Elsevier

Remote Sensing of Environment

Volume 149, June 2014, Pages 227-238
Remote Sensing of Environment

Monitoring multi-layer canopy spring phenology of temperate deciduous and evergreen forests using low-cost spectral sensors

https://doi.org/10.1016/j.rse.2014.04.015Get rights and content

Highlights

  • LED-sensors can measure NDVI and LAI concurrently at multiple canopy depths.

  • Integration of NDVI and LAI can monitor multi-layer canopy phenology accurately.

  • Understory canopy development determined the onset of greenness in MODIS NDVI.

Abstract

Emerging near-surface remote sensing techniques have advanced our ability to monitor forest canopy phenology. Thus far, however, little effort has been made to monitor the phenologies of the various canopies of multi-layer forests separately, despite their importance in regulating forest biogeochemical cycles. Here we report phenological changes in multi-layer canopies of deciduous broadleaf and evergreen needleleaf forests in the Republic of Korea during the spring of 2013. We installed light-emitting diode (LED) sensors at four different canopy heights at two sites to measure the normalized difference vegetation index (NDVI) using red and near-infrared (NIR) spectral reflectance and to estimate leaf area index (LAI) using the blue band gap fraction. LED-sensors identified leaf-out dates of over- and understory canopies at both sites; leaves unfolded 8–11 days earlier in the understory canopy than the overstory canopy. At the deciduous forest site, LED-NDVI failed to capture the leaf-out date in the overstory canopy, because all four LED-sensors started to see green-up from the understory canopy while the overstory canopy was leafless. LED-LAI identified different leaf-out dates for the over- and understory canopy, because the gap fraction was measured explicitly for each canopy layer. In the evergreen forest site, LED-NDVI signals between the top of the tower and beneath the overstory canopy were decoupled because of the dense evergreen overstory canopy. Both LED-NDVI and LED-LAI identified new needle expansion in the overstory canopy and understory canopy development. MODIS NDVI agreed well with LED-NDVI data (R2 = 0.96, RMSE = 0.04) at the deciduous forest site, and we discovered that understory canopy development determined the onset of greenness based on MODIS NDVI data. LED-LAI data agreed well with independent estimates from the other instruments, indicating that LED-sensors could be used to monitor multi-layer canopy LAI. Continuous, in-situ observation of multi-layer canopy phenology will aid in the interpretation of satellite remote sensing phenology products and improve land surface models that adopt a multi-layer canopy model.

Introduction

Forest structures, which typically form multi-layer canopies, are complex in space and time. Multi-layer canopies exhibit different phenological patterns that modify species composition, light capture, carbon, water, and nutrient cycles in the forest (Baldocchi et al., 2004, Pearcy, 1990, Seiwa, 1998). For example, understory species maximize annual mass gain and survival by unfolding leaves earlier than overstory species in deciduous broadleaf forests (Seiwa, 1998). Although field observations can be used to monitor multi-layer canopy phenology (Richardson & O'Keefe, 2009), it is unclear how to monitor multi-layer canopy phenology automatically and continuously. Strengths and limitations of different vegetation metrics, such as Normalized Difference Vegetation Index (NDVI) (Tucker, 1979) and leaf area index (LAI) for monitoring multi-layer canopy phenology also have to be established.

To monitor forest canopy phenology, researchers generally obtain spectral data from the sky toward the forest canopy, spectral data from the forest floor toward the sky, and use quantum sensors to measure light attenuation through the canopy. Numerous studies have focused on monitoring canopy phenology from the sky. Instruments used include radiometric sensors (Huemmrich et al., 1999, Schmid et al., 2000, Soudani et al., 2012), digital cameras (Nagai et al., 2010, Richardson et al., 2007, Sonnentag et al., 2012), and satellite remote sensing (Ganguly et al., 2010, White et al., 2009, Zhang et al., 2003). Looking at the forest from the sky enables forest phenology to be monitored from the plot level to continental scales. However, those sensors mostly see the top of canopies, and it is unclear how in-depth the sensors look at the forests, as this depends on forest openness and sun–target–sensor geometries (Pisek & Chen, 2009). Thus observing the forest from the sky is unlikely to capture multi-layer canopy phenology in forests, particularly dense forests.

