Elsevier

Agricultural Water Management

Volume 98, Issue 2, 15 December 2010, Pages 261-270
Agricultural Water Management

Using vegetation indices from satellite remote sensing to assess corn and soybean response to controlled tile drainage

https://doi.org/10.1016/j.agwat.2010.08.019Get rights and content

Abstract

Controlled tile drainage (CTD) is a management practice designed to retain water and nutrients in the field for crop use. CTD has shown promise for improving water quality and augmenting crop yields but findings are often restricted to field and plot scales. Remote sensing is one of the alternatives to evaluate crop responsiveness to CTD at large spatial scales. This study compared normalized and green normalized difference vegetation indices (NDVI and GNDVI) for corn (Zea mays L.) and soybean (Glycine max L.) among CTD and uncontrolled tile drainage (UCTD) fields in a ∼950 ha experimental watershed setting in Ontario, Canada from 2005 to 2008. The indices were derived from Landsat-5 and SPOT-4 satellite imagery. Log-transformed NDVI and GNDVI for soybean (R3–R6 growth stage) and corn (VT to R5–R6 growth stage) crops were higher significantly (p  0.05) for CTD, relative to UCTD for 50% (soybean) and 72% (corn) of both the log-transformed NDVI and GNDVI image acquisitions compared; only 17% and 13% were significant (p  0.05) in the reverse direction (UCTD > CTD). Log-transformed NDVI and GNDVI standard errors for CTD, relative to UCTD fields, were lower for 65% of the significant corn and 71% of the significant soybean NDVI and GNDVI comparisons for the growth stages noted above. This finding suggested overall more uniform crop growth for CTD fields relative to UCTD fields. Observed yields from a subset of commonly managed CTD and UCTD fields in the study area were not significantly different from each other (p > 0.05) with respect to tile drainage management practice; however, 87% of these paired yield comparisons indicated that CTD mean corn/soybean grain yields were greater than or equal to those for UCTD. On average, CTD observed corn and soybean grain yields were 3% and 4%, respectively, greater than those from UCTD. From observed yield and NDVI and GNDVI observations, vegetation indices vs. yield linear regression models were developed to predict grain yields over a broader land base in the experimental watershed area. Here, predicted mean yields were 0.1–11% higher for CTD corn and −5% to 4% higher for CTD soybean, relative to UCTD crops; but results varied between manured and non-manured fertilizer practices. Eighty-nine percent of the standard deviations for these yield predictions were lower for CTD relative to UCTD. The results of this study indicate that at a minimum, CTD did not adversely impact corn and soybean grain yields over the time span and field environments of the study, and based on the weight of evidence presented here, CTD shows general promise for augmenting crop performance. Finally, remote sensing derived vegetation indices such as NDVI and GNDVI can be used to assess the impact of agricultural drainage management practices on crop response and production properties.

Research highlights

This study demonstrated that utilizing high spatial resolution imagery from multi-spectral remote sensing platforms to derive vegetation indices can be useful for determining the impact of controlled tile drainage on crop growth at large spatial scales. Such information can be especially valuable for crop growth modeling and yield forecasting. This study found: (i) at growth stages ≥V10 for corn and ≥R4 for soybean, a majority of the statistical comparisons conducted indicated that controlled tile drainage increases significantly NDVI and GNDVI for corn and soybean over the experimental study area for manured and non-manured fields, (ii) controlled tile drainage increases modestly the uniformity in vegetation indices of corn and soybean, relative to conventional tile drainage, (iii) there was no significant difference among observed controlled tile drainage and conventional tile drainage crop grain yields at the p = 0.05 level, yet 87% of the observed grain yield comparisons indicated field mean grain yields for controlled drainage were greater than or equal to those for conventional tile drainage (observed corn and soybean yields were 3% and 4% greater under controlled tile drainage practices), and (iv) using site-based grain yield vs. vegetation index regression models for specific crop growth stages to predict crop yield from vegetation index pixel information (NDVI and GNDVI) over a broader spatial extent in the experimental area, it was shown that predicted non-manured corn and soybean grain yields under controlled tile drainage were (7–11%) and (3–4%) higher relative to conventional tile drainage, respectively. These percentages for manured corn and soybean were (0.1–3%) and (−5% to −2%), respectively. Hence, although CTD effects on observed grain yield were insignificant from a statistical standpoint, the results here are promising in that crops were also not adversely impacted by CTD. This, coupled with the water quality benefits of CTD, suggests broadly that CTD is a practice that has multiple benefits to the producer and the environment.

