Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Agricultural and Forest Meteorology 325 (2022) 109127

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology


journal homepage: www.elsevier.com/locate/agrformet

Upland reclamation promotes forest evaporative losses in the Boreal Plains


of Canada: A comparison of carbon and water fluxes
M. Graham Clark a, *, Richard M. Petrone b, Sean K. Carey a
a
School of Earth, Environment & Society, McMaster University, Hamilton, Ontario, Canada
b
Geography and Environmental Management, University of Waterloo, Waterloo, Ontario, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: Reclaiming upland boreal forests as part of post-mining watershed construction in the Athabasca Oil Sands
Reclamation Region (AOSR) requires evaluation criteria as these novel ecosystems develop. Here, we analyzed 55 site-years of
Boreal forests eddy covariance observations of constructed forests and soils on formally pit-mined landscapes (9 sites) and
Oil sands
contrasted them to 18 site-years of post-harvested ecosystems (3 sites) and 38 site-years of mature Boreal Plains
Evapotranspiration
Productivity
ecosystems (3 sites). After approximately 5 years, the post-harvested sites had fluxes of gross primary produc­
Eddy Covariance tivity (GPP), evapotranspiration (ET), and water use efficiency (WUE) within the variability of the widely studied
mature Boreal Plains BERMS FLUXNET sites. However, even after 10 years the constructed forests had signifi­
cantly lower WUE than the mature sites. High ET fluxes drove low WUE in the constructed upland conifer sites
despite similar rates of GEP. Conversely, in the constructed broadleaf forests low GEP, despite similar ET,
resulted in low WUE. A climate sensitivity analysis showed that there was little impact of abnormal hot, cold,
wet, or dry growing seasons on GEP or evapotranspiration (ET) at the constructed forests. It is presumed that the
high moisture retaining properties of the soils used in reclamation produced the low WUE and resilience to dry
and hot conditions in constructed forests. Placing moisture retaining soils incorporates a degree of resilience to
climate variability but also limits downgradient water yields to low lying wetlands in the relatively dry AOSR
climate. This highlights a potential shortcoming of reclamation objectives developed for specific ecosystems
when scaling from ecosystem to watershed scale construction. Finally, robust relationships between satellite
observed greenness (MODIS EVI and NDVI) and ecosystem scale fluxes highlight how remote sensing-based
metrics can be used by land managers to identify regions within landscape units that may be under performing.

1. Introduction AOSR reclamation projects to “display characteristics of resilience to


natural disturbances and climatic variability” has no clear criteria upon
Landscapes impacted by oil sands mining in the Athabasca Oil Sand which to evaluate and certify success (Alberta Environment and Sus­
Region (AOSR) are required to be restored to the “equivalent capability” tainable Resource Development, 2013).
of the landscape prior to disturbance (Section 146(b) of the Environ­ The evaluation of forest resilience to climate extremes has extensive
mental Protection and Enhancement Act; Government of Alberta, 2020). precedent. Typically, it involves monitoring productivity and/or water
The current objective is to reconstruct integrated Boreal Plains land­ use in relation to stressors (e.g. Anderegg et al., 2018; a global assess­
scapes, including upland forests, wetlands, and lakes to benchmarked ment of species diversities impact on forest-drought resilience). Resil­
criteria. Once reclamation criteria are met, the operator may apply for a ience in this specific context is the capacity to restore pre-disturbance
reclamation certificate, which releases the owner from further obliga­ growth rates (productivity), water loss/retention (evapotranspiration),
tions (Gosselin et al., 2010). The reclamation criteria in the document or both (water use). Both productivity and evapotranspiration are
“Criteria and Indicators Framework for Oil Sands Mine Reclamation regularly measured through the eddy covariance (EC) technique, the
Certification” (Alberta Environment and Sustainable Resource Devel­ most widespread method for observing ecosystem-atmosphere exchange
opment, 2013) are based on point in time structural qualities of eco­ of water, energy, and carbon. Within the Canadian context, EC has been
systems or are vaguely defined functions. Specifically, the objective of applied to assess the resilience of northern forest systems relative to

* Corresponding author.
E-mail address: dr.mg.clark@gmail.com (M.G. Clark).

https://doi.org/10.1016/j.agrformet.2022.109127
Received 24 January 2022; Received in revised form 14 June 2022; Accepted 8 August 2022
Available online 23 August 2022
0168-1923/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

disturbances such as fire (e.g. Amiro et al., 2003; Mkhabela et al., 2009; 1996) indicate that the study period (2003–2020) included nine
Coursolle et al., 2012), insect bark and leaf damage (e.g. Speckman growing seasons (June-August) with temperatures in the top quartile of
et al., 2015; Stephens et al., 2018), drought (e.g. Kljun et al., 2006; all 71 years of reanalysis growing season temperatures (2006, 2008,
Petrone et al., 2015; Anderegg et al., 2018), and harvesting (e.g. Hum­ 2011–2017), and two years in the bottom quartile (2005, 2018). Using
phreys et al., 2005; Mkhabela et al., 2009; Coursolle et al., 2012; Pet­ the difference between the annual precipitation and PET data from the
rone et al., 2015). In this study EC will be used to contrast ecosystem same reanalysis product indicated the study covered five wet years in
recovery after harvesting to recovery after construction and compare the top quartile (2007, 2014, 2016, 2018, 2020) and four dry years in
both to a mature system. This comparison will include analyzing how the bottom quartile (2004, 2009, 2012, 2015).
climate impacts productivity and water use as a specific metrics of In total 9 constructed forests sites were analyzed in various stages of
resilience. reclamation covering 55 growing seasons (Table 1; Fig. 1). Six of the
An independent expert panel, assembled by the Royal Society of sites were all located on the Syncrude Canada Ltd. Mildred Lake Mine
Canada, encouraged a whole new conceptualization of reclamation to lease: South Bison Hill (CONB-SB30), Sandhill Fen Watershed Upland
deal with the large scales involved in returning open pit AOSR mines to (CONB-SFWUP), Sandhill Fen Watershed Perched (CONB-SFWP), Coke
the equivalence of landscapes prior to disturbance (Gosselin et al., Beach (CONC–CKB), Jack Pine (CONC-JP), and Southwest Sand Stor­
2010). Broadly, this new approach was to perceive the new landscapes age (CONC-SWSS). Three sites were located on the SUNCOR Millennium
as “hydrological response areas”, with ecosystems integrated as specific mine lease: Nikanotee Fen Uplands (CONC–NIU), Wapisiw (CONB-
“hydrological units” or building blocks within the broader landscape WAP), and Cell-11A (CONC–C11A). All sites are within 40 km of Fort
(Devito et al., 2012). This design philosophy hinges on the movement, McMurray (56◦ 43′ 35′′ N, 111◦ 22′ 49′′ W), in northern Alberta, Canada.
and storage of water in the landscape, something that is critical to All nine constructed forests were built as forested upland Boreal
maintain diverse Boreal Plains ecosystems in the cyclic wet/dry climate Plains hydrological units. Various combinations of Jack Pine, White
of the region and to the transport of contaminants of interest that can Spruce, Black Spruce, Balsam Poplar, Trembling Aspen, and Siberian
impact human or ecological health. Thus, the flux of water is something Larch (Larix sibirica) were planted at the nine sites (Table 1). Despite the
closely monitored within the constructed landscapes already and EC variety of species planted at each site, the seasonality of the fluxes from
systems are widely deployed to monitor the evapotranspiration of wet­ the forest sites typically fell into one of two general categories: broadleaf
lands (Nicholls et al., 2016; Ketcheson et al., 2017; Scarlett et al., 2017; and conifer. Conifer sites were CONC-JP, CONC-SWSS, CONC–NIU,
Biagi et al., 2021; Volik et al., 2021, 2021b), lakes (Clark et al., 2021), CONC–CKB and CONC–C11A, which were either planted with domi­
and forests (Carey 2008; Ketcheson et al., 2017; Strilesky et al., 2017; nant covers of Siberian Larch (CONC-SWSS), White Spruce
Biagi et al., 2021; Chasmer et al., 2018; Sutherland et al., 2017), as well (CONC–CKB) or Jack Pine (CONC-JP, CONC–NIU, CONC–C11A).
as to constrain modeling estimates upon which hydrological response Except for CONC–CKB all the other sites with spruce plantings were
areas and units are designed (Lukenbach et al., 2019). The extensive grouped with broadleaf because of the extensive aspen or poplar
monitoring network on constructed landscapes in the AOSR, both
spatially and temporally, may also be used to evaluate the development Table 1
of reconstructed upland forests and contrast that development to nearby A list of sites and features. Cover depth is the designed depth of soil cover over
harvested sites and mature sites. the fill. Fill may be tailings, coke, or some other combination of mine waste
In this study, we seek to illustrate how the EC technique in con­ material. Soil covers are a mix of mineral soils and peat. Initial species planting
structed and harvested landscapes can be used to assess the development densities are reported where available, otherwise just species names are given. A
of productivity and water use, and its variance in response to climate. In planting density of 0 indicates that it was specifically highlighted as natural
this assessment, we analyze an extensive monitoring network containing colonization in Straker et al. (2019) and N/A indicates the data was unavailable.
55 site-years of EC data from nine constructed forest systems in the Site Planted Species Year EC Years Cover
AOSR and compare it to 56 site years of mature and post-harvest (original planting Planted depth
density in stems [cm] (peat
reference sites. The main objectives of this study are: 1) determine the
per hectare) content
time it takes for surface-atmosphere exchanges of carbon and water to [%])
return to levels of natural variability; and 2) assess the resilience of the
CONB-SB30 White Spruce 2003 2003–2015, 100+ (80)
constructed forests by analyzing both productivity and water use in Aspen (1944 2017–2019
years of climatic stress. combined)
CONB- Aspen (0) 2012 2013–2015, 55 (90)
2. Methods SFWUP Balsam Poplar (0) 2017–2019
White Spruce
CONB-SFWP Aspen (0) 2012 2014–2015 75 (90)
2.1. Study sites Balsam Poplar (0)
CONC–CKB White spruce (N/A) 2007 2017 50 (9)
The AOSR is within the Boreal Plains ecozone, which is a landscape CONC-JP Jack pine (2900) 1993 2007–2012 55 (8)
CONC-SWSS Jack Pine 1996 2005–2006, 55 (5)
of subdued relief composed of low-lying, typically peat filled, valleys
White Spruce 2014–2015
and forested plain ecosystems (Wiken et al., 1996). Dominant tree spe­ Balsam Poplar
cies are White Spruce (Picea glauca), Black Spruce (Picea mariana), Trembling Aspen
Balsam Fir (Abies balsamea), Jack Pine (Pinus banksiana), Tamarack Siberian Larch
(Larix laricina), Trembling Aspen (Populus tremuloides), Balsam Poplar (Various
combinations at
(Populus balsamifera), and Birch (Betula spp.) (Wiken et al., 1996). AOSR 2000 total)
climate is dry, with short hot summers and long cold winters (Köppen CONC–NIU Jack pine (572) 2014 2013–2020 35 (0)
Dfc). At the Fort McMurray A meteorological station (Climate ID Black Spruce (900)
3062693), the climate normal (1981–2010) mean temperature was 1 ◦ C CONC–C11A Jack pine (1565) 2006 2017–2018 25 (48)
Trembling Aspen
with an average of 419 mm a− 1 of precipitation. However, annual po­
(400)
tential evapotranspiration (PET) is 520 (±30 SD) mm and the region Paper Birch (200)
typically experiences dry summers that are occasionally punctuated CONB-WAP Aspen (N/A) 2013 2018–2020 51 (N/A)
with wet years (roughly twice every decade) providing some ground­ Balsam Poplar (N/
water recharge (Devito et al., 2012). Global NCEP/NCAR climate A)
Jack Pine (N/A)
reanalysis data (resolution of 1◦ , spanning 1950–2021) (Kalnay et al.,

