Proximal NDVI derived phenology improves in-season predictions of wheat quantity and quality

Highlights

• Daily NDVI best predicted harvesting metrics during the mid- and late season.

• Phenological metrics were derived using smoothed, daily NDVI data.

• NDVI derived phenological metrics improved predictions of harvest metrics.

• Grain protein and N were primarily driven by phenology during reproductive stages.

Abstract

Automated, low-cost and field-deployable remote sensing tools are well suited for continuously monitoring crop growth and providing growers with timely information about crop performance. Because automated sensors provide information about crop development and performance across time, we examined the hypothesis that ground-based canopy reflectance data might define crop phenology in new ways over the course of the season that can better forecast crop yield, protein, biomass, and grain nitrogen at harvest. This study examines the utility of daily Normalized Difference Vegetation Index (NDVI) data to monitor crop phenology over two complete growing seasons. Spectral reflectance data was collected at a total of sixteen plots under four different applied nitrogen (N) and soil water availability scenarios in rainfed soft white spring wheat (Triticum aestivum L.). Using NDVI at solar noon, four phenological periods were derived from the data using a non-parametric regression locally weighted smoothing parameter (loess) to account for day to day variability, and piecewise linear regression to determine inflection points in the seasonal NDVI curve. The NDVI derived phenological metrics (i.e. the change in NDVI per day, and duration (in days) of each phenological period) were compared against daily NDVI values throughout the season to predict harvest metrics. Daily NDVI data were generally poor predictors of harvest metrics early in the growing season, and reached maximum predictive power at the onset of heading, and the middle of ripening for biomass and yield (R² ∼ 0.50 and ∼0.25 during heading, respectively, and R² ∼ 0.50 during early ripening). Conversely, using both simple and multiple regression analysis, we found that harvest metrics were better explained using the rate and duration of NDVI derived phenological periods. Simple regressions between NDVI derived phenological metrics revealed several physiologically and management relevant correlations including strong, statistically significant (p < 0.05) relationships between the rate of tillering and stem extension and total biomass (R² = 0.63 and 0.54, respectively), the duration of heading and yield (R² = 0.67), the rate of ripening and grain protein concentration (R² = 0.45), and the duration of ripening and grain N content (R² = 0.43), for example. Using multiple regression analysis, 83% of the variance in yield, 67% in protein concentration, 87% in total biomass, and 80% in grain N was explained by two to three NDVI derived phenological metrics. Further, multiple regression analysis using NDVI derived phenological metrics from the early season (tillering and stem extension) substantially improved early prediction of yield and biomass as compared to daily NDVI data, whereas protein and grain N were primarily driven by metrics associated with the reproductive development of the crop (heading and ripening). This work has implications for improving in-season management decisions and understanding of the phenological drivers of harvest metrics using daily NDVI data as an evaluation tool.

Introduction

Daily information regarding crop development patterns is important for monitoring and predicting grain quantity and quality. Highly resolved, objective, and real-time information about crop phenology and growth can aid in within-season farm management decisions, but also can improve the understanding of abiotic and physiological processes controlling plant N uptake, yield, and protein content. The motivation behind tracking key phenological phases in crop development is not new (Hanway, 1963, Cleary and Waring, 1969, Zadoks et al., 1974); but its quantification has improved significantly in recent years due to the advent of canopy reflectance data available from spectro-radiometric platforms (Goodin and Henebry, 1997, Raun et al., 2001, Viña et al., 2004). More recently, several low-cost approaches toward continuous monitoring of plant phenology have been conducted using digital time-lapse cameras (Rundel et al., 2009, Richardson et al., 2007, Sakamoto et al., 2012), filtered photodiodes (Garrity et al., 2010, Magney et al., 2016), light-emitting diodes (Ryu et al., 2010), and autonomously operating terrestrial laser scanners (Eitel et al., 2013). The rapid development of low-cost, easily-interpretable, and field-ready ground based remote sensing systems is the result of a growing interest in tracking temporal and spatial changes in the physiological and phenological status of vegetation (Gamon et al., 2006, Richardson et al., 2012).