Introduction

Controlled tile drainage (CTD) is a management practice that regulates the amount of tile drainage waters that can leave a field (Gilliam et al., 1979). This practice can increase water table elevations and soil water contents in the field, thus allowing for crops to more readily access water and nutrients during critical growth stages. As a result, controlled tile drainage has been shown to increase crop yields (Tan et al., 1999, Mejia et al., 2000, Ng et al., 2002, Ma et al., 2007). Furthermore, CTD can improve surface water quality of adjacent water resources by virtue of reducing the net export of nutrients, and other agriculturally derived contaminants, via tile drainage systems (Drury et al., 1996, Wesström and Messing, 2007). In many temperate regions, tile drains are allowed to freely drain for a period of time in spring in order to facilitate normal and timely spring field operations, and often at or just after planting, tile drainage is restricted by a water flow control structure like those shown in Fig. 1. The water flow control structures that restrict discharge from a tile network can reduce the potential risk of anoxia stress to crops by permitting tiles to drain once a critically high water table is achieved in the system. Overall, CTD is a flexible management system that can be set to accommodate specific crops, topographic, and soil characteristics (ASABE, 1990, Madramootoo et al., 1993, Evans et al., 1995, Drury et al., 1996, Paasonen-Kivekas et al., 1996).

Corn (Zea mays L.) and soybean (Glycine max L.) have a relatively high demand for nutrients and water at the growth stages termed V10 and R4–R5 (Table 1, Table 2), respectively (Bruyn et al., 1995, McWilliams et al., 1999, Morrison et al., 2006). By V12 stage, corn becomes increasingly susceptible to yield reduction by moisture and nutrient deficiencies (McWilliams et al., 1999, OMAFRA, 2009a); whereas this critical time for soybean is achieved at approximately R4–R6 stage of development (Bruyn et al., 1995, Morrison et al., 2006, OMAFRA, 2009b). Depending on weather conditions, CTD could reduce nutrient and water stress during these critical growth periods.

Normalized Difference Vegetation Index (NDVI) (Tucker, 1979) and Green Normalized Difference Vegetation Index (GNDVI) (Gitelson et al., 1996) are well established and easily acquired spectral reflectance indicators of crop environmental stress (Tucker, 1979, Tucker and Sellers, 1986, Shanahan et al., 2001, Jackson et al., 2004, Vina et al., 2004); thus, these indices could potentially be used as indicators of the responsiveness of crops to augmented nutrient and water availability imposed by CTD practices. Gitelson et al. (1996) introduced the GNDVI to overcome issues of saturation observed for NDVI for some vegetation types at later growth stages. This index essentially substitutes the green band for the red band in the NDVI estimator and therefore, GNDVI has been considered more useful for assessing leaf chlorophyll variability when leaf area index is moderately high (Gitelson et al., 1996). Gianelle et al. (2009) confirmed that GNDVI was less affected by saturation and thus a good indicator of a number of predictors of vegetation productivity. A number of optical satellite sensors acquire data in the spectral wavelengths needed to produce these vegetation indices at the pixel level. Several of these satellites, including the Landsat and SPOT series, have spatial resolutions suitable for generating field-level information with swaths covering large spatial extents (e.g., Doraiswamy et al., 2004). Since the relative impact of CTD on crop performance will vary as a result of soil, weather, and topographic disposition, remote sensing derived indices such as NDVI and GNDVI could serve as a means to gauge the efficacy of CTD on crop growth and yield (e.g., Shanahan et al., 2001, Baez-Gonzalez et al., 2005, Elwadie et al., 2005) at the landscape scale where detailed crop information (e.g., yield monitoring) may be limited. If CTD is employed broadly in a region or en masse in a particular area, then vegetation indices derived from remote sensing could be used to help determine where the practice is, and is not, most effective in terms of yield augmentation (e.g., topographic, weather, and soil limiting factors). Moreover, such information may be crucial for estimating or informing at policy-relevant scales of investigation, like watersheds and counties; cost-benefits (e.g., Moran, 1994) of emerging management practices like controlled tile drainage, the implication of the drainage management practice in terms of developing/augmenting cost-share programs for specific regions (e.g., http://www.nrcs.usda.gov/PROGRAMS/EQIP; http://www.nation.on.ca/clean-water.htm), nutrient and water balances (e.g., Bella et al., 2000), and yield forecasting (e.g., Rasmussen, 1997).

We hypothesize that NDVI and GNDVI can provide a spatial indicator of crop responsiveness to increased water and nutrients resulting from CTD intervention. The main objectives of this paper are to: (i) compare NDVI and GNDVI at specific corn and soybean crop growth stages for fields in a watershed setting under CTD and uncontrolled tile drainage (UCTD) practices for 3 years of observation, and (ii) contrast observed grain yields for these tile drain management practices with those predicted over the watershed using vegetation indices vs. grain yield linear regression models.