2
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

Fig. 1. The study region, 40 km north of Fort McMurray. Inset shows the location of the constructed forests in the Athabasca Oil Sand Region (ASOR), the harvested
sites at Utikuma lake (P40) and the mature sites at the BERMS FluxNet sites in Canada. These maps were made with freely available data products from Google Earth
(main image) and Statistics Canada (2016).

overstory in the first decade of growth at each site (Straker et al., 2019). the Boreal Plains ecozone. Although all three regions (AOSR constructed
Worth noting, despite planting jack pine at CONC-SWSS in 1996, by sites, Utikuma Lake harvest sites, and Mature BERMS sites) have the
2014 the CONC-SWSS site was dominated by Siberian larch (11%), same continental subarctic climate regime (Köppen classification of DfC)
aspen (7%) poplar (6%) and understory vegetation (Straker et al., 2019). the relevant differences from the nearest climate station are also pre­
Three harvested sites from the Utikuma Lake area (250 km from the sented here (Table 2).
AOSR sites) were used to contrast the development of the AOSR con­
structed forests. The uplands in the Utikuma Lake study area in which 2.2. Data collection and processing
these forest sites are located is primarily composed of low relief rolling
moraines and the associated glacial overburden soils. Red Earth Creek 1 All flux data was computed using the EC technique (Baldocchi et al.,
(HAR-P40S) and 2 (HAR-P40N) were harvested a year apart, in 2006 1988) by determining the covariance between the 3-dimensional wind
and 2007, respectively. There is 10 years (HAR-P40S) and 4 years (HAR- and gas concentrations at 10 Hz to calculate the 30 min block averaged
P40N) of data post-harvest plus one year of pre-harvest for each site. flux. Density (Webb et al., 1980), as well as low and high pass filtering
There was also an EC tower erected directly within the harvested site at corrections were applied as needed. Gap filling is necessary in EC
HAR-P40N in the 2 years following logging, hereafter called HAR-P4SN.
The harvested forests were all broadleaf forests of Aspen and Balsam and
Table 2
are naturally regenerating. Detailed site information for the harvested
Climatological (1981–2010) differences between the nearest Environment and
forests is given in Brown et al. (2014) and Petrone et al. (2015). Climate Change Canada meteorological station (Fort MacMurray A [ECCC
In addition to the comparison to the harvested sites, gap-filled fluxes Climate ID 3062693] is approx. 30 km south of the constructed sites, Wasbasca
from three mature (i.e. “old growth”) Boreal Plains sites were down­ RS [ID 3076908] is approx. 100 km east of the harvested sites, and Prince Albert
loaded from FluxNet (https://fluxnet.org) for an undisturbed mature [ID 4056240] is approx. 80 km south of the mature sites. *Wasbasca RS has no
flux reference. These sites are part of the Boreal Ecosystem Research and available cloud cover data, so instead cloud data comes from the next closest
Monitoring Sites (BERMS) just over 500 km away from the AOSR and station Peace River A [ID 3075040] ~110 km west of the harvested sites.
cover 15 years of an Aspen dominated forest (MATB-Oas), 12 years of a Region Constructed Harvested Mature
Black Spruce dominant forest (MATC–Obs), and 10 years of a Jack Pine sites
dominant forest (MATC–Ojp). In addition to the Aspin overstorey at Climate station elevation 369.1 544.7 428.2
MATB-Oas, there is 2-m understory of hazelnut and the occasional bal­ Mean annual temperature 1.0 1.7 1.4
sam poplar. The soils at MATB-Oas are Orthic Gray Luvisol. MATC–Obs Mean annual precipitation 418 462.1 427.7
Hours per year of 0–20% cloud cover 2159.2 2107.3* 2548.8
also has ~10% tamarack coverage with hummock-hollow micro
Hours per year of 30–70% cloud cover 2019.4 2223.8* 198.7
topography and feathermoss/lichen forest floors. The soils at Hours per year of 80–100% cloud cover 4587.7 4433.7* 4229.9
MATC–Obs are classified as Peaty (20–30 cm) Orthic Gleysols. Within Days with >= 0.2 mm precipitation 142.5 116.4 121.3
the MATC–Ojp site, there are some isolated groups of green alder and Days with > 5 mm precipitation 23.6 27.7 25.2
the forest floor is mostly lichen and feathermosses and the topography is Days with maximum temperature > 174.7 172.8 175.2
10 ◦ C
negligible. MATC–Ojp soils are classified as either degraded Eutric
Days with minimum temperature > 153.1 155.1 156.2
Brunisol or Orthic Eutric Brunisols. All 24 sites in this study are within 0 ◦C

