You are on page 1of 24

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/351822091

Spatiotemporal Peatland Productivity and Climate Relationships Across the


Western South American Altiplano

Article in Journal of Geophysical Research: Biogeosciences · June 2021


DOI: 10.1029/2020JG005994

CITATIONS READS

3 210

5 authors, including:

Duncan A. Christie Roberto Chávez


Universidad Austral de Chile Pontificia Universidad Católica de Valparaíso
82 PUBLICATIONS 3,902 CITATIONS 29 PUBLICATIONS 434 CITATIONS

SEE PROFILE SEE PROFILE

Matías Olea
Pontificia Universidad Católica de Valparaíso
4 PUBLICATIONS 64 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

AVOID - Adaptation and Vulnerability of Nothofagus forests to drought-Induced risks and disasters in Central and Southern Chile View project

Development of durum wheat in Chile View project

All content following this page was uploaded by Duncan A. Christie on 17 June 2021.

The user has requested enhancement of the downloaded file.


RESEARCH ARTICLE Spatiotemporal Peatland Productivity and Climate
10.1029/2020JG005994
Relationships Across the Western South American
Key Points:
• W
 e develop temporally continuous
Altiplano
Normalized Difference Vegetation Talia G. Anderson1,2 , Duncan A. Christie3,4 , Roberto O. Chávez5 , Matias Olea5 , and
Index data sets at the individual
Kevin J. Anchukaitis1,2
unit, 1 km2 and 5 km2 scales for
the network of Chilean Altiplano 1
School of Geography, Development and Environment, University of Arizona, Tucson, AZ, USA, 2Laboratory of
peatlands
• Productivity of peatlands is strongly Tree-Ring Research, University of Arizona, Tucson, AZ, USA, 3Laboratorio de Dendrocronología y Cambio Global,
associated with lagged water Instituto de Conservación Biodiversidad y Territorio, Universidad Austral de Chile, Valdivia, Chile, 4Center for Climate
availability from rain and snow in and Resilience Research (CR)2, Santiago, Chile, 5Laboratorio de Geo-Información y Percepción Remota, Instituto de
the western Altiplano
Geografía, Pontificia Universidad Católica de Valparaíso, Valparaíso, Chile
• The recent jump in productivity
appears to be associated with years
of anomalously high precipitation
Abstract The South American Altiplano is one of the largest semiarid high-altitude plateaus in
the world. Within the Altiplano, peatlands known as “bofedales” are important components of regional
Supporting Information:
hydrology and provide key water resources and ecosystem services to Andean communities. Warming
Supporting Information may be found
in the online version of this article.
temperatures, changes in hydroclimate, and shifting atmospheric circulation patterns all affect peatland
dynamics and hydrology. It is therefore urgent to better understand the relationships between climate
variability and the spatiotemporal variations in peatland productivity across the Altiplano. Here, we
Correspondence to:
T. G. Anderson,
explore climate influences on peatland vegetation using 31 years of Landsat data. We focus specifically on
taliaanderson@email.arizona.edu the bofedal network in the western Altiplano, the driest sector of the plateau, and use the satellite-derived
Normalized Difference Vegetation Index (NDVI) as an indicator of productivity. We develop temporally
Citation: and spatially continuous NDVI products at multiple scales in order to evaluate relationships with climate
Anderson, T. G., Christie, D. A., variables over the past three decades. We demonstrate that cumulative precipitation and snow persistence
Chávez, R. O., Olea, M., & Anchukaitis, over the prior 2 years are strongly associated with growing season productivity. A step change in peatland
K. J. (2021). Spatiotemporal
peatland productivity and climate
productivity between 2013–2015 drives an increasing trend in NDVI and is likely a response to consecutive
relationships across the western years of anomalously high snow accumulation and rainfall. Early summer minimum temperatures
South American Altiplano. Journal of emerge as a secondary influence on productivity. Understanding large-scale productivity dynamics and
Geophysical Research: Biogeosciences,
126, e2020JG005994. https://doi.
characterizing the response of bofedales to climate variability over the last three decades provides a
org/10.1029/2020JG005994 baseline to monitor the responses of Andean peatlands to climate change.