Section snippets

Materials and methods

The study site is situated within ∼950 ha of an experimental watershed system located in eastern Ontario, Canada (45.26 N, 75.18 W). Dominant soils at the site are Bainsville silt loams (Wicklund and Richards, 1962). Lower hydraulic conductivity clayey soils underlay these soils at approximately 1.0–1.5 m depth. Local slope of the study area is generally <1%. The land occupied by soybean and corn under conventional and controlled tile drainage is given in Table 3 for years 2005–2007. Tile drain

General information

Satellite imagery was selected for the V1–R6 growth stages for corn and <V3-R6 growth stages for soybean (Fig. 2, Fig. 3, Fig. 4, Fig. 5 and Table 4, Table 5). The more advanced vegetative and the reproductive growth stages are associated with rooting systems that would be capable of accessing elevated water tables and heightened soil water contents at depth in soil resulting from CTD, relative to the very early crop growth stages where root activity is more shallow and roots less developed.

Summary and conclusions

This study demonstrated that utilizing high spatial resolution imagery from multi-spectral remote sensing platforms to derive vegetation indices can be useful for determining the impact of controlled tile drainage on crop growth at large spatial scales. Such information can be especially valuable for crop growth modeling and yield forecasting. This study found: (i) at growth stages ≥V10 for corn and ≥R4 for soybean, a majority of the statistical comparisons conducted indicated that controlled

Acknowledgements

Support for this project was provided by Agriculture and Agri-Food Canada's (AAFC) Watershed Evaluation of Beneficial Management Practice (WEBs) program and an NSERC of Canada USRA scholarship to H. Cicek. We would like to thank Dr. Malcolm Morrison (AAFC-Ottawa) and Dr. Gilles Lamothe (Univ. Ottawa) for advice on this document.

References (48)

  • I. Wesström et al.

    Effects of controlled drainage on N and P losses and N dynamics in loamy sand with spring crops

    Agric. Water Manage.

    (2007)
  • ASABE (American Society of Agricultural Biological Engineers)

    Standard ASAE EP479 Design, Installation and Operation of Water Table Management Systems for Subirrigation/Controlled Drainage in Humid Regions

    (1990)
  • A.D. Baez-Gonzalez et al.

    Large-area maize yield forecasting using leaf area index based yield model

    Agron. J.

    (2005)
  • C.M.D. Bella et al.

    Evapotranspiration estimates using NOAA AVHRR imagery in the Pampa region of Argentina

    Int. J. Remote Sens.

    (2000)
  • L.P. Bruyn et al.

    Water-Sensitive periods during the reproductive growth phase of Glycine max L. I. Regression model for estimating daily leaf water potential

    J. Agron. Crop Sci.

    (1995)
  • J. Cohen

    A power primer

    Psychol. Bull.

    (1992)
  • J. Cohen

    Statistical Power Analysis for the Behavioural Sciences

    (1988)
  • C.F. Drury et al.

    Influence of controlled drainage-subirrigation on surface and tile drainage nitrate loss

    J. Environ. Qual.

    (1996)
  • L.M. Dwyer et al.

    Rooting characteristics of corn, soybeans, and barley as a function of available water and soil physical characteristics

    Can. J. Soil Sci.

    (1988)
  • M.E. Elwadie et al.

    Remote sensing of canopy dynamics and biophysical variables estimation of corn in Michigan

    Agron. J.

    (2005)
  • R.O. Evans et al.

    Controlled versus conventional drainage effects on water quality

    J. Irrig. Drain. Eng.

    (1995)
  • Evans, R., Skaggs, R.W., 1996. Operating Controlled Drainage and Subirrigation Systems. North Carolina Cooperative...
  • N.R. Fausey et al.

    Emergence of inbred and hybrid corn following flooding

    Agron. J.

    (1985)
  • D. Gianelle et al.

    Ecosystem carbon fluxes and canopy spectral reflectance of a mountain meadow

    Int. J. Remote Sens.

    (2009)
  • Cited by (57)

    • Water productivity and crop yield: A simplified remote sensing driven operational approach

      2018, Agricultural and Forest Meteorology
      Citation Excerpt :

      In addition, Sadras et al. (2014) proposed the differentiation of mechanistic approaches into assimilation and biomass production and partitioning models. Regression models are data-driven approaches developed on the basis of the strong relationships between biophysical parameters derived from RS and yield data (e.g., Shanahan et al. (2001), Cicek et al. (2010), Panda et al. (2010)). Assimilation models are based on the optimization of one or more crop characteristics or initial conditions to better fit the behavior of the biophysical parameters derived from RS data (e.g., Sibley et al. (2014), Bouman (1992), Clevers et al. (1994)).

    • Remote Sensing and Machine Learning for Riparian Vegetation Detection and Classification

      2023, 2023 IEEE International Workshop on Metrology for Agriculture and Forestry, MetroAgriFor 2023 - Proceedings
    View all citing articles on Scopus
    View full text