3
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

processing to determine monthly and seasonal fluxes as all EC systems following the methods of Helbig et al. (2020). Because of the inconsis­
have periods where data needs to be removed (e.g. maintenance, pre­ tency between the seasonal deployment of EC equipment, growing
cipitation events, stability, etc.) and general signal quality control pro­ season fluxes were defined as the period from June to August, inclu­
cessing. A mean of 29.4% of the growing season fluxes were missing and sively, to limit the amount of shoulder season data that would otherwise
required gap filling for seasonal totals (see supplementary table ST1 for require interpolation to produce seasonal fluxes at some sites.
breakdown of gaps by year and site). Gaps were filled with artificial The surface area of photosynthetic activity is directly related to the
neural networks following the process outlined in Clark et al. (2019). ecosystem productivity (Reed et al., 1994; Pettorelli et al., 2005). The
Daily fluxes were calculated only from days with less than 6 h of gap typical metric of surface area is leaf area index (LAI). LAI was observed
filled data (i.e. at least 18 h of observations) and monthly mean fluxes at most sites in mid-summer. However, some sites were never measured
were only reported if there were at least 22 days in the month of daily and due to logistical challenges, some years were missed at other sites. In
data. Site specific instrumentation is reported in the supplementary addition to missing data, LAI observations are at times limited to one
material (Table ST2) but most sites used a combination of LI-7500 IRGA observation a year and the available record contains some difference in
(LI-COR Biosciences) and CSAT-3 anemometer (Campbell Scientific the collection methods over the decades of observation (Straker et al.,
Inc.). 2019). As an additional metric of photosynthetic surface area, the
The flux of CO2, or net ecosystem exchange (NEE), was partitioned 16-day 250 m resolution quality controlled enhanced vegetation index
into the difference between gross ecosystem productivity (GEP) and (EVI) and normalized difference vegetation index (NDVI) from the
ecosystem respiration (ER). ER was directly observed at night (when MODIS satellite platform were downloaded from the USGS EARTHDATA
incoming shortwave radiation was <10 W m− 2 and friction velocities portal (https://lpdaacsvc.cr.usgs.gov/appeears/). EVI and NDVI indices
>0.3) when GEP was assumed to be zero. Occasionally young sites (11 provide both a continuous index of photosynthetic surface area for all
site-years of constructed and 8 site-years of harvested) did not have site years and the phenology of the growing season. For each site, the
sufficient data at night with friction velocities >0.3, for these either a 0.2 250 m pixel that contained the flux tower was downloaded for the entire
or 0.1 threshold was used depending on sample depth (a list of site-years age of the site. The exception is CONC-JP and CONC-SWSS, which were
and minimum threshold is provided in supplementary table ST3). An planted before either MODIS instruments were operational. For those
exponential relationship between night-time ER and soil temperature two sites, the MODIS coverage is limited to 2003–2020. Pixels flagged by
was used to interpolate ER for all periods and was adjusted for season­ the USGS as cloud covered or erroneous were removed from the record
ality over a moving window each year following the methods described and a 64-day moving widow (4 observations) was used to downscale the
in Humphreys et al. (2014). Daytime GEP was then calculated as the data by linear interpolation to a daily frequency. Then, as suggested by
difference of measured NEE and modelled ER. The rate of evaporation Pettorelli et al. (2005), each timeseries was smoothed with a 60-day
relative to productivity, the water use efficiency (WUE), was calculated running mean to limit the influence of artifacts in the imagery. In
based on the amount of GEP (in g of C–CO2) per kg of water evaporated addition to limiting the impact of artifacts, smoothing the satellite data
from the ecosystem. Surface conductance (Gs) and aerodynamic reduces the impact of spatial heterogeny (e.g. proximity to mining
conductance (Ga) were from inverting the Penmen-Monteith equation infrastructure) that could bias a single 250 m MODIS pixel covering a

Fig. 2. The nine sites used in this study. Left image for each site is the earliest growing season image available on Google Earth, right image is August 2020. The dates
the images were taken is provided in month/year format in the upper corner of each image. Yellow circles indicate a 250 m radius around the eddy covariance
instruments, the maximum extent that could be integrated into a single 250 m MODIUS pixel.

4
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

smaller site (Fig. 2). Monthly and seasonal mean values were then performed using a one-way ANOVA for mean ET, GEP, fluxes and WUE
calculated from the downscaled daily data. Although the smoothing of the mature sites and the disturbed sites in age classes of <5 years, 5–9
method reduces the max greenness observed in the growing season, it is years, and 10+ years since disturbance. Significant differences in means
consistently applied across the harvested, constructed and mature study were determined by a post-hoc Bonferroni test an alpha of 0.05.
sites and thus should not impact this analysis but should be noted for
inter-study comparability. 3. Results

2.3. Data analysis 3.1. Evapotranspiration

All regressions reported in this paper were either ordinary least The ET-age relationship explained a similar amount of variance be­
squares (OLS), hereafter referred to as ‘linear’, or a negative exponential tween the constructed conifer (RMSE = 32.3 mm) and constructed
non-linear least squares of the form β1(1-e(β2(x+ β3)), hereafter referred broadleaf (RMSE = 40.1 mm) forest groups (Fig. 3A). Generally,
to as simply ‘non-linear’. Only significant (α = 0.05) R2 are reported for broadleaf sites had greater rates of ET for the same age (Fig. 3A). Year-
linear models. However, all non-linear models are reported with the root to-year variance in ET was large compared to the age trend and any age
mean square error (RMSE) to indicate the average deviation from the trend appeared negligible after 5 years (Fig. 3A). Overall, there was a
curve. To simplify the visualization of plots with daily data (due to the minor positive age trend in drought sensitivity (Fig. 3B) and drought
high number of growing season days, n > 5000) the Y data was binned coupling (Fig. 3C) in the constructed sites (RMSE of the age sensitivity
and the mean value of each bin is plotted instead. Each plot had 18 bins, model is 0.23 and 0.20 and coupling model is 0.63 and 1.57 for
where data was grouped by 5 percentile increments of the X data in the broadleaf and conifer, respectively). The trend was more defined with a
plot from 5 to 95. Such that the first point is the mean of 5th < bin ≤ lower error in the harvested sites (RMSE 0.13 and 0.63). SB30 at age 4
10th percentile of the X data, the second point is the mean of the 10th < and 14 had the strongest drought coupling (0.77 and 0.78), but both of
bin ≤ 15th percentile, and so on. For example, when ET was plotted as a those years had typical sensitivity (1.0 and 1.1; Fig. 3B). The greatest
function of the vapor pressure deficit (VPD), only the mean ET corre­ sensitivity was in SWSS at age 10 (4.3) but it was not well coupled
sponding to each 5-percentile bin of VPD would be plotted. To assess the (0.28). While ET was closely tied to GPP (R2 = 0.37 and 0.48, p-values <
role of daily atmospheric conditions on drought, ET was modelled using 0.01, for broadleaf and conifer, respectively; Fig. 3D), the relationship
a multivariate linear regression of the predictors: incoming shortwave was much tighter for the mature sites (R2 = 0.84, p-value < 0.01).
radiation (Kdn), mean air temperature (Ta), VPD and Gs. To demon­ ET fluxes from the harvested forests were similar to the constructed
strate how observed fluxes can be scaled to the whole hydrological unit, forests. The variance in the age-trend was lower than the constructed
the relationship between ET, GEP and WUE were modelled with OLS to forests (RMSE = 17.9). The age trend was also similar, but with a steeper
the MODIS derived greenness indexes. slope (i.e. faster recovery) when compared to constructed broadleaf
The resilience of the forest to climate stress was tested in two ways. forests (Fig. 3A). It is likely that the steepness of the slope, and thus the
First, a drought coupling and drought sensitivity index were calculated rate of recovery, was driven by the oldest two site years with much
at the daily scale via the methods of Anderegg et al. (2018), which were higher ET fluxes (Fig. 3A). The drought sensitivity of the harvested sites
modified from Sulman et al. (2016) and Novick et al. (2016). Anderegg had a more rapid age trend than the constructed forests and the drought
et al. (2018) demonstrated how this approach can be utilised to compare coupling showed a similar lack of age trend (Fig. 3B,C). However, the ET
drought resilience across the wide diversity of forests around the world. fluxes from the harvested sites were more closely coupled to produc­
Briefly, the drought coupling is the coefficient of determination from a tivity than the constructed sites (R2 of 0.77) but not as tight as the
standardized linear model of ET as a function of VPD, soil moisture, and mature sites. The ET flux in both the mature and harvested sites had a
their interaction term for June, July, and August. Likewise, the drought greater response to productivity (greater slope) and lower base ET
sensitivity was the sum of the absolute coefficients, including the (smaller intercept) when compared to the constructed forests (Fig. 3D).
interaction term of soil moisture and VPD. The second method assessed Atmospheric conditions had a greater impact on ET than stand
resilience with a climate sensitivity analysis using the impact of hot, development. Specifically, Kdn, Ta, VPD and Gs were the dominant
cold, wet, and dry years on the relationship between greenness and ET, drivers of ET at the half hourly scale for all ages (Fig. 4). Ordinary least
GEP fluxes and WUE. Hot (cold) years were all growing seasons with squares regression of ET vs these variables resulted in R2 for 0.63, 0.25,
temperatures greater than the 75th percentile (lower than the 25th 0.25, and <0.01 for Kdn, Ta, VPD, and Gs, respectively (all p-value <
percentile) and wet (dry) growing seasons with total precipitation 0.0001 except Gs which was insignificant with a p-value of 0.36).
greater than the 75th percentile (lower than the 25th percentile). Since Combining the four variables (Kdn, Ta, VPD, and Gs) in a single multi­
there is an underlying age trend to the data, absolute fluxes are difficult variate linear regression accounted for 66% of the variance in the daily
to compare in a year-to-year evaluation of the impact of climate. Instead, ET observations across all constructed and harvested sites in this study
given a quantity of satellite observed photosynthetic surface area, stress (p-value < 0.0001). There was a clear increase in ET for any given
should manifest itself as lower productivity or greater water loss per unit variable with age (Fig. 4), but the ET response to Ta and VPD in the older
of productivity. This approach can be justified because there were no disturbed broadleaf sites was less than the equivalent in the mature sites.
significant differences in EVI between climate years binned by normal, Harvested sites had the greatest ET fluxes in the 5+ age group for similar
wet, dry, hot, and cold (One-Way ANOVA; p-value 0.58 and 0.31 for rates of Gs (Fig. 4). The ET response to all four variables was greater in
normal/wet/dry and normal/hot/cold years, respectively). Since the the constructed conifer sites compared to the mature sites.
product of EVI and Ta, the product of EVI and Kdn, and NDVI was found
to explain the most variance in ET, GEP, and WUE, respectively (see 3.2. Productivity
analysis below), these relationships between fluxes and greenness were
used as the basis to evaluate the impacts of climate variability by The GEP at both conifer and broadleaf sites had a strong relationship
analyzing the regression coefficients of a multilinear model using hot, with age (RMSE = 75 and 75 g C m− 2, respectively; Fig. 5A). Although
cold, wet, or dry years as binary variables in addition to the linear the rate of change per year was slowing, it remains unclear at what point
relationship of greenness on ET, GEP and WUE, respectively. To mitigate it will reach an asymptote. Unlike the mature and harvested sites,
the need for a large model of interaction terms that would over-fit the respiration and productivity were not closely associated in the con­
dataset, wet/dry were analyzed separately from hot/cold since those structed forests (R2 of 0.05 and 0.16. p-values >0.05 for broadleaf and
years are mutually exclusive (i.e. hot and cold years are not possible, confier, respectively) (Fig. 5B). This is likely due to the use of stockpiled
unlike hot wet years). Finally, direct comparison of fluxes was soils with high organic content and the increased CO2 associated with

