*2.1. Study Area*

To find out the response of phenological events according to the climatic condition, we selected three areas different in land use intensity, such as urban, rural, and natural areas on the same latitude. The phenological signal is evident in deciduous broad-leaved forests because the changes in the canopy are large depending on the stage of the phenology [60]. In this study, therefore, we selected the target species as *Quercus mongolica* Fisch. ex Ledeb., a dominant species in the deciduous broad-leaved forest of Korea. Since *the Q. mongolica* that belongs to the *Quercus* genus grows at the highest elevation in South Korea, it is considered to be sensitive to temperature rises due to climate change. Seven sites including Mts. Nam, Mido, and Umyeon in the urban center, Mts. Cheonggye and Buram in the suburbs, Gwangneung (Mt. Sori) in a rural area, and Mt. Jeombong in a natural area were selected for analysis (Figure 1, Table 1). The urbanization ratio was calculated from the ratio of development area to the total area within a 5 km radius from the study area by a geographic interpolation system (GIS) using the national land use map (National Geographic Information Institute, 2016). The urbanization rates of Mts. Nam, Mido, Umyeon, Cheonggye, Buram, and Sori (Gwandneung) were 76.07, 70.35, 52.44, 35.80, 49.60, and 6.38, respectively, and Mt. Jeombong was not urbanized at all as it is located in a natural area (Table 2).

**Figure 1.** A map showing the location of the study sites.


**Table 1.** Description of study sites. *DoY*: day of year.

**Table 2.** Urbanization rate in study sites.


#### *2.2. Digital Camera and Satellite Image Acquisition*

We installed digital cameras (Model Ltl-6210M, Little Acorn Outdoors, Denmark, WI, USA) near the top of each tower or tree, looking north and angled slightly downward, providing a view across the canopy. To acquire daily photos, each camera was set to record JPEG images three times per day (09:00, 12:30, and 14:30). In order to maintain consistency, only 12:30 images were used for the analysis (Table 3).


As the notion the shorter the collection cycle, the higher the clarity applies [61], we used MODIS (MODerate-Resolution Imaging Spectroradiometer) 500 m resolution land surface imagery (MOD09GA), which is supplied at daily intervals as multi-spectral satellite images. The MODIS satellite is a payload scientific instrument placed in the earth's orbit by NASA in December 1999 on the Terra (EOS AM) satellite. The MODIS satellite imagery measures the surface temperature of the land and ocean, and the distribution map of the earth's vegetation is re-synthesized with control variables such as clouds and distributed free of charge by NASA. The MODIS satellite is suitable for monitoring phenological changes because the sensor incorporates enhanced cloud detection, atmospheric correction, georeferencing, and the ability to monitor vegetation [39,62].

#### *2.3. Digital Camera Image Analysis*

An annual vegetation phenological cycle inferred from remote sensing is characterized by four stages that define the key phenological phases at annual time scales: (1) green-up, (2) maturity, (3) senescence, and (4) dormancy [17,60,63]. Phenological signals remain low values during the dormancy phase and then increase rapidly as the green-up phase begins. When the leaves reach maturity, signals no longer increase but maintain a high value. After that, the plants enter into the senescence phase, and the signals decrease radically. As the dormancy phase begins, the phenological signals return to the lowest value of the initial phase. As such, an inflection point at which the curvature rapidly changes in the phenological signal curve may be interpreted as the start date of each stage [17,63]. The formula for obtaining the curvature K value of the inflection point is as follows:

$$f(t) = a + \frac{c}{1 + \exp(a + bt)} + d \tag{1}$$

$$K = \frac{f''(t)}{\left(1 + \left(f'(t)\right)^2\right)^{\frac{3}{2}}} \tag{2}$$

where *t* is time, *c* is the amplitude of an increase or decrease in the green value, *d* is the baseline value of the dormant season, and *a* and *b* control the lower and upper limits of the function [17,64–66].

