International Journal of Applied Earth Observation and Geoinformation
Estimation of sub-canopy solar radiation from LiDAR discrete returns in mixed temporal forest of Białowieża, Poland
Graphical abstract
Introduction
Solar radiation drives numerous biological and ecological processes that are essential to life. For instance, it affects the rate of photosynthesis and transpiration of plants from which other biotic elements depend upon. In forest understory, solar radiation also influences regeneration and understory growth (Grant, 1997; Anderson and Denhead, 1969; Van der Zande et al., 2010; Mücke and Hollaus, 2011; Sakai and Akiyama, 2005), soil conditions (Musselman et al., 2013; von Arx et al., 2012) and snow melting phenomena (Golding and Swanson, 1978; Musselman et al., 2013). Insects and micro-organisms in the forest also have a range of light preferences in order for them to survive (Battisti et al., 2013; Weiss et al., 1991). The level of radiation that penetrates the canopy is therefore an essential input for modeling these domain of studies. In addition, knowledge about the amount of light in the forest can be used to calculate the canopy transmittance which is defined as the fraction of radiation that will penetrate at a particular site relative to the radiation above the canopy (Anderson, 1964). Transmittance is useful in developing stand management prescriptions to create favorable conditions for the growth of understory trees (Comeau, 2000; Jennings, 1999). In spite all of these importance, efforts to calculate below canopy solar radiation with high spatial resolution at wider scale are scarce.
Quantifying light conditions for vegetation studies is normally measured from 400 to 700 nm of the electromagnetic spectrum known as the photosynthetically active radiation (PAR). Since photosynthesis is more closely related to the amount of photons per unit area rather than its total energy, PAR is normally expressed as photosynthetic photon flux density (PPFD) with units of μmol m−2 s−1 (Biggs, 1986; McKree, 1981). For in situ measurements, PAR sensors are used that are based on photodiodes equipped with filters to restrict their response range (Schleppi and Paquette, 2017). These sensors are either stationary which can record PAR at point location over a period of time or as handheld instrument for sub-meter measurements (as in the case of sensors in arrays) (Comeau, 2000; Schleppi and Paquette, 2017). Another method of measuring PAR is the use of hemispherical photography (HP). A widely used technique in forestry studies, HP involves photo acquisitions by looking upward from beneath a forest or other plant canopy. The geometric analysis of HP is used to characterize the light environment under the canopy or the attributes of the canopy itself (Fournier and Hall, 2017; Rich, 1989). While each of these instruments have their own strengths and weaknesses, their ground-based operations have a lot of common which is time-consuming, labor intensive and expensive to some extent. For example, in order to increase the spatial coverage of the measurement or to lessen the variability of the sunlight, large amount of samples are needed by either adding more sensors placed at different locations or moving from one plot to another (Comeau, 2000). Further, it may also be impractical for areas difficult to access (Olpenda, 2018).
The technology of remote sensing (RS) and geographic information system (GIS), combined with field data, have proven to be useful in generating large-scale spatial information of sunlight even below canopies (e.g. Bode et al., 2014; Riaño et al., 2004). A critical pre-requisite is needed though which is the calibration of RS dataset relative to the ground data. The relationship between these two data sources must be established at optimum potential of the sensor describing the characteristics of forest canopy. Previous related studies have shown that optical images were found to be less useful due to low spatial resolution and saturation of spectral bands with high LAI values (Breda, 2003; Jensen et al., 2008; Yamamoto et al., 2015). According to a review (Olpenda et al., 2018), among the type of sensors that researchers used for analyzing understory light condition, the light detection and ranging (LiDAR) is the most preferred one. Since the incoming solar radiation is mainly affected by the canopy structure, it is therefore rational to have details about it as high as possible of which LiDAR can offer (Jupp et al., 2008; Lefsky et al., 2002; Zhao and Popescu, 2009). One advantage of the laser technology is its capability to produce 3D information of the canopy and then simulate light attenuation through the Beer’s Law. However, such principle requires parameters that are species specific (Bolibok et al., 2015; Kobayashi et al., 2012). It also assumes that the forest canopy is a turbid medium which is appropriately applicable to homogeneous environment but impractical in natural, heterogeneous systems where species diversity is high (Bode et al., 2014; Kobayashi et al., 2012; van Leeuwen et al., 2013). Another common approach of modeling sub-canopy radiation or transmittance with LiDAR is the use of derivatives (e.g. Riaño et al., 2004; Yamamoto et al., 2015) that are generated at different levels of processing (Evans et al., 2009). First order derivatives include the digital elevation model (DEM) and digital surface model (DSM) where simulation of light regime is conducted usually in a GIS platform using solar radiation tools (e.g. Bolibok et al., 2016, 2015). There are several, hundreds in fact, second order LiDAR derivatives that can be used for the analysis but the most commonly cited are the ratio metrics particularly the laser penetration index (LPI) (Hopkinson and Chasmer, 2007; Olpenda et al., 2018; You et al., 2017). Barilotti et al. (2006) is one of the first to introduce LPI for leaf area index (LAI) estimation and defines it as the ratio of the points that reach the ground over the total points in an area. Bode et al. (2014) on the other hand used LPI to approximate understory solar radiation (in watts m−2) by treating the laser beams analogous to the sun rays that penetrates the canopy. The model was successful, at least to direct and total radiation, by generating the product of LPI and the solar radiation above canopy. Since the latter is unavailable for this research, we turned to traditional approaches and modified some of the previous techniques.