5
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

Fig. 3. (A) Growing season total ET, (B) drought coupling, and (C) drought sensitivity as a function of site age. D) Seasonal total evapotranspiration as a function of
gross ecosystem productivity. The curves for the non-linear (A,B, and C) and linear (D) models are shown for the constructed conifer sites (dark green), broadleaf
(light green), harvested (yellow) and mature (red) sites. Mature sites are not included in the age analyses.

Fig. 4. Mean evaporation from each 5-percentile bin of Kdn, Ta, VPD and Gs for mature, harvested and constructed forests with the harvested and constructed forests
sorted into age classes of <5, 5–9, and 10 or more years since disturbance. Only half hour observations between 10:00 h and 16:00 h local time are shown.

decomposition of the organics. 3.3. Water use


Compared to the constructed forests, the harvested sites had a
greater GEP at every age but with a greater variance (RMSE = 115 g C There was a rapid increase in WUE followed by a relatively stable
m− 2). Both the harvested sites and the mature sites had a close rela­ asymptote for the constructed sites beginning between 5 and 10 years
tionship between GEP and ER (R2 of 0.89, 0.95 and 0.65, p-values all after planting. WUE in both constructed conifer and broadleaf sites had
<0.001 for the harvested, mature conifer and mature broadleaf, similar variance around the non-linear trend (RMSE = 0.31 and 0.27 g
respectively) (Fig. 5B). The mature broadleaf forests and older harvested kg− 1, respectively) (Fig. 6). The harvested sites had a similar pattern but
sites (age 5+) had greater productivity than every grouping of con­ reached an asymptote slightly earlier (approximately after 3 years) with
structed forest for the same level of incoming shortwave radiation. slightly greater variance (RMSE = 0.38 g kg− 1) and a much higher WUE
Unlike the broadleaf, the mature conifer and 10+ aged constructed than the constructed forests at every age.
confier sites had similar productivity responses to incoming shortwave
radiation (Fig. 5C).

6
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

Fig. 5. (A) Total growing season gross ecosystem productivity as a function of age. (B) Growing season total productivity as a function of seasonal total ecosystem
respiration. Regression lines are only shown for significant relationships (p-value < 0.05). (C) Mean binned daily gross ecosystem productivity as a function of age,
binned in 5 percentile increments of Kdn. Dashed lines are conifer sites and solid lines are broadleaf.

values < 0.0001) (Non-linear curves shown in Fig. 7A). EVI had a
slightly weaker, but still significant, relationship to age (r = 0.83 and
0.64 for broadleaf and conifer, respectively, p-values < 0.0001) (Non-
linear curves shown in Fig. 7B). EVI at the sites with the dense Jack Pine
planting (CONC-JP and CONC–C11A) was very stable with age. The
harvested sites had a small age trend in NDVI but none in EVI (Fig. 7A,
B). This is likely because tree removal at HAR-P40S and HAR-P40N took
place over an area of <25 000 m2, or <37% of a 250 m × 250 m MODIS
pixel even if it captured the entire logged region. At the constructed
forests sites LAI was positively correlated with both EVI and NDVI (r =
0.72 and 0.60, respectively, both p-value < 0.0001) (Fig. 7C for EVI,
NDVI not shown). The harvested sites had a slightly weaker relationship
between NDVI and LAI (r = 0.57, p-value = 0.017, not shown), but an
insignificant relationship to EVI (Fig. 7C). Both are likely due to the scale
of the MODIS pixel discussed above. The variance in the monthly EVI
and NDVI values were very similar for the study sites (r = 0.84, p-value
< 0.0001) (Fig. 7D).
Correlations between EVI and NDVI and monthly mean ET, GPP, and
WUE indicate that satellite derived greenness provide a robust estimate
of the carbon and water fluxes. All the following OLS regression models
are significant at p-value < 0.0001. Between the different greenness
products analyzed here, EVI x Ta, provided a slightly stronger relation­
ship to ET than NDVI x Ta, (all groups pooled, R2 = 0.64 and 0.47,
respectively) (Fig. 8A,B). Likewise, EVI x Kdn was a similar fit to GEP as
NDVI x Kdn (pooled, R2 = 0.75 and 0.80, respectively) (Fig. 8C,D).
However, the relationship between EVI and WUE (pooled, R2 = 0.19) is
weaker than the relationship between NDVI and WUE (pooled, R2 =
0.57) (Fig. 8E,F). There was negligible difference between the harvested
and mature sties in greenness and ET, GEP and WUE relationships,
Fig. 6. Growing season water use efficiency as a function of age. The colours
represent constructed conifer sites (dark green), broadleaf (light green), and except that the harvested and mature sites had greater intercepts in GEP
harvested sites (yellow). and WUE due to higher productivity noted above. Because NDVI cap­
tures more of the variance in WUE and the inconsistent relationship of
EVI to WUE, NDVI will be used as the greenness index for the climate
3.4. Greenness
stress analysis on WUE below.
NDVI had a strong relationship to age (Pearson correlation coeffi­
cient (r) of 0.89 and 0.75 for broadleaf and conifer, respectively, p-

7
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

Fig. 7. The MODIS derived NDVI (A) and EVI (B) as a function of age. EVI as a function of site MATC–Observed LAI values (C) and relationship between NDVI and
EVI (D). Regression lines shown for conifers (dark green), broadleaf (light green), harvested (yellow) or all the data (black).