Another approach to monitor forest phenology is to look upwards at the forest from the forest floor. This method has the advantage that details of the forest inside can be obtained. In an open forest, such as that found in a savanna ecosystem where the tree canopy height is 10 m, upward-pointing digital cameras accurately monitored tree canopy phenology despite slight variations in LAI over the seasons (0 to 0.9) (Ryu et al., 2012). A recent study of tall, dense, closed canopies reported that hemispherical photographs taken from the forest floor did not capture the canopy phenology of a coniferous forest well, although the same technique captured the canopy phenology of a deciduous forest well (Nagai et al., 2013). To the best of our knowledge, no study has evaluated if upward-pointing digital cameras can capture multi-layer canopy phenology.

Light attenuation through canopies is a good indicator of canopy phenology. One century ago, Salisbury (1916) observed that changing light environments at the forest floor are related to phenology. In the 1970s, an innovative field campaign was conducted to measure the forest light environment using traversing radiometer systems that horizontally moved through Sitka Spruce forest canopies at different canopy depths (Norman and Jarvis, 1974, Norman and Jarvis, 1975). A similar system was established in an Oak-Hickory forest in the1980s to monitor multi-layer canopy phenology (Baldocchi et al., 1984, Baldocchi et al., 1985, Chason et al., 1991, Hutchison et al., 1986). Traversing radiometer systems have been shown to be able to capture spatial variations in light environments in forests (Brown, 1973, Herrington et al., 1972). However, these types of system require regular and careful maintenance of sophisticated infrastructure such as rails, motors, cables, masts and radiometers, which hampers continuous observation of multi-layer phenology over seasons and years. As an alternative, researchers have observed light attenuation through canopies by placing photosynthetically active radiation (PAR) sensors at fixed locations above and below canopies to monitor over- and understory canopies together, not separately (Nagai et al., 2013, Novick et al., 2004, Richardson et al., 2007). In spite of the advancements in observing light penetration through canopies over the century, one challenging task that remained is to monitor multi-layer canopy phenology separately and automatically.

Two most widely used variables for monitoring vegetation phenology are NDVI and LAI. NDVI reflects vegetation activity, whereas LAI is a key canopy structural variable that controls land–atmosphere interactions (Norman and Jarvis, 1974, Ryu et al., 2011). Although both metrics are related to canopy phenology, their utility for monitoring multi-layer canopy phenology is less well known. A series of studies measured NDVI in the field above the canopy, not inside the canopy (Fensholt and Sandholt, 2005, Ryu, Baldocchi, et al., 2010, Soudani et al., 2012). LAI is typically estimated by measuring the gap fraction (GF) (Miller, 1967, Monsi and Saeki, 1953, Welles and Norman, 1991), which requires measuring light intensity at two different heights. Several studies measured overstory and understory LAI separately using quantum sensors, but monitoring was manual (Baldocchi et al., 1984, Barr et al., 2004, Nasahara et al., 2008). The emergence of inexpensive but reliable spectral sensors that measure both PAR and NIR regions separately (Garrity et al., 2010, Ryu, Baldocchi, et al., 2010) has made it possible to monitor NDVI and LAI concurrently at multiple canopy depths.

Multi-layer canopy phenology monitoring data are needed to better interpret and evaluate satellite-derived phenology metrics. During a green-up period, satellite-derived phenological metrics are influenced by both over- and understory canopies and the forest floor (Ahl et al., 2006, Eriksson et al., 2006, Nagai et al., 2010, Suzuki et al., 2011). Ahl et al. (2006) evaluated a series of phenological metrics derived from MODIS Land products and concluded that the over- and understory canopies should be monitored separately in the field. Furthermore, the coarse temporal revisit frequency of satellites and temporal composites in satellite imagery are additional sources of uncertainty when quantifying phenological metrics (Morisette et al., 2009). Continuous observation of over- and understory canopy phenologies might aid in the interpretation of satellite remote sensing phenology products.