Some of the initial development of remotely sensed vegetation indices, such as the Normalized Difference Vegetation Index (NDVI) (Tucker, 1979) – a differenced ratio of reflected energy in the red and near-infrared portions of the electromagnetic spectrum – was prompted by the motivation to indirectly predict grain yield using bands available from space using Landsat satellite data (Rouse et al., 1974, Tucker et al., 1980, Pinter et al., 1981, Aase and Siddoway, 1981). Following these early explorations, issues such as sensor view angle, solar angle, atmospheric conditions, radiometric calibration, canopy architecture, and soil background were determined to be important factors confounding crop canopy reflectance (Verhoef, 1984, Huete, 1987, Jackson and Huete, 1991, Eitel et al., 2009). Since then, interest has grown quickly around selection of the optimal sensor angles, wavelengths, measurement frequencies, spatial resolution, radiometric resolution, and technical capabilities of instruments used for monitoring and predicting crop growth parameters (see review by Mulla, 2013). For nearly four decades (Tucker, 1979), NDVI has remained one of the most consistently measured vegetation indices across a wide variety of sampling platforms, prompting its widespread use in agriculture.

Highly temporally resolved NDVI data have been widely used to track seasonal phenology of green-up and senescence over a wide variety of ecosystems from space using NOAA's advanced high resolution radiometer (AVHRR, e.g., Justice et al., 1985, Myeni et al., 1997, Brown and de Beurs, 2008) and NASA's Moderate Resolution Imaging Spectrometer (MODIS, Fisher and Mustard, 2007, Soudani et al., 2008). Additionally, more immediately available information has been used at much finer spatial and temporal resolutions using ground-based instruments (Vierling et al., 1997, Huemmrich et al., 1999, Viña et al., 2004, Huete, 2012, Soudani et al., 2012, Magney et al., 2016). Accurate interpretation of satellite-based data requires robust, and highly temporally resolved, ground-based reference data (Nguy-Robertson et al., 2013). Furthermore, the inherent complexity, processing time, and difficulties in acquiring and interpreting satellite images at the field scale can make this technology inaccessible for many growers to make within-season management decisions. Ground-based sensors therefore can inform both agricultural management decisions, as well as to track plant phenological variation that can be used by crop scientists seeking a more process-based understanding of dynamics controlling biomass accumulation and grain-fill (e.g., Oscarson, 2000, Farooq et al., 2014).

Because NDVI is an integrated measure of canopy greenness, and ground-based radiometers can repeatedly collect NDVI readings at consistent height and viewing geometry in <1 second, these data can be compared across different field locations, nutrient plots, irrigation regimes, and over cultivars with traits engineered to maximize reproductive growth (e.g., Borrell et al., 2014). The plant breeding community has a keen interest in developing crops that ‘stay-green’ longer, increasing duration of grain-fill, and decreasing the rate of senescence (Christopher et al., 2014, Gaju et al., 2014). Phenological data has already been used to investigate the physiological and morphological traits necessary to increase grain quality and quantity using visual observations and instruments that quantify the loss of chlorophyll using a handheld chlorophyll meter (Borrell et al., 2000). Using chlorophyll meters to derive phenological metrics such as the onset and rate of senescence, researchers have found that ‘stay-green’ phenotypes can retard senescence, promoting a longer grain fill period (Harris et al., 2007). Additionally, phenological dynamics during the post-anthesis period in wheat have been described using logistical models (Pepler et al., 2005). However, hand-held chlorophyll meters are limited in both time and space, with data collections that can suffer from subjective measurement bias due to leaf selection, and infrequent field visits. High temporal automated measurements of NDVI have provided a more robust and objective approach to indirectly estimate crop phenological expression (Lopes and Reynolds 2012).

The overall goal of this study was to investigate the utility of NDVI to derive physiologically and management relevant phenological periods, and investigate their capacity to make within-season predictions of biomass, grain yield, N uptake, and grain protein concentration. Our specific objectives were to: (1) investigate the predictive capacity of daily NDVI values throughout the growing season to model end of season grain yield, biomass, grain N uptake, and protein concentration; and (2) compare daily NDVI values to the rate and duration of four phenological periods (onset of tillering, stem extension, heading, and ripening) in predicting harvest metrics (biomass, grain yield, N uptake, and grain protein concentration).