Fig. 8. The relationship between satellite data (EVI on top and NDVI on bottom) and ET with air temperature (Ta, A and B), GPP with incoming shortwave radiation
(Kdn, C and D), and WUE (E and F). OLS regression line shown in black, significant at p-value < 0.0001.

3.5. Climate stress (Table 3 and Fig. 9). ET was slightly higher in dry years, particularly in
low EVI or cool sites (Table 3, Fig. 9A,B). Wet years had little overall
For clarity of interpretation, we only focused on the constructed impact on ET, but there was a small decoupling of the relationship
forests and sub-set the data into two analyses. Although the interaction (shallower slope) between ET and greenness in both wet and dry years
between the subsets may be of value, there were too few years that relative to the ‘normal’ (Table 3 and Fig. 9A,B). Hot years had a positive
represented all nine possible interactions of normal, hot, cold, wet and impact on the intercept of the ET model where cool years had a negative
dry years running that would likely overfit the model. Therefore, two impact on the slope. Productivity as a function of greenness had a slight
separate OLS models were constructed, one of normal, hot and cold negative impact from dry or wet conditions. The slope of the relation­
years (mutually exclusive conditions) and one of normal, dry, and wet ship between GEP and EVIxKdn was reduced in cold years (Table 3 and
years (also mutually exclusive conditions). These basic linear models Fig. 9C,D). WUE was overall greater in most cold years than hot or dry
indicate how the above relationships between ET, GEP and WUE to years, as indicated by the intercept, but the relationship between
greenness only had slight changes during periods of climate variability greenness and WUE broke down in the cold years as the slope

8
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

Table 3
Standardized regression coefficients for models of climate stress on ecosystem fluxes (equivalent of the non-standardized models shown in Fig. 9). Regressions are
performed on the standardized data (z-score) for comparability, resulting in coefficients that are equivalent to the% variance explained by each variable and intercepts
in units of 1 standard deviation. X is the MODIS derived greenness indicator described in text (when appropriate it is multiplied by the mean Ta or Kdn for that month).
Wet, Dry, Cold, Hot are boolean variables in the model indicating if the observation occurred during the respective climate period and Conif indicates a Boolean if the
observation was from a conifer site. All models have p-values < 0.0001.
Int. X Wet Dry Conif Wet x X iDry x X Conif x X R2

ET − 0.08 0.78 − 0.01 0.11 0.11 − 0.07 − 0.06 − 0.05 0.58


GEP 0.16 0.77 − 0.05 − 0.09 − 0.52 − 0.07 − 0.06 0.05 0.76
WUE 0.17 0.59 − 0.06 − 0.11 − 0.63 − 0.03 0.07 0.16 0.58
Int. X Cold Hot Conif Cold x X Hot x X Conif x X R2
ET − 0.07 0.77 0.05 0.14 0.06 − 0.03 0.03 0.03 0.58
GEP 0.18 0.78 − 0.03 − 0.03 − 0.57 − 0.07 0.08 − 0.10 0.76
WUE 0.21 0.57 − 0.03 − 0.07 − 0.70 − 0.12 0.09 − 0.14 0.59

Fig. 9. Plots of multilinear regression models of key ecosystem response to MODIS derived greenness products during hot, cold, wet or dry years. All OLS models are
significant (p-values < 0.0001). Interaction terms make some lines appear non-linear. The standardized coefficients of each linear model are provided in Table 3.

approached 0 (Table 3, Fig. 9E,F). any age (2.6 ± 0.3 vs 3.5 ± 0.5 and 3.7 ± 0.3 g kg− 1, respectively)
(Fig. 10). Likewise, the WUE at the constructed conifer sites (10+ aged)
were significantly lower than the mature conifer sites (2.1 ± 0.2 and 3.1
3.6. Comparing the constructed, harvested and mature sites
± 0.2 g kg− 1, respectively).
Mean growing season ET in the 10+ aged AOSR reclaimed broadleaf
4. Discussion
forests was not significantly different from the BERMS mature MATB-
Oas site or the harvested sites older than 5 years (264 ± 30, 264 ± 43,
The widespread use of EC allows for a practical assessment of pro­
and 263 ± 34 mm, respectively). However, ET from the constructed
ductivity, evaporation and water use as integrative measures of ‘equiv­
conifer sites greater than 10+ years (211 ± 38) was significantly greater
alence’ to other Boreal Plains ecosystems. This approach is useful for
than the BERMS mature MATC–Obs and MATC–Ojp conifer sites
developing criteria and indicators for post-mining forest recovery and
(combined mean of 165 ± 21 mm) (Fig. 10). The mean growing season
the high correlation to satellite products further expands the ability to
GEP flux in the 10+ aged constructed broadleaf forests was significantly
monitor forest resilience during periods of climate extremes in the
less than both the BERMS mature MATB-Oas site and the harvested sites
AOSR. Furthermore, it is likely this type of analysis would work with
older than 5 years (679 ± 101 vs 920 ± 92 and 1001 ± 109 g C m− 2,
other stressors, such as disease and insect damage, but there were no
respectively) (Fig. 10). The seasonal GEP flux from the constructed
outbreaks during this study period upon which to test this theory. The
conifer sites aged 10+ years were not significantly different from the
following section further discusses: 1) how the constructed AOSR forests
BERMS mature conifer sites (478 ± 65 and 502 ± 68 g C m− 2, respec­
are recovering with age relative to other published Canadian forests; and
tively) (Fig. 10). The WUE in the constructed broadleaf forests (10+
2) how effective this assessment is for monitoring the resilience of the
aged) were also significantly less than the mature or harvested sites of

9
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

Fig. 10. A comparison of the mature (Mat) Boreal Ecosystem Research and Monitoring broadleaf (MATB-Oas) and conifer (MATC–Obs, MATC–Ojp) sites and the
reclaimed AOSR sites of ages <5, 5–9, and 10+ years since planting. Individual site years are plotted with the site symbol. Red line is the median, black line is the
mean, with boxes indicating the interquartile range of all sites in that group. Boxes are coloured according to grouping; constructed conifer (CONC), constructed
broadleaf (CONB), harvested (HAR) and mature conifer (MATC) and broadleaf (MATB). Significantly different (α = 0.05) means were determined with a Bonferroni
test post hoc test, presented in supplementary Table ST4.

constructed forests. required to maintain soil water availability regardless of vegetation


development during extreme climate cycles. The modeling in Huang
et al. (2015) determined that ET increased with soil depth as a
4.1. Post-construction age response non-linear function of depth with the rate decreasing rapidly around
100 cm. Here the total seasonal ET does increase with soil depth be­
Our results suggest that after 5 years the reclaimed broadleaf forests tween sites (Table 1, Fig. 3), however this relationship is confounded by
have recovered their ET functions on the landscape as shown by the the age trend as well as the species present and should not be assumed to
similar rates of ET as the mature sites (Fig. 10), however ET remains be robust to statistical scrutiny. ET as a function of soil depth is only
higher per unit of productivity for all constructed forests (Fig. 3D). The mentioned here to highlight that both models and eddy covariance ob­
broad goal of the AOSR reclamation projects is to create integrated servations indicate soil depth and composition impacts upland ET rates.
landscapes and therefore the similar hydraulic function of constructed This could have big implications on constructing hydrological response
upland forests to the natural analogues is paramount for supplying areas (i.e. watershed scale reclamation) when scaling guidelines devel­
appropriate moisture to downgradient ecosystems (Devito et al., 2012; oped for specific ecosystems.
Ketcheson et al., 2016). However, evaporation rates from constructed Productivity is closely related to stand age (Fig. 5). Although, the
conifer sites are higher than the natural analogues (Fig. 10) and the high 10+ year old constructed broadleaf forests appear to be reaching an
year-to-year variability makes it difficult to determine if rates are still asymptote well below their mature counterparts, the conifers appear to
increasing with age after 17 years (Fig. 3). High rates of ET relative to have reached their expected levels of productivity (Fig. 10). Unlike the
natural conifer stands is likely due to the moisture retaining properties constructed site, the harvested sites appear to have recovered to the
of the soils used in reclamation. Elsewhere in the Boreal Plains, the mature productivity levels by their 5th year of growth. All sites are ex­
dominant control in both above and below canopy ET rates was soil pected to continue increasing based on the non-linear trends presented
moisture (Brown et al., 2014). In this study, CONC–C11A, CONC-SWSS, here (Fig. 5). These results align with the findings of Coursolle et al.
CONC–CKB and CONC-JP are all mesic soils, CONC–NIU is classified (2012) who found a chronosequence of Canadian boreal forest EC sys­
as sub-mesic, and all contain more moisture than the xeric soils found tems showed a peak productivity around ~20 years, with older (~80+
nearby in an undisturbed jack pine site (Straker et al., 2019). Further­ years) forests showing slightly lower fluxes. However, the close rela­
more, although the underlying tailings are primarily sand, the soil cover tionship between ET and GEP (Fig. 3) implies that this could lead to
contains peat additions and high clay sized fraction content. The per­ increased ET fluxes without substantial increases in WUE (which appear
centage of sand in the sub-2 mm fraction of the soil cover used in the to have plateaued much lower than the harvested or mature sites, Figs. 6
constructed of the conifer forests was 60.8% (CONC–C11A), 31.7% and 10). While it is challenging to untangle ET, GEP, fluxes and WUE,
(CONC-SWSS), 32.4% (CONC-JP), and 49.7% (CONC–NIU) compared the generally greater WUE in harvested and matures sites combined with
to 97.4% in that same nearby undisturbed jack pine site (Straker et al., the age trajectories of both GEP and ET suggest that water losses from
2019). The two mature BERMS conifer sites had a mean total sand the constructed forests may increase in the future and impact down­
content of 83.3 ± 14.0% in the mineral soils at MATC–Obs (i.e. gradient ecosystems. This may already be occurring at the constructed
non-peat) and 92.2 ± 12.4% at MATC–Ojp (Anderson, 2000). The low conifer sites, which show greater rates in the half-hour ET fluxes relative
sand content and enhanced storage capacity near the surface of the to the mature sites for any level of Kdn, Tair, VPD, or Gs (Fig. 4).
constructed conifer forests likely limits drainage, increases soil mois­
ture, and results in the increased evaporation reported here. Another
possible cause of the large ET fluxes from the constructed conifer sites 4.2. EC derived fluxes as a resilience metric for reclamation
could be the soil depth used to cover the tailings/overburden during
construction. Huang et al. (2015) determined that 100 cm of soil was The purpose of analyzing the impact of climate variability on