In this study, we report how we used LED-sensors (Ryu, Baldocchi, et al., 2010) to monitor the multi-layer canopy phenology of a temperate deciduous broadleaf forest and an evergreen needleleaf forest in Korea during the spring of 2013. We installed LED-sensors, which measure spectral irradiance of red, blue, green, and near-infrared bands, at four different canopy depths to measure NDVI and LAI concurrently for each canopy layer at both sites. To evaluate the efficacy of LED-sensors at detecting phenological events, we integrated in-situ observations, upward-pointing digital camera data, and Plant Canopy Analyzer, LAI-2200 (LI-COR Biosciences, Lincoln, NE) data. We then evaluated MODIS NDVI-derived phenology metrics by scaling up in-situ phenological records through LED-NDVI and Landsat NDVI. Our goal in this study was to characterize phenological changes of multi-layer canopies for two different forest types, deciduous broadleaf forest and evergreen needleleaf forest. The scientific questions that we addressed are as follows: 1) Are phenological metrics in multiple canopy layers consistent between sensors and indices (NDVI and LAI)? 2) How do vertical profiles of NDVI and LAI differ before and after leaf expansion in deciduous and evergreen forests? 3) What does MODIS NDVI see during a green-up period?

Section snippets

Site description

Study sites were a deciduous broadleaf forest (DBF) and an evergreen needleleaf forest (ENF) in Gwangneung Experimental Forest, which is part of the Korea Flux Network (Kim et al., 2006) (Fig. 1). Gwangneung Experimental Forest is located in the mid-western part of the Korean Peninsula, and has a typical cool-temperate climate. Annual maximum, minimum, and mean temperatures are 35, − 15, and 10 °C, respectively. Annual mean precipitation is 1365 mm (Lim, Shin, Jin, Chun, & Oh, 2003). The DBF site

Seasonal evolution of LAI at the DBF site

Three independent estimates of LAI derived from LED, LAI-2200, and upward-pointing digital cameras showed consistent seasonal patterns and magnitudes of LAI at the DBF site (Fig. 2). LED-sensors captured different seasonal evolutions of the over- and understory LAI. LED-sensor-derived leaf-out dates of the over- and understory were DOY 124 and 116, respectively, which were almost identical to those based on in-situ observations (DOY 126 and 115) (Fig. 3). Understory LAI ranged from 0.5 to 1,

Are phenological metrics in multiple canopy layers consistent between sensors and between indices?

There were notable discrepancies in phenological metrics across both sensors and indices. We start by discussing the upward-pointing digital camera results. Camera installed on the forest floor at the DBF site was only able to identify leaf-out and full-leaf dates for the understory canopy (Fig. 3). We speculate that the longer length of the path from the camera to the overstory canopy prohibited detection of the overstory phenology. To determine LAI from digital cameras, a total canopy

Summary and conclusions

To monitor multi-layer canopy phenology separately, we employed LED-sensors at four different canopy heights in a deciduous broadleaf forest and an evergreen needleleaf forest. LED-sensor enabled us to monitor NDVI and LAI concurrently at different canopy heights. Key findings are as follows: 1) LED-sensors identified leaf-out and full-leaf dates for over- and understory canopies at both sites; understory canopies unfolded their leaves 8–11 days earlier than the overstory canopies; 2) LED-LAI

Acknowledgment

This study was supported by Korea-Americas Cooperation Program through the National Research Foundation of Korea (NRF) funded by the Korean Ministry of Education, Science and Technology (NRF-2011-0030485) and the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (2009-0083527), and the BK21 Plus Project in 2014 (Seoul National University Interdisciplinary Program in Landscape Architecture, Global leadership program towards innovative green infrastructure).

References (72)

  • A. Huete et al.

    Overview of the radiometric and biophysical performance of the MODIS vegetation indices

    Remote Sensing of Environment

    (2002)
  • K. Hufkens et al.

    Linking near-surface and satellite remote sensing measurements of deciduous broadleaf forest phenology

    Remote Sensing of Environment

    (2012)
  • H. Kobayashi et al.

    On the correct estimation of gap fraction: How to remove scattered radiation in gap fraction measurements?

    Agricultural and Forest Meteorology

    (2013)
  • A.R.G. Lang et al.

    Estimation of leaf area index from transmission of direct sunlight in discontinuous canopies

    Agricultural and Forest Meteorology

    (1986)
  • B.Y.H. Liu et al.

    The interrelationship and characteristic distribution of direct, diffuse and total solar radiation

    Solar Energy

    (1960)
  • C. Macfarlane

    Classification method of mixed pixels does not affect canopy metrics from digital images of forest overstorey

    Agricultural and Forest Meteorology

    (2011)
  • S. Nagai et al.