Multiple metrics have been previously utilized for the estimation of LAI and transmittance but not on the level of radiation itself. Moreover, past models’ effectivity relied on the height threshold for canopy identification in LiDAR and how large should the area be (e.g. plot radius) for extracting the dataset. Many authors of LAI and transmittance studies have varying interpretations on canopy height threshold ranging from 1 m to as high as 3.6 m (e.g. Barilotti et al., 2006; Zhao and Popescu, 2009). Others preferred to use the diameter at breast height (DBH) (e.g. 1.37 m) or the camera height. Though McCallum et al. (2014) recommends height cut-offs for two distinct forest types, there are still no clear standards agreed upon by the scientific community. Meanwhile, plot radius for ALS extraction were suggested by various authors, albeit mostly for LAI studies. Riaño et al. (2004) reported that LAI was better predicted using LiDAR metrics from 10 to 12.5 m radii for pine forest. On the other hand, Zhao and Popescu (2009) estimated LAI by regression and found that the best predictor can be derived using a longer plot radius of 25 m. Yamamoto et al. (2015) estimated relative illuminance in Cypress stands derived from the HP-measured transmittance and found a shorter range of effective radius at 7.5–15 m. Aside from the LiDAR data quality (e.g. number of hits per m2), it seems that the performance of these metrics is also a function of the forest structure being studied. The use and combinations of LiDAR metrics and the selection of the best plot radius and canopy height threshold therefore are yet to be validated in a mixed type of forest in Poland.
The ultimate goal of this paper was to develop regression models of sub-canopy sunlight condition from LiDAR metrics in heterogeneous forest under various protection regimes and silviculture strategies. We deliberately modeled not only the direct or total radiation but more importantly the diffuse component. Very few RS-based studies have been conducted that focused on the diffuse sunlight despite its prevailing roles to plants for gaining higher productivity and efficiency (Gu et al., 2002; Li et al., 2014; Olpenda et al., 2018). Though Bode et al. (2014) included to model the diffuse radiation in their analysis, the predictive value was low (R2 = 0.30). Aside from being more strongly related to PAR compared to its direct counterpart (Lieffers et al., 1999; Yamamoto et al., 2015), the total transmittance can be characterized by a single measurement of diffuse radiation during growing season (Parent and Messier, 1996 in Hale et al., 2009). In this paper, field-based calibration of the model was done through hemispherical photography where the final output is the average daily PPFD. Hemispherical photographs were also used by Hopkinson and Chasmer (2007) (R2 = 0.92) and Yamamoto et al. (2015) (R2 = 0.95) for calibrating their respective transmittance models from LiDAR and did not use any optical sensors.
Section snippets
Study area
The study was conducted in the Polish portion of Białowieża Forest (BF) which is about 190 km east-northeast of Warsaw, Poland (52° 43’ 39”, 23° 53’ 57”) (Fig. 1). Composed of diverse forest communities, this forest complex of almost 60,000 ha is situated in the transition between the boreal and temperate zone (UNESCO, 2014). BF is generally flat in terms of topography with an altitude ranging from 135 to 185 m a.s.l. The mean annual sum of precipitation in the area is 627 mm while the average
Sunlight measurements from hemispherical photographs
Out of the expected 500 hemispherical photos (5 photos × 100 plots), only 390 were selected as those with proper exposure and suitable quality for sunlight analysis. These resulted to the exclusion of four plots. Table 2 shows the final distribution of the screened photos that were analyzed. Out of the total 96 plots left for the analysis, only 52 had complete 5 photos equivalent to 260 photographs.
