Introduction
Forest loss is one of the most critical threats to the biosphere in the
Anthropocene (Ruddiman, 2013; Malhi et al., 2014). Impacts of forest
loss are extensive and far-reaching (Morris, 2010)—driving species
extinctions (Brook et al. 2003; Giam, 2017), climate change (Ramankutty
et al., 2006; Lawrence & Vandecar, 2015), and threatening the health of
local human communities (Myers et al., 2013). The consequences of forest
loss are relatively well-documented (Gibson et al., 2011; Barlow et al.,
2016) so our understanding of its impacts on terrestrial ecosystems and
biodiversity is fairly robust. However, these repercussions are rarely
localised to the immediate (terrestrial) area (Lawton et al., 2001;
MacKenzie, 2008) and often affects adjacent ecosystems (Maina et al.,
2013). Forest streams, for example, are impacted by a reduction in
terrestrial inputs (e.g., leaf litter, woody debris) which are important
for the creation and maintenance of microhabitats (Giam et al., 2015;
Naman et al., 2018). This results in simplified systems which lack niche
diversity, and are hence less speciose (Loke & Todd, 2016).
Unsurprisingly, forest loss is generally detrimental to freshwater
biodiversity and commonly results in the extirpation of sensitive
species (Liew et al., 2018a, Wilkinson et al., 2018).
In addition, allochthonous terrestrial inputs have, until recently, been
thought of as constituting an important resource subsidy for low-order
forest streams (Lau et al., 2013). This seemed logical given the
expected abundance of such inputs (e.g., high standing stocks of leaf
litter), and lower internal (aquatic) primary productivity resulting
from light attenuation under dense canopy cover. As such, it was
generally assumed that forest loss in catchments would result in
resource limitations, thus impacting stream food webs from a bottom-up
(resource-driven) direction (Liew et al., 2018). Recent data appear to
cofound these expectations, however. Stream consumers grow at a
significantly lower rate on diets comprising exclusively of leaf litter
(Lau et al., 2013; Guo et al., 2016), suggesting that terrestrial
resource subsidies may play a more peripheral role in steam food webs
than previously supposed (Lau et al., 2009; Brett et al., 2017).
Current uncertainties about the role of allochthonous terrestrial
subsidies in stream food webs complicate predictions about the effects
of forest loss and its underlying mechanisms. It is perhaps for this
reason that studies investigating changes in aquatic food webs in
response to anthropogenic impacts sometimes report conflicting findings.
For example, Takimoto et al. (2008) observed no significant links
between disturbance and food webs—or more specifically, food chain
length—while McHugh et al. (2010) found disturbance to be one of the
primary predictors of food chain length. The lack of a clear consensus
is critical from a conservation perspective. Disrupted food webs are
thought to impair ecosystem functions, especially those related to
energy flow (Holt & Loreau, 2001). Moreover, communities associated
with functionally-impacted ecosystems are also often more vulnerable to
further species loss (Chua et al., 2019).
In this paper, we aim to clarify the relationship between forest loss
and freshwater food webs using a longitudinal (i.e., before and after)
study design to help minimise potential confounders (e.g., baseline
differences in the communities surveyed). We do this by measuring
changes in forest cover and food web structure at two time points
approximately two decades apart in tropical Southeast Asia, a region
presently undergoing significant forest loss (Miettinen et al. 2012). We
used a combination of state-of-the-art ecological tracers—i.e., amino
acid-specific Carbon-13 (Liew et al., 2019) and Nitrogen-15 (Chikaraishi
et al., 2009) stable isotopes—and high-resolution remote-sensing, as
both methods are currently the most precise tools for measuring food web
indices and catchment forest cover in a longitudinal design.
At each time point, we measure two food web indices, namely, the
proportion of consumer tissue comprising assimilated terrestrial organic
carbon (which we interpret as the extent of allochthony ) and
trophic position (TP ) of representative apex predators (Choy et
al. 1996). Together, allochthony and TP provide
information about the role of terrestrial subsidies as a basal resource
(Liew et al., 2019) and the vertical complexity of food webs (Takimoto
& Post, 2013; Digel et al., 2014). Conversely, we measured catchment
forest cover at both time points by generating raster data of cloud-free
USGS Landsat mosaic images using machine-learning algorithms. With this,
we asked: (1) how have freshwater food webs changed over time?; and (2)
can these changes be predicted by the extent of forest loss in the
respective catchments?