To extract phenological signals, we collected images from the digital cameras periodically and classified them into red, green, and blue bands. Using digital numbers for each band, we calculated the average excess green index (ExG) for each ROI based on the equation [64,65]

$$\text{ExG} = \text{2} \times \rho\_{\text{GREEN}} - \left(\rho\_{\text{RED}} + \rho\_{\text{BLUE}}\right) \tag{3}$$

where *ρRED*, *ρGREEN*, and *ρBLUE* are values in red, green, and blue channels acquired from digital camera images, respectively. The region of interest (ROI) is used when digital camera images are analyzed to clarify phenological changes [67]. As the digital camera images include a mixture of the sky, landscape, and other factors, the ROI is limited to the crown layer to extract the phenological signal from the images. Furthermore, because, in these study sties, the *Q. mongolica* stands are mixed with *Quercus variabilis* Blume and *Quercus serrata* Murray stands, and other species, the ROI was limited to pure stands of *Q. mongolica*, and we tried to avoid mountains and sky [60,64] (Figure 2). In this study, we set up a number of ROIs for the images of the *Q. mongolica* community in each site and extracted the ExG index.

**Figure 2.** A field of view from the digital camera at the Mt. Jeombong site. Regions of interest (ROIs) 1–3 are indicated in red.

#### *2.4. Analysis of Satellite Images*

The MOD09GA datasets are comprised of seven bands, including visible light bands and near-infrared bands. The EVI in MODIS was calculated using red (Band 1: 620–670 μm), green (Band 4: 545–565 μm), blue (Band 3: 459–479 μm), and near-infrared (Band 2: 841– 876 μm) based on the equations. The vegetation index uses the EVI index, which is an improvement over other indexes. The EVI index is an improved vegetation index to reduce the effects of spatial differences by using blue bands in areas with large spatial differences and is suitable for observing seasonal changes in vegetation by reflecting the characteristics of the canopy [62]. The EVI calculations used in the analysis are as follows:

$$EVI = 2.5 \times (\rho \text{nir} - \rho \text{red}) / (\rho \text{nir} + (6 \times \rho \text{red} - 7.5 \times \rho \text{blue}) + 1) \tag{4}$$

where *ρnir*, *ρred*, and *ρblue* are values in near-infrared, red, and blue bands. MODIS satellite images were reprojected to TM (transverse Mercator) coordinates because they use a sinusoidal projection. Based on the extracted data, the EVI index for each study site was derived. Then, the EVI was obtained using the smooth curve fitting method to remove variation and to gather trends because interpretation error can occur due to data errors and variation depending on weather conditions [17]. In this study, the EVI was smoothed to the 80th percentile using an exponentially weighted moving average (EWMA). The EWMA was defined as

$$S\_l = \mathfrak{a} \times \mathbb{Y}\_l + (1 - \mathfrak{a}) \times S\_{l-1} \text{ ( $t \gg 1$ ,  $S\_l = \mathcal{Y}\_l$ )}\tag{5}$$

where *t* is the day of year (DoY); *St* is the EWMA value at the DoY; *Yt* is the EVI value at the DoY; and *α* is the smoothing coefficient.

#### *2.5. Sap Flow Measurement*

To analyze the relationship between the phenological signal and the physiological responses of plants, we collected data from a sap flow measurement instrument (model SFM1 Sap Flow Meter, ICT international, Armidale, Australia) installed in the study sites. Measured individuals were randomly selected from individuals included in the ROI. Sap flow velocity (cm<sup>3</sup>·hr−1) was calculated by heat pulse, and temperature was measured from the thermistor inserted 7.5 mm and 22.5 mm inside the removed bark [68]. The seasonal trajectory of sap flow was interpreted using curvature K (formula 2) based on the daily sap velocity. The transition date of the sap flow was compared with the phenological transition date obtained from the digital camera and MODIS installed at the same site.
