Development of a footprint description tool utilizing SMEAR Estonia eddy-covariance data and footprint modelling in combination with remote sensed forest species and land cover data

SMEAR Estonia (58.2714°N, 27.2703°E, 36 m a.s.l.) is situated at the Järvselja Experimental Forestry Centre. The forest ecosystem within the station’s FFP is a hemiboreal forest comprising, silver birch (Betula pendula Roth) and downy birch (Betula pubescens Ehrh.), Scots pine (Pinus sylvestris L.), Norway spruce (Picea abies (L.) H. Karst.), alder species (Alnus spp.), and common aspen (Populus tremula L.). The mean annual temperature in the area varies between 4 °C and 6 °C, the annual precipitation is 500–750 mm with about 40–80 mm as snow, and the growing season length is about 200–220 days (Noe et al., 2016; Kollo et al., 2023). The experimental site consists of a 130 m tall main flux tower, the main cottage for power supply, internet access, a storage for online data, and the pumping facilities and gas analysers for the flux tower (Noe et al., 2015). Different atmospheric greenhouse gases, such as carbon dioxide (CO₂), water vapour (H₂O) and methane (CH₄) are measured together with reactive trace gases, such as ozone (O₃), nitrogen oxides (NO_x = NO + NO₂) and sulphur dioxide (SO₂).

All flux data were collected at SMEAR Estonia (Noe et al., 2015) by continuous high-frequency (10 Hz) measurements (Noe et al., 2021). Raw data from the flux tower were stored at half-hour intervals and cover the years from 2015 to 2020. Half-hourly data files were organised into separate years as input data. The eddy covariance system consists of a sonic anemometer (METEK uSonic-3 Class A) and an infrared gas analyser (Licor LI-7200, Lincoln, NE, USA) deployed at two heights, 30 m and 70 m a.s.l. at the flux tower. Data to determine the FFP, horizontal wind direction (wd) and speed (u), for advection and diffusion of gases and particles in this region – the surface friction velocity (u*), the standard deviation of lateral wind speed deviations (σ_v), and Obukhov length (L) were obtained by using anemometer readings and applying EddyPro software (LI-COR, Lincoln, NE, USA). A set of basic parameters (Table 1) is needed to calculate the FFP plus the displacement height (d), which was chosen constant at 13.4 m determined by the average canopy height of the area and the measurement height above ground (zm) which depends on the height of the receptor – either 30 or 70 m.

To calculate the annual FFP climatology, we used the stand-alone Python version of the FFP model (Kljun et al., 2015).

The workflow of the FFP (Figure 1) calculation was as follows: using the Kljun FFP model (Kljun et al., 2004, 2015) each year’s FFP were calculated separately to get the area and shape of the FFP at the 10%, 50%, 60%, 70%, 80% and 90% contours of cumulative source weights (Figure 2). We chose these percentages to get a more detailed description of the FFP. We first filtered out all non-available (NaN) values, then eliminated data where friction velocity (u*) parameter was smaller than 0.2 to avoid the use of non-turbulent atmospheric conditions. The next step was checking wind direction values for proper range (0–360 degrees) and after that we created the needed input vectors for the Kljun model algorithm. Finally, we run the FFP algorithm to get the x-y distances from the tower for each contour, which then were saved separately for each year in csv format files to further use them as input to the geographic information system (GIS). For this work we used data measured at 30 and 70 m.

Figure 1.

Figure 2.

To compare to the annual FFP shape and area we assessed the annual heterogeneity of the horizontal wind regime. For that, we calculated the horizontal wind speed and direction density using Mathematica (Wolfram Research, Inc., Mathematica, Version 12.3.1, Champaign, IL, USA) for both measurement heights and each year of our experiment.

Basal data of forest stand parameters related with the stand height increment and volume stock are usually measured at 5-year intervals in Estonia. Remotely sensed LIDAR height measurements have a frequency of 4 years (https://geoportaal.maaamet.ee/). Therefore, it is necessary to assess the yearly figures of forest stand parameters by modelling that allows to map the years when no actual measurement took place.

We therefore modelled the increase in the height of all stand elements using the models for normal forest stands as proposed by Kiviste & Kiviste (2009). The volume stock was modelled by using the national regulations according to the Forest Inventory Act (2009).

To create a forest mask, we used everything from the woody vegetation layer of the Base Map from the Estonian Topographic Data Collection (Maa-amet, 2017) and added from the wetland layer the areas with woody vegetation (column PUIS_T contains the value “Yes”). Since the forest surface layer also contains the surface areas of infrastructure elements (roads, ditches, railways, power lines, and quarter boundaries) which are usually mapped in GIS layers as so-called “line type elements” with no spatial extent we needed to determine the surfaces of such line elements. Therefore, it was first necessary to generate for each line type a fitting surface type representing the area that can then be added as a new layer into the forest mask. In this work, we selected only the layers of line elements that are passing through forests. For this purpose, we introduced a layer of roads, layer of ditches, layer of railways, layer of power lines and a layer of forest quarter boundaries. For roads, ditches, railways, and forest quarter boundaries we used again information that can be retrieved from the Base Map. For power line routes we used data from the national transmission operators Elering and Fortum. To avoid possible double accounting, we deleted areas overlapping with the surfaces of line elements from the forest layer.