10
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

constructed forests in the AOSR is twofold. First, to quantify if the fluxes. This is consistent with the greenness reported in this study
constructed forests have similar resilience to disturbance as natural (Fig. 7), however after 10 years the rate of year-to-year change in NDVI
boreal stands for criteria under the Indicators Framework for Oil Sands or EVI is still increasing in the constructed conifer forests. There are
Mine Reclamation Certification (Alberta Environment and Sustainable fewer studies of other forms of disturbance on boreal forests. In one,
Resource Development, 2013). Second, to better understand how Stephens et al. (2018) found that even during extreme defoliation by
climate anomalies may compound impacts to downgradient ecosystems tent caterpillars it took only months for the BERMS MATB-Oas site to
in the integrated landscapes of these large-scale reclamation projects. return to expected productivity levels. Aside from the year of caterpillar
Kljun et al. (2006) showed that multiple sequential drought years are outbreak, and the three-year drought (analyzed and reported by Kljun
required to negatively impact boreal forest ecosystems, but such et al., 2006), the biggest impact to annual productivity rates was the
multi-year dry conditions did not occur over this study period. Assess­ timing of spring thaw. In this study however, mean May temperature
ment of climate variability on within-year fluxes demonstrated small was not significantly correlated to seasonal GEP (r = 0.20, p-value =
impacts to ET, GEP, and WUE (Fig. 9). This finding is similar to the lack 0.185, analysis not shown) and the constructed AOSR forests had a
of wet/cold years impact on AOSR fen ecosystems found by Volik et al. negligible productivity response to overall hotter growing seasons
(2021a). Although the drought coupling increased with age (i.e. the (Fig. 9; Table 3).
relationship between soil moisture, VPD and ET was tighter) it was no Overall, the constructed AOSR forests appear to be responding in a
more sensitive with age or climate anomalies (i.e. the slopes did not manner similar to other disturbed sites but may have a longer trajectory
increase). The lack of persistent impacts following a dry, hot, wet, or back to peak productivity than the harvested sites compared here
cold year suggests that these constructed forests maintain a degree of (Fig. 5). The minimal response to dry or hot conditions also suggests a
resilience to climate variability. Mature Boreal Plains forests increase resilience to climate variability. It is unlikely that this resilience mech­
WUE during drought periods (Kljun et al., 2006) but unexpectedly the anism is a function of the diversity found in undisturbed forests (i.e.
constructed AOSR forests slightly decreased WUE during dry and hot Anderegg et al., 2018), but rather by soil attributes designed to promote
years (Fig. 9E,F). This inverse response to dry conditions could again be high moisture retention during climate cycles (Huang et al., 2015).
related to the soil properties of the constructed AOSR forests. In Boreal Although it is favorable for the forests to be resilient to drought,
Plains wetlands, deep, high organic content, and low porosity soils may excessive retention and evaporation can deprive downgradient ecosys­
have been shown to be part of the mechanism for how WUE increases tems of water. Furthermore, this study joins other work (e.g. Chasmer
under wetter conditions (Volik et al., 2021b). In the meantime, et al., 2018) to highlight how eddy covariance systems can be coupled
short-term dry periods are unlikely to impact boreal forest productivity with satellite derived greenness indexes to monitor entire constructed
(Kljun et al., 2006), especially if moisture retention in the soils is high forests giving land managers an effective and easily obtainable tool to
particularly since we defined dry years as a 1 in 4-year event (75th monitor forest development. Although MODIS is used in this study
percentile). These short-term dry periods likely increase overall evapo­ because of its large temporal coverage, higher resolution satellite
ration from moist soils which in turn leads to a reduced ecosystem level derived greenness indexes are available for more contemporary periods.
WUE. The similarity between NDVI and EVI suggests that these relationships
The constructed AOSR forests analyzed here have a similar response are likely robust to slight differences in wavelengths between various
to other studies of boreal forest disturbance. Mkhabela et al. (2009) satellite-based platforms. It is expected that a wider range of spectral
found that three post-harvesting jack pine sites (355 ± 137 g m− 2) had indexes can be used to extrapolate flux data to the reclamation areas
slightly lower productivity than three post-fire jack pine sites (589 ± 67 increasing what tools are available to land managers to isolate and
g m− 2) due to a greater diversity of species in the post-fire sites relative intervene in regions which may not be performing as expected.
to the post-harvest sites. The 10+ year GEP fluxes from the constructed
conifer sites (478 ± 65 g m− 2) is very similar to the mean GEP flux 10+ 5. Conclusion
year post-logging recovery presented in Mkhabela et al. (2009).
Although unlike what was presented in Mkhabela et al. (2009), diversity Seasonal EC fluxes provide a practical assessment tool for upland
may not be the key indicator for productivity in the constructed forests. forest systems in the AOSR, particularly as part of the indicators of
The greatest productivity in the constructed forest was a site with resilience to climatic variability and criteria of equivalence. It is
dominant stands of both Aspen and White Spruce (CONB-SB30), but the particularly important to align hydrological fluxes with expectations as
greatest diversity in tree species was found in CONC-SWSS (Straker the scale of reconstruction in the AOSR requires constructed watersheds,
et al., 2019), which only had average (in this study) seasonal GEP fluxes. or ‘hydrological response areas’ to be fully integrated ecosystems, or
However, it may explain the high productivity in the naturally regen­ ‘hydrological units.’ Here we have shown that broadleaf forests return to
erated harvested sites (1001 ± 109 g m− 2). In a chronosequence study of natural levels of ET within 5–10 years, but it is uncertain if GEP will ever
Canadian forest recovery, Coursolle et al. (2012) determine that GEP at reach the same range, or if future increases in GEP will drive higher ET
black spruce, jack pine, aspen and mixed wood forests return to natural rates. The conifer sites returned to natural variability of GEP within 10
variance after ~20 years following disturbance. In the Coursolle et al. to 20 years, but ET remains high and is possibly continuing to increase
(2012) study, jack pine and black spruce fluxes stabilized between 500 - with age. Due to these high fluxes, both constructed conifer sties
800 g C m− 2 y− 1, while the aspen and mixed wood forests GEP was broadleaf sites had low WUE compared to mature and post-harvested
between 800 and 1200 g C m− 2 y− 1 (both ranges interpreted from sites. We suggest here that the high rates of ET relative to GEP is
publication figure). These ranges are greater than the 10+ aged GEP likely due to the soil properties used in construction and will likely
from the constructed conifer (478 ± 65 g C m− 2 y− 1) and broadleaf (679 impact downgradient ecosystems. However, the high-water retention in
± 101 g C m− 2 y− 1) forests reported here. It could be that these numbers the soils also likely produced a resilience to dry and/or hot years,
are a low estimate since the reported growing season in this study is only revealing a trade-off when scaling reclamation objectives from
June, July, and August vs the full annual flux reported in Coursolle et al. ecosystem to watershed scales. Finally, the scalability to satellite based
(2012). Indeed, Coursolle state the MATB-Oas growing season runs from observational platforms provides a useful tool to land managers in
17 May to 25 September (131 days) and MATC–Ojp runs from 16 April determining the water lost to the atmosphere across the entire hydro­
to 13 October (180 days), substantially longer than the 91 days included logical unit.
in the growing season reported here. In another national study, Hicke
et al. (2003) utilized satellite data to determine it took on average 9 Supplementary
years for productivity to return to Canadian boreal forests after fire, but
this is based on satellite observed ‘greenness’, not carbon or water Table ST1: The amount of observations that passed QA/QC in the