Received 3 AUG 2020 Plain Language Summary A unique network of peatlands, locally known as bofedales,
Accepted 14 MAY 2021 are found in the highlands of the Central Andes in South America. We specifically focus on the Chilean
bofedal network and evaluate year to year changes in vegetation productivity, as represented by
“greenness” (Normalized Difference Vegetation Index). We find that the accumulation of snow and rain
over two years is an important climate influence on subsequent bofedal growing season productivity
across the region. We also show that early summer minimum temperatures have a secondary influence
on bofedal productivity at the regional level. Finally, we show that a recent greening (2013–2015) was
preceded by years of high snow and rain accumulation.

1. Introduction
Wetlands play a significant role in regulating global climate by sequestering large amounts of carbon
and releasing major greenhouse gases including carbon dioxide, methane, and nitrous oxide (Moomaw
et al., 2018). Recent estimates indicate that peat-accumulating wetlands alone contain more than 20% of the
global soil organic carbon stock, making them one of the largest carbon sinks in the world (R. B. Jackson

only ∼3% of Earth’s surface, these carbon stocks are comparable to the amount stored in forest biomass
et al., 2017; Köchy et al., 2015; Leifeld & Menichetti, 2018; Yu et al., 2010). Even though peatlands cover

globally (Moomaw et al., 2018; Y. Pan et al., 2011; Yu et al., 2010). Consequently, the responses of peat-
© 2021. American Geophysical Union. lands to climate variability and change have important implications for the global carbon cycle (Charman
All Rights Reserved. et al., 2015). While most peatlands are concentrated in high-latitude boreal and subarctic regions of the

ANDERSON ET AL. 1 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

Figure 1. (a) Study area (pink) and mean annual precipitation (mm; Southern Hemisphere phenological year) for 1980 to 2018. Precipitation is from the
CR2MET product (Boisier, 2019). Our peatland study area was defined by Chávez, Christie et al. (2019) and includes the entire Chilean Altiplano above 3,500
masl and a 30 km buffer beyond Chile’s country limits; however, the few bofedales below 3,700 masl were excluded due to their small size (Chávez, Christie
et al., 2019). (b) Photo from October 2019 of Andean peatland from the area of Cosapilla in Chile (17.8°S, 69.4°W) at approximately 4,230 masl.

Northern Hemisphere (NH) (Yu et al., 2010), those in the Southern Hemisphere (SH) Central Andes have
some of the largest long-term carbon accumulation rates known for high-elevation or high-latitude peat-
land ecosystems globally (Cooper et al., 2015; Earle et al., 2003; Hribljan et al., 2015).

Central Andes peatlands, locally known as “bofedales,” are found from 3,200 masl to over 5,000 masl across
the South American Altiplano, a high-elevation, semi-arid plateau situated between the east and west An-
des Cordilleras. Unlike NH peatlands that are primarily dominated by Sphagnum mosses, bofedales are
most commonly composed of peat-forming cushion plants in the Juncaceae, Asteraceae, and Plantaginace-
ae families (Cooper et al., 2010; Ruthsatz, 2012; Squeo, Warner et al., 2006). Bofedales are an important fea-
ture of local hydrology; they provide accessible water and critical forage resources for wild herbivores and
the livestock of local indigenous communities and support a variety of endemic mammal and bird species
(Engel et al., 2014; Gandarillas et al., 2016; Ledru et al., 2013; Ruthsatz, 2012; Schittek et al., 2016; Squeo,
Warner et al., 2006; Yager et al., 2019). Although the western Altiplano is the driest sector of the high-eleva-
tion plateau and is marked by difficult growing conditions including short growing seasons, high-intensity

ales in the Chilean sector covers over 500 km2 and ranges from ∼3,700–5,100 masl between 17.3°S–24°S
winds, large fluctuations in temperature, and daily frosts (Squeo, Warner et al., 2006), the network of bofed-

(Figure 1) (Chávez, Christie et al., 2019). In the otherwise arid and sparsely vegetated landscape bordering
the Atacama desert, bofedales are important water resources for local communities and components of
regional hydrology (Squeo, Warner et al., 2006).