The cumulated FFP covers a surrounding area of up to 600 m from the main tower if we use flux data measured at 30 m height (Figure 3). In 2015 the FFP covered 61.6 ha, in 2016 it covered 65.4 ha, in 2017 60.2 ha, in 2018 62.3 ha, in 2019 and 2020 the FFP covered areas of 61.4 ha and 58.3 ha, respectively. The average FFP area over the six-year period was 61.5 ha. The source area naturally depends on three main parameters: measurement height, wind speed and direction (Figure 3). However, the area is not only sensitive to the previous factors, but is also dependent on surface roughness and atmospheric stability (Vesala et al., 2008). The general shape of the FFP remained intact over the different years, although the FFP climatology is of a slightly different area for each year mostly due to changes in wind speed and direction. Figure 4 gives additional information on the density of wind data, where the darker colour shows a higher density of input data to the FFP model calculation. The FFP area in 2016 was 6.3% bigger than in 2015, but in 2017, it was smaller by 7.9% than in 2016. In 2018, the area was again bigger than in 2017 by 3.5% and was again smaller in 2019 than in the previous year by 1.5%. In 2020, the area was 5.0% smaller than in 2019. On average, the difference over the years was 4.8%. The smallest and the biggest area covered by flux data appeared to be in 2017 and 2020, respectively.

Figure 3.

Figure 4.

89.4 % of the FFP area regarding the 30 m high measurement point is covered by forests (Table 2). The main tree species growing in the area are the most widespread and economically most valuable species in Estonia: Scots pine (Pinus sylvestris) which cover 28.2 ha on average, silver and downy birch (Betula spp.) 6.8 ha on average, Norway spruce (Picea abies) 15.4 ha on average, common aspen (Populus tremula) 4.3 ha on average, grey (Alnus incana (L.) Moench) and black alder (Alnus glutinosa (L.) Gaertn.), 0.1 ha and 0.2 ha on average, respectively. Beside forest, there are other land types in the FFP as well (Table 2), for example, power lines, buildings like the SMEAR main cottage and the shelter for power supply of the station, roads and ditches, and some clear areas and isles between the forest quarters. The main forest site types in the FFP are mineral dump, Myrtillus, Uliginosum, Filipendula, Oxalis, Oxalis-Myrtillus, Oxycoccus and drained swamp (Table 3).

Growing stocks and yearly increment were calculated for each year and are listed in Table 4. Depending on the year, the figures decreased or increased. The reasons for the decrease and increase might be different: 1) thinning and clear-cutting; 2) smaller/larger area of the FFP for a particular year. If the FFP is smaller, it means that the area where fluxes are detected from is closer to the flux tower. Therefore, the forest growing on the far edges of the FFP gets out of sight of the flux tower. Given there was no thinning/clear-cutting brought about the change in the growing stock perhaps due to shifting stands with different growing stock in or out of the FFP area. Thinning was done only in 2018 and 2019 when 273 m³ and 501 m³ were cut (Table 4), thus such small amounts had no significant impact on the shape of the FFP.

The area of the FFP measured from a height of 70 m was significantly bigger (Figure 5) and covers the surrounding area of up to 4 km from the main tower. In 2015, the FFP covered 3,288 ha, in 2016 3,317.3 ha, in 2017 3,241.8 ha, in 2018 3,332.7 ha, in 2019 and 2020 the FFP covered areas of 3,323.8 ha and 3,272.3 ha, respectively. The average FFP area over the six-year period was 3,296 ha. The main shape of the FFP remained intact, although the FFP climatology were slightly different for each year as well as for the smaller 30 m FFP. The FFP area in 2016 was 0.9% bigger than in 2015, but in 2017 it was smaller by 2.3% than in 2016. In 2018, the area was again bigger than in 2017 by 2.8%, and the FFP area was again smaller in 2019 than in the previous year by 0.3%. In 2020, it was smaller by 1.6% than in 2019. On average, the difference over the years was only 1.6%. The main growing tree species in the FFP area are Scots pine (511.2 ha on average), silver and downy birch (1,007.9 ha on average), Norway spruce (450 ha on average), common aspen (242.7 ha on average), grey and black alder 37.4 ha and 464.5 ha on average, respectively. Bogs and fens cover 86 ha of the area and 40.7 ha of them are covered by Scots pine. Overall, 2,897.3 (87.9 %) ha of the FFP area is categorized as forest land, the remaining 398.7 ha are covered with different types of land (12.1 %) (Table 5). The main forest site types for this FFP are described in Table 6.

Figure 5.

Growing stocks were calculated for each year (Table 7), in some years the growing stock decreased or increased as compared to other years. The reasons behind these changes are similar to those given in the previous section.

At 70 m height, the dominant wind directions in 2015 to 2017 ranged from the southwest to south, like the 30 m anemometer measurements showed. In the more recent years, the general wind directions were also frequently from the south to southeast, but also northeast wind directions are prevalent at this height. The darker colour on the figure denotes a higher density in wind direction and speed and therefore a higher contribution from those directions to the overall FFP (Figure 6). Thus, wind speed and directions in those particular regions are more abundant which makes the input to calculate the overall FFP shape more robust.

Figure 6.