11
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

June, July and August period (i.e. the non-gap filled proportion). Clark, M.G., Drewitt, G.B., Carey, S.K., 2021. Energy and carbon fluxes from an oil sands
pit lake. Sci. Total Environ. 752, 141966 https://doi.org/10.1016/j.
Table ST2: Sonic anemometer and infrared gas analyzer used at each
scitotenv.2020.141966.
site. Coursolle, C., Margolis, H.A., Giasson, M.A., Bernier, P.Y., Amiro, B.D., Arain, M.A.,
Table ST3: Site-years which had insufficient data to use a 0.3 u* Barr, A.G., Black, T.A., Goulden, M.L., McCaughey, J.H., Chen, J.M., 2012. Influence
threshold for nighttime respiration modeling. of stand age on the magnitude and seasonality of carbon fluxes in Canadian forests.
Agric. For. Meteorol. 165, 136–148. https://doi.org/10.1016/j.
Table ST4: Significantly different means as determined by a Bon­ agrformet.2012.06.011.
ferroni post-hoc test. X indicates means are significantly different. Data Devito, K., Mendoza, C., Qualizza, C., 2012. Conceptualizing water movement in the
is plotted in Figure 10. Boreal Plains. Implications for watershed reconstruction. Synthesis report prepared
for the Canadian Oil Sands Network for Research and Development, Environmental
and Reclamation Research Group. 164 pp. 10.7939/R32J4H.
Gosselin, P., Hrudey, S.E., Naeth, M.A., Plourde, A., Therrien, R., Van Der Kraak, G.,
Declaration of Competing Interest Xu, Z., 2010. Environmental and Health Impacts of Canada’s oil Sands Industry.
Royal Society of Canada, Ottawa, ON, p. 10.
Government of Alberta, 2020. Environmental Protection and Enhancement Act. Revised
The authors declare that they have no known competing financial statues of Alberta 2000, Chapter E-12. Government of Alberta. Available at. htt
interests or personal relationships that could have appeared to influence ps://www.qp.alberta.ca/documents/acts/e12.pdf [Accessed 27 January 2021].
Helbig, M., Waddington, J.M., Alekseychik, P., Amiro, B.D., Aurela, M., Barr, A.G.,
the work reported in this paper. Black, T.A., Blanken, P.D., Carey, S.K., Chen, J., Chi, J., 2020. Increasing
contribution of peatlands to boreal evapotranspiration in a warming climate. Nat.
Clim. Chang. 10 (6), 555–560. https://doi.org/10.1038/s41558-020-0763-7.
Acknowledgements Hicke, J.A., Asner, G.P., Kasischke, E.S., French, N.H., Randerson, J.T., James Collatz, G.,
Stocks, B.J., Tucker, C.J., Los, S.O., Field, C.B., 2003. Postfire response of North
This work was funded by Syncrude Canada Ltd. research grant American boreal forest net primary productivity analyzed with satellite MATC-
Observations. Glob. Chang. Biol. 9 (8), 1145–1157. https://doi.org/10.1046/j.1365-
(460010687); Canada First Research Excellence Fund, Global Waters 2486.2003.00658.x.
Futures program; Collaborative Research and Development grant Huang, M., Barbour, S.L., Carey, S.K., 2015. The impact of reclamation cover depth on
(CRDPJ477235-14) and industry partners Syncrude Canada Ltd and the performance of reclaimed shale overburden at an oil sands mine in Northern
Alberta, Canada. Hydrol. Process. 29 (12), 2840–2854. https://doi.org/10.1002/
Canadian Natural Resources Limited; as well as Collaborative Research
hyp.10229.
and Development grant (418557-2011) and industry partners Suncor Humphreys, E.R., Andrew Black, T., Morgenstern, K.A.I., Li, Z., Nesic, Z., 2005. Net
Energy Inc., Esso Imperial Oil Ltd., and Shell Canada Ltd.), Northern ecosystem production of a Douglas-fir stand for 3 years following clearcut
harvesting. Glob. Chang. Biol. 11 (3), 450–464. https://doi.org/10.1111/j.1365-
Scientific Training Program and Ducks Unlimited. We would also like to
2486.2005.00914.x.
thank Mike Treberg, Gord Drewit, Adam Green, Ronald Peter Van Humphreys, E.R., Charron, C., Brown, M., Jones, R., 2014. Two bogs in the Canadian
Haarlem, Scott Brown, and George Sutherland for all their hard work in Hudson Bay Lowlands and a temperate bog reveal similar annual net ecosystem
the field and in front of a computer collecting and organizing data. exchange of CO2. Arct. Antarct. Alp. Res. 46 (1), 103–113. https://doi.org/10.1657/
1938-4246.46.1.103.
Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M.,
Supplementary materials Saha, S., White, G., Woollen, J., Zhu, Y., 1996. The NCEP/NCAR 40-year reanalysis
project. Bull. Am. Meteorol. Soc. 77 (3), 437–472. https://doi.org/10.1175/1520-
0477(1996)077<0437:TNYRP>2.0.CO;2.
Supplementary material associated with this article can be found, in Ketcheson, S.J., Price, J.S., Carey, S.K., Petrone, R.M., Mendoza, C.A., Devito, K.J., 2016.
the online version, at doi:10.1016/j.agrformet.2022.109127. Constructing fen peatlands in post-mining oil sands landscapes: challenges and
opportunities from a hydrological perspective. Earth Sci. Rev. 161, 130–139.
https://doi.org/10.1016/j.earscirev.2016.08.007.
References Ketcheson, S.J., Price, J.S., Sutton, O., Sutherland, G., Kessel, E., Petrone, R.M., 2017.
The hydrological functioning of a constructed fen wetland watershed. Sci. Total
Environ. 603, 593–605. https://doi.org/10.1016/j.scitotenv.2017.06.101.
Anderegg, W.R., Konings, A.G., Trugman, A.T., Yu, K., Bowling, D.R., Gabbitas, R.,
Kljun, N., Black, T.A., Griffis, T.J., Barr, A.G., Gaumont-Guay, D., Morgenstern, K.,
Karp, D.S., Pacala, S., Sperry, J.S., Sulman, B.N., Zenes, N., 2018. Hydraulic diversity
McCaughey, J.H., Nesic, Z., 2006. Response of net ecosystem productivity of three
of forests regulates ecosystem resilience during drought. Nature 561 (7724),
boreal forest stands to drought. Ecosystems 9 (7), 1128–1144. https://doi.org/
538–541. https://doi.org/10.1038/s41586-018-0539-7.
10.1007/s10021-005-0082-x.
Anderson, D., 2000. BOREAS TE-01 SSA Soil Lab Data. ORNL DAAC, Oak Ridge,
Lukenbach, M.C., Spencer, C.J., Mendoza, C.A., Devito, K.J., Landhäusser, S.M., Carey, S.
Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/530.
K., 2019. Evaluating how landform design and soil covers influence groundwater
Amiro, B.D., MacPherson, J.I., Desjardins, R.L., Chen, J.M., Liu, J., 2003. Post-fire carbon
recharge in a reclaimed watershed. Water Resour. Res. 55 (8), 6464–6481. https://
dioxide fluxes in the western Canadian boreal forest: evidence from towers, aircraft
doi.org/10.1029/2018WR024298.
and remote sensing. Agric. For. Meteorol. 115 (1–2), 91–107. https://doi.org/
Mkhabela, M.S., Amiro, B.D., Barr, A.G., Black, T.A., Hawthorne, I., Kidston, J.,
10.1016/S0168-1923(02)00170-3.
McCaughey, J.H., Orchansky, A.L., Nesic, Z., Sass, A., Shashkov, A., 2009.
Alberta Environment and Sustainable Resource Development, 2013. Criteria and
Comparison of carbon dynamics and water use efficiency following fire and
Indicators Framework For Oil Sands Mine Reclamation Certification. Prepared by
harvesting in Canadian boreal forests. Agric. For. Meteorol. 149 (5), 783–794.
Mike Poscente and Theo Charette of Charette Pell Poscente Environmental
https://doi.org/10.1016/j.agrformet.2008.10.025.
Corporation for the Reclamation Working Group of the Cumulative Environmental
Nicholls, E.M., Carey, S.K., Humphreys, E.R., Clark, M.G., Drewitt, G.B., 2016. Multi-year
Management Association. Alberta Environment and Sustainable Resource
water balance assessment of a newly constructed wetland, Fort McMurray, Alberta.
Development, Fort McMurray, AB. September 18, 2012.
Hydrol. Process. 30 (16), 2739–2753. https://doi.org/10.1002/hyp.10881.
Baldocchi, D.D., Hincks, B.B., Meyers, T.P., 1988. Measuring biosphere-atmosphere
Novick, K.A., Ficklin, D.L., Stoy, P.C., Williams, C.A., Bohrer, G., Oishi, A.C., Papuga, S.
exchanges of biologically related gases with micrometeorological methods. Ecology
A., Blanken, P.D., Noormets, A., Sulman, B.N., Scott, R.L., 2016. The increasing
69 (5), 1331–1340. https://doi.org/10.2307/1941631.
importance of atmospheric demand for ecosystem water and carbon fluxes. Nat.
Biagi, K.M., Clark, M.G., Carey, S.K., 2021. Hydrological functioning of a constructed
Clim. Chang. 6 (11), 1023–1027. https://doi.org/10.1038/nclimate3114.
peatland watershed in the Athabasca oil sands region: potential trajectories and
Petrone, R.M., Chasmer, L., Hopkinson, C., Silins, U., Landhäusser, S.M., Kljun, N.,
lessons learned. Ecol. Eng. 166, 106236 https://doi.org/10.1016/j.
Devito, K.J., 2015. Effects of harvesting and drought on CO2 and H2O fluxes in an
ecoleng.2021.106236.
aspen-dominated western boreal plain forest: early chronosequence recovery. Can. J.
Brown, S.M., Petrone, R.M., Chasmer, L., Mendoza, C., Lazerjan, M.S., Landhäusser, S.M.,
For. Res. 45 (1), 87–100. https://doi.org/10.1139/cjfr-2014-0253.
Silins, U., Leach, J., Devito, K.J., 2014. Atmospheric and soil moisture controls on
Pettorelli, N., Vik, J.O., Mysterud, A., Gaillard, J.M., Tucker, C.J., Stenseth, N.C., 2005.
evapotranspiration from above and within a Western Boreal Plain aspen forest.
Using the satellite-derived NDVI to assess ecological responses to environmental
Hydrol. Process. 28 (15), 4449–4462. https://doi.org/10.1002/hyp.9879.
change. Trends Ecol. Evol. 20 (9), 503–510. https://doi.org/10.1016/j.
Carey, S.K., 2008. Growing season energy and water exchange from an oil sands
tree.2005.05.011 (Amst.).
overburden reclamation soil cover, Fort McMurray, Alberta, Canada. Hydrol.
Reed, B.C., Brown, J.F., VanderZee, D., Loveland, T.R., Merchant, J.W., Ohlen, D.O.,
Process. Int. J. 22 (15), 2847–2857. https://doi.org/10.1002/hyp.7026.
1994. Measuring phenological variability from satellite imagery. J. Veg. Sci. 5 (5),
Chasmer, L., Baker, T., Carey, S.K., Straker, J., Strilesky, S., Petrone, R., 2018.
703–714. https://doi.org/10.2307/3235884.
Monitoring ecosystem reclamation recovery using optical remote sensing:
Scarlett, S.J., Petrone, R.M., Price, J.S., 2017. Controls on plot-scale evapotranspiration
comparison with field measurements and eddy covariance. Sci. Total Environ. 642,
from a constructed fen in the Athabasca oil sands region, Alberta. Ecol. Eng. 100,
436–446. https://doi.org/10.1016/j.scitotenv.2018.06.039.
199–210. https://doi.org/10.1016/j.ecoleng.2016.12.020.
Clark, M.G., Humphreys, E., Carey, S.K., 2019. The initial three years of carbon dioxide
Speckman, H.N., Frank, J.M., Bradford, J.B., Miles, B.L., Massman, W.J., Parton, W.J.,
exchange between the atmosphere and a reclaimed oil sand wetland. Ecol. Eng. 135,
Ryan, M.G., 2015. Forest ecosystem respiration estimated from eddy covariance and
116–126. https://doi.org/10.1016/j.ecoleng.2019.05.016.

