on learning from imbalanced data sets

    ACM SIGKDD Explor. Newslett.

    (2004)
  • J. Cohen

    A coefficient of agreement for nominal scales

    Educ. Psychol. Meas.

    (1960)
  • CWC

    Water and Related Statistics

    (2010)
  • C. Dazzi et al.

    Anthropogenic soils: general aspects and features

    Ecocycles

    (2015)
  • S. De Jong

    The analysis of spectroscopical data to map soil types and soil crusts of Mediterranean eroded soils

    Soil Techn.

    (1992)
  • J.A.M. Dematte et al.

    Is it possible to classify topsoil texture using a sensor located 800 km away from the surface?

    Revista Brasileira de Ciência do Solo

    (2016)
  • D.P. Deshmukh et al.

    Characterization of physical attributes of tank silt of natural reservoirs in western Melghat region of Dharni

    Int. J. Chem. Stud.

    (2019)
  • DHAN Foundation

    Tank silt Application for Agricultural Production Enhancement - Scope, Issues and Challengers

    (2020)
  • G. Douzas et al.

    Geometric SMOTE a geometrically enhanced drop-in replacement for SMOTE

    Inf. Sci.

    (2019)
  • G. Douzas et al.

    Imbalanced learning in land cover classification: improving minority classes’ prediction accuracy using the geometric SMOTE algorithm

    Remote Sens.

    (2019)
  • B. Efron et al.

    An Introduction to the Bootstrap

    (1993)
  • European Space Agency (ESA)

    Sentinel-2 User Handbook

    (2015)
  • W.J. Frampton et al.

    Evaluating the capabilities of Sentinel-2 for quantitative estimation of biophysical variables in vegetation

    ISPRS J. Photogramm. Remote Sens.

    (2013)
  • J. Gaillardet et al.

    OZCAR: the French network of critical zone observatories

    Vadose Zone J.

    (2018)
  • A. Gholizadeh et al.

    Soil organic carbon and texture retrieving and mapping using proximal, airborne and Sentinel-2 spectral imaging

    Remote Sens. Environ.

    (2018)
  • C. Gomez et al.

    Analysis of the uncertainties affecting predictions of clay contents from VNIR/SWIR hyperspectral data

    Remote Sens. Environ.

    (2015)
  • C. Gomez et al.

    Use of sentinel-2 time-series images for classification and uncertainty analysis of inherent biophysical property: case of soil texture mapping

    Remote Sens.

    (2019)
  • Y. Gunnell et al.

    Past and present status of runoff harvesting systems in dryland peninsular India: a critical review

    AMBIO: J. Hum. Environ.

    (2003)
  • J. Hardaker
  • N. Japkowicz et al.

    The class imbalance problem: a systematic study

    Intell. Data Anal.

    (2002)
  • P.V. Karanam et al.

    Economic evaluation of sediments as a source of plant nutrients

    Current Science

    (2008)
  • J. Kuhn et al.

    Caret: Classification and Regression Training

    (2016)
  • Cited by (9)

    • Assessment of spectral reduction techniques for endmember extraction in unmixing of hyperspectral images

      2024, Advances in Space Research
      Citation Excerpt :

      Other regions that indicate the presence of black soils in the abundance map could be agricultural fields that have tank silt applied over the surface in the interest of soil fertility. This application of tank silt is also a common practise observed in the region and was studied by Gomez et al. (2021). The abundance map of crops shows the presence of crop fields distributed over the top right and bottom half of the image.

    • Multi-label sub-pixel classification of red and black soil over sparse vegetative areas using AVIRIS-NG airborne hyperspectral image

      2023, Remote Sensing Applications: Society and Environment
      Citation Excerpt :

      Multiple labels were assigned to each sample according to the availability of red or black soil, vegetation or crop residue at the sample location. In general, the success rate of the single label classifier for mapping the soil types depend on the bare soil pixels (Gomez et al., 2021). An increase in vegetation canopy cover can change the shape of soil reflectance spectra (Gomez et al., 2021; Ouerghemmi et al., 2016).

    View all citing articles on Scopus
    View full text