    Field experiments to test the use of the normalized-difference vegetation index for phenology detection

    Agricultural and Forest Meteorology

    (2010)
  • S. Nagai et al.

    Utility of information in photographs taken upwards from the floor of closed-canopy deciduous broadleaved and closed-canopy evergreen coniferous forests for continuous observation of canopy phenology

    Ecological Informatics

    (2013)
  • K.N. Nasahara et al.

    Vertical integration of leaf area index in a Japanese deciduous broad-leaved forest

    Agricultural and Forest Meteorology

    (2008)
  • J. Pisek et al.

    Mapping forest background reflectivity over North America with Multi-angle Imaging SpectroRadiometer (MISR) data

    Remote Sensing of Environment

    (2009)
  • M. Rautiainen et al.

    Seasonal reflectance dynamics of common understory types in a northern European boreal forest

    Remote Sensing of Environment

    (2011)
  • Y. Ryu et al.

    Testing the performance of a novel spectral reflectance sensor, built with light emitting diodes (LEDs), to monitor ecosystem metabolism, structure and function

    Agricultural and Forest Meteorology

    (2010)
  • Y. Ryu et al.

    Evaluation of land surface radiation balance derived from Moderate Resolution Imaging Spectrometer (MODIS) over complex terrain and heterogeneous landscape on clear sky days

    Agricultural and Forest Meteorology

    (2008)
  • Y. Ryu et al.

    On the correct estimation of effective leaf area index: Does it reveal information on clumping effects?

    Agricultural and Forest Meteorology

    (2010)
  • Y. Ryu et al.

    How to quantify tree leaf area index in a heterogeneous savanna ecosystem: A multi-instrument and multi-model approach

    Agricultural and Forest Meteorology

    (2010)
  • Y. Ryu et al.

    Continuous observation of tree leaf area index at ecosystem scale using upward-pointing digital cameras

    Remote Sensing of Environment

    (2012)
  • H.P. Schmid et al.

    Measurements of CO2 and energy fluxes over a mixed hardwood forest in the mid-western United States

    Agricultural and Forest Meteorology

    (2000)
  • C. Song et al.

    Classification and change detection using Landsat TM Data: When and how to correct atmospheric effects?

    Remote Sensing of Environment

    (2001)
  • O. Sonnentag et al.

    Tracking the structural and functional development of a perennial pepperweed (Lepidium latifolium L.) infestation using a multi-year archive of webcam imagery and eddy covariance measurements

    Agricultural and Forest Meteorology

    (2011)
  • O. Sonnentag et al.

    Digital repeat photography for phenological research in forest ecosystems

    Agricultural and Forest Meteorology

    (2012)
  • K. Soudani et al.

    Ground-based Network of NDVI measurements for tracking temporal dynamics of canopy structure and vegetation phenology in different biomes

    Remote Sensing of Environment

    (2012)
  • R. Suzuki et al.

    NDVI responses to the forest canopy and floor from spring to summer observed by airborne spectrometer in eastern Siberia

    Remote Sensing of Environment

    (2011)
  • C.J. Tucker

    Red and photographic infrared linear combinations for monitoring vegetation

    Remote Sensing of Environment

    (1979)
  • E.F. Vermote et al.

    Atmospheric correction of MODIS data in the visible to middle infrared: First results

    Remote Sensing of Environment

    (2002)
  • Y.Q. Zhang et al.

    Determining digital hemispherical photograph exposure for leaf area index estimation

    Agricultural and Forest Meteorology

    (2005)
  • X.Y. Zhang et al.

    Monitoring vegetation phenology using MODIS

    Remote Sensing of Environment

    (2003)
  • Cited by (96)

    • Tracking diurnal to seasonal variations of gross primary productivity using a geostationary satellite, GK-2A advanced meteorological imager

      2023, Remote Sensing of Environment
      Citation Excerpt :

      CRK consists of Oryza Sativa L, Odae, which is harvested one time per year. GDK is also part of KoFlux and AsiaFlux and is located in the mid-western part of the Korean Peninsula (Ryu et al., 2014; Yun and Chun, 2018), which consists of Quercus acutissima, Quercus serrata, and Carpinus laxiflora. The HOC site consists of Larix kaempfere, Styrax obassia, and Cornus controversa.

    View all citing articles on Scopus
    View full text