12
M.G. Clark et al. Agricultural and Forest Meteorology 325 (2022) 109127

chamber measurements under high turbulence and substantial tree mortality from severely as dry soil. Geophys. Res. Lett. 43 (18), 9686–9695. https://doi.org/
bark beetles. Glob. Chang. Biol. 21 (2), 708–721. https://doi.org/10.1111/ 10.1002/2016GL069416.
gcb.12731. Sutherland, G., Chasmer, L.E., Kljun, N., Devito, K.J., Petrone, R.M., 2017. Using high
Straker, J.R., Carey, S.K., Petrone, R.M., Baker, T.D., Strilesky, S.L. 2019. Developing a resolution LiDAR data and a flux footprint parameterization to scale
functional approach to assessment of land capability: utilizing ecosystem water and evapotranspiration estimates to lower pixel resolutions. Can. J. Remote Sens. 43 (2),
carbon nutrient fluxes as integrated measure of reclamation performance. Canada’s 215–229. https://doi.org/10.1080/07038992.2017.1291338.
Oil Sands Innovation Alliance, 2021, www.cosia.ca. Volik, O., Kessel, E., Green, A., Petrone, R., Price, J., 2021a. Growing season
Statistics Canada, 2016. 2016 Census - Boundary files. https://www12.statcan.gc. evapotranspiration in boreal fens in the Athabasca oil sands region: variability and
ca/census-recensement/2011/geo/bound-limit/bound-limit-2016-eng.cfm. environmental controls. Hydrol. Process. 35 (2) https://doi.org/10.1002/hyp.14020
(Accessed 15 March 2022). p.e14020.
Stephens, J.J, Black, T.A., Jassal, R.S., Nesic, Z., Grant, N.J., Barr, A.G., Helgason, W.D., Volik, O., Petrone, R., Kessel, E., Green, A., Price, J., 2021b. Understanding the peak
Richardson, A.D., Johnson, M.S., Christen, A., 2018. Effects of forest tent caterpillar growing season ecosystem water-use efficiency at four boreal fens in the Athabasca
defoliation on carbon and water fluxes in a boreal aspen stand. Agric. For. Meteorol. oil sands region. Hydrol. Process. https://doi.org/10.1002/hyp.14323 p.e14323.
253, 176–189. https://doi.org/10.1016/j.agrformet.2018.01.035. Webb, E.K., Pearman, G.I., Leuning, R., 1980. Correction of flux measurements for
Strilesky, S.L., Humphreys, E.R., Carey, S.K., 2017. Forest water use in the initial stages density effects due to heat and water vapour transfer. Q. J. R. Meteorol. Soc. 106
of reclamation in the Athabasca oil sands region. Hydrol. Process. 31 (15), (447), 85–100. https://doi.org/10.1002/qj.49710644707.
2781–2792. https://doi.org/10.1002/hyp.11220. Wiken, E., Gauthier, D., Marshall, I., Lawton, K., Hirvonen, H., 1996. A Perspective On
Sulman, B.N., Roman, D.T., Yi, K., Wang, L., Phillips, R.P., Novick, K.A., 2016. High Canada’s Ecosystems: an Overview of the Terrestrial and Marine Ecozones, 14.
atmospheric demand for water can limit forest carbon uptake and transpiration as Canadian Council on ecological Areas, Ottawa, Ontario, Canada. Occasional paper.

13

You might also like