Basic statistics of all components of sunlight are shown in Table 3. It can be seen how high the
Effects of using multiple photographs
Multiple acquisitions (5 photos per plot) were critically important for the model because methods of measurements done by HP are different from LiDAR metrics. Due to lens projection, HP technique is actually assessing canopy closure which is defined as the proportion of the sky hemisphere obscured by vegetation when viewed from a single point (Jennings, 1999). In other words, HP-derived sunlight is integrated from the whole hemisphere that the camera “saw” because fish-eye lens has an angle of
Conclusion
The recognition of ratio metrics in this paper being the most effective predictor for understory sunlight was consistent with other studies. LPI alone was sufficient to model the sunlight below canopy, most especially the diffuse component. Previous attempts in estimating sunlight below canopy at high resolution maps had poor model performance. The usage of the default height threshold used in this study corresponding to the camera height (as what many previous studies followed as well) did not
Recommendations
Based on the results and experiences encountered during the conduct of this study, we enumerate the following initiatives for the improvement of the model.
- 1
The addition of more samples is strongly suggested to make the model more robust and increase the precision statistically. Increasing the sample size will also prompt the allotment of validation sets. Further, increasing also the number of hemispherical photographs to acquire per plot is proposed (e.g. from 5 to 9).
- 2
A point of reference for
Author contributions
A.O. and K.S. conceived the ideas and designed the methodology; A.O. analysed the data, performed the analysis and led the writing of the manuscript. Both K.S. and K.B. contributed critically to the drafts and gave final approval for publication. K.S. was responsible for gaining financial support for the project leading to this publication.
Conflicts of interest
The authors declare no conflict of interest.
Acknowledgments
This work was supported financially by the Project LIFE+ ForBioSensing PL “Comprehensive monitoring of stand dynamics in the Białowieża Forest as supported by remote sensing techniques”. The Project has been co-funded by Life + (contract number LIFE13 ENV/PL/000048) and Poland’s National Fund for Environmental Protection and Water Management (contract number 485/2014/WN10/OP-NM-LF/D). We would like to thank Małgorzata Białczak, Rafał Sadkowsk, Bartłomiej Kraszewski and Maciej Lisiewicz for
References (86)
- et al.
Airborne laser scanner (LiDAR) proxies for understory light conditions
Remote Sens. Environ.
(2013) - et al.
Subcanopy Solar Radiation model: predicting solar radiation across a heavily vegetated landscape using LiDAR and GIS solar radiation models
Remote Sens. Environ.
(2014) - et al.
Comparison of five canopy cover estimation techniques in the western Oregon Cascades
Forest Ecol. Manage.
(2006) The effect of thinning intensity on the below-canopy light environment in a Sitka spruce plantation
For. Ecol. Manage.
(2003)- et al.
Comparison of film and digital hemispherical photography across a wide range of canopy densities
Agric. For. Meteorol.
(2002) - et al.
Discrete return lidar-based prediction of leaf area index in two conifer forests
Remote Sens. Environ.
(2008) - et al.
Review of methods for in situ leaf area index determination
Agric. For. Meteorol.
(2004) - et al.
Modeling energy and carbon fluxes in a heterogeneous oak woodland: a three-dimensional approach
Agric. For. Meteorol.
(2012) - et al.
On the correct estimation of gap fraction: how to remove scattered radiation in gap fraction measurements?
Agric. For. Meteorol.
(2013) - et al.
Methodology comparison for canopy structure parameters extraction from digital hemispherical photography in boreal forests
Agric. For. Meteorol.
(2005)
Classification method of mixed pixels does not affect canopy metrics from digital images of forest overstorey
Agric. For. Meteorol.
Photographic exposure affects indirect estimation of leaf area in plantations of Eucalyptus globulus Labill
Agric. For. Meteorol.
Estimation of LAI and fractional cover from small footprint airborne laser scanning data based on gap fraction
Remote Sens. Environ.
Estimation of solar direct beam transmittance of conifer canopies from airborne LiDAR
Remote Sens. Environ.
Evaluating models of shortwave radiation below Eucalyptus canopies in SE Australia
Agric. For. Meteorol.
Estimation of leaf area index and covered ground from airborne laser scanner (Lidar) in two contrasting forests
Agric. For. Meteorol.
Quantifying the spatio-temporal variability of net primary production of the understory species, Sasa senanensis, using multipoint measuring techniques
Agric. For. Meteorol.
Important LiDAR metrics for discriminating forest tree species in Central Europe
ISPRS J. Photogramm. Remote. Sens.
Automated reconstruction of tree and canopy structure for modeling the internal canopy radiation regime
Remote Sens. Environ.
Spatio-temporal effects of forest canopy on understory microclimate in a long-term experiment in Switzerland
Agric. For. Meteorol.
Determining digital hemispherical photograph exposure for leaf area index estimation
Agric. For. Meteorol.
Lidar-based mapping of leaf area index and its use for validating GLOBCARBON satellite LAI product in a temperate forest of the southern USA
Remote Sens. Environ.
Studies of the woodland light climate I. The photographic computation of light conditions
J. Ecol.
Radiation and crop structure
Shortwave radiation on inclined surfaces in model plant communities
Agron. J.
LAI determination in forestry ecosystem by LiDAR data analysis
Proceedings of Workshop 3D Remote Sensing in Forestry, Vienna
Solar radiation directly affects larval performance of a forest insect: effects of solar radiation on larval performance
Ecol. Entomol.
Advanced Agricultural Instrumentation: Design and Use
Zmienność całkowitego promieniowania słonecznego na obszarze Polski w okresie 1961–1995. Meteorologia, Materiały Badawcze IMGW
Equipment and photo exposure in hemispherical photography in forest research
For. Res. Pap.
Wpływ wysokości otaczającego drzewostanu, wystawy i pochylenia terenu na mikroklimat radiacyjny na gnieździe – wyniki modelowania
Sylwan
Latitudinal variability of the radiation microclimate in artificial forest gaps in Poland—the modelling perspective
For. Res. Pap.
Ground-based measurements of leaf area index: a review of methods, instruments and current controversies
J. Exp. Bot.
Ligth regimes beneath closed canopies and tree-fall gaps in temperate and tropical forest
Can. J. For. Res.
Estimating basal area and stem volume for individual trees from Lidar data
Photogramm. Eng. Remote Sens.
Digital hemispherical photography for estimating forest canopy properties: current controversies and opportunities
iForest Biogeosci. For.
Tree canopy leaf area index (LAI) measurements with the hemispherical photography at a Tuczno forest
Acta Agrophys.
Measuring Light in the Forest: Technical Report
Light attenuation following selection harvesting in northern hardwood forests
Forest Ecol. Manage.
Collinearity: a review of methods to deal with it and a simulation study evaluating their performance
Ecography
Discrete return Lidar in natural resources: recommendations for project planning, data processing, and deliverables
Remote Sens.
A Guide to LIDAR Data Acquisition and Processing for the Forests of the Pacific Northwest. (No. PNW-GTR-768)
Cited by (8)
QIENet: Quantitative irradiance estimation network using recurrent neural network based on satellite remote sensing data
2024, International Journal of Applied Earth Observation and GeoinformationCanopy gap characteristics and regeneration patterns in the Białowieża forest based on remote sensing data and field measurements
2022, Forest Ecology and ManagementCitation Excerpt :As a result, a map was created for the entire Polish part of the Białowieża Forests. The details of the methodology are described in Olpenda et al. (2019). The variables containing supporting and contextual information (e.g., the percentage of the specific tree species covering the land around the gap or height of vegetation in the 10 m buffer from the GIS gap contour) are included in the browsing and contextual information categories.
Characterizing the three-dimensional spatiotemporal variation of forest photosynthetically active radiation using terrestrial laser scanning data
2021, Agricultural and Forest MeteorologyCitation Excerpt :Obtaining accurate estimates of the 3-D spatiotemporal distribution of forest canopy PAR is essential for studying the photosynthetic process and carbon cycle of vegetation (Li et al. 1997; Running and Coughlan 1988). In most of previous studies, the PAR under the canopy was estimated using coarse-resolution lidar data (Bode et al. 2014; Olpenda et al. 2019; Peng et al. 2014), and methods were validated using pyranometers or digital hemispherical photos placed at heights of 1–1.5 m above the ground. In this study, it was shown that the proposed method could estimate the 3-D spatiotemporal distribution of PAR within and under the canopy at a fine spatial resolution, and the results were compared with measurements from pyranometers placed 0–3 m above the ground, some of which were even placed within crowns at plots 1–3 (Fig. 1).
VEGETATION INFLUENCE ON CANOPY HEIGHT MODELS DERIVED FROM AIRBORNE LASER SCANNING
2021, International Multidisciplinary Scientific GeoConference Surveying Geology and Mining Ecology Management, SGEM