In addition to these ecosystem services, bofedales are important paleoenvironmental and paleoclimate ar-
chives. Bofedales can have peat bodies as much as 10 m thick and may accumulate organic matter over
thousands of years (Earle et al., 2003; Hribljan et al., 2015; Schittek et al., 2016). Andean peatland ages range
from around 1,000 years to over 8,000 years before present, which is indicative of various periods of forma-
tion throughout the Holocene and changes in vegetation composition over time (Earle et al., 2003; Hribljan

ANDERSON ET AL. 2 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

et al., 2015; Schittek et al., 2016; Squeo, Warner et al., 2006). Previous studies of bofedal sediment cores
have shown that they are sensitive to moisture conditions over multiple millennia and can provide insight

bofedales at the southern-most margin of the Chilean Altiplano ( ∼27°S), Earle et al. (2003) found carbon
into environmental changes at sub-centennial to decadal scales (Ledru et al., 2013; Schittek et al., 2016). In

accumulation rates at least an order of magnitude higher than NH peatlands. High carbon accumulation
rates have similarly been found in Andean peatlands near the Chilean border in Bolivia (Cooper et al., 2015;
Hribljan et al., 2015). However, the expansive bofedal network could potentially shift from a major carbon
sink to a source as a result of environmental pressures including land use change, climate change, over-
grazing, and large-scale mining. Since these ecosystems have specifically been identified as sentinels to
the effects of recent climate change (Dangles et al., 2017), projections of future peatland productivity and
its effect on the global carbon cycle need to account for potentially spatially heterogeneous sensitivities to
climate factors (Charman et al., 2015).

While previous studies have evaluated bofedal productivity and climate relationships in different sectors
of the Altiplano plateau at varying spatiotemporal scales, none have yet characterized these relationships
in the Chilean Altiplano (Casagranda et al., 2019; Chávez, Christie et al., 2019; Dangles et al., 2017; Gar-
cia & Otto, 2015; García et al., 2019; Polk et al., 2017; Squeo, Warner et al., 2006). Earlier assessments of
bofedal productivity show unique connections to climate with variable strengths across distinct watersheds
and locations throughout the Altiplano (Casagranda et al., 2019; Dangles et al., 2017; Garcia & Otto, 2015;
García et al., 2019; Polk et al., 2017). While productivity is often linked to moisture availability, only one
study characterizes Altiplano peatland responses to climate changes continuously over multiple decades
(Dangles et al., 2017). The spatially differentiated peatland responses to climate and lack of temporally con-
tinuous studies highlight the need to evaluate these relationships in Chile specifically. Although it has been
established that Chilean bofedales are sustained by precipitation and the perennial inflow of groundwater,
likely stemming from a combination of rain and snowmelt (and glaciers in a few instances), it is uncertain
how each factor contributes to overall productivity and if temperatures or growing season length play a
significant role in interannual vegetation productivity in this region (Squeo, Warner et al., 2006). Evidence
that warming is more rapid in mountain regions (Pepin et al., 2015) further underscores the urgency in
understanding how climate variability affects these ecosystems.

Given that bofedales are located near both elevation and climate thresholds for plant life, they are expected to
be highly sensitive to the changing climate conditions in the Altiplano. Observations already reveal shrink-
ing of glaciers and warming temperatures in the Central Andes highlands (Rabatel et al., 2013; Thibeault
et al., 2012; Vuille et al., 2003, 2015). The observed and projected temperature increases in the Altiplano are
larger than the global mean and are expected to continue to be so (Bradley et al., 2006; Russell et al., 2017).
Despite these clear tendencies, recent trend analyses of Andes snow cover duration (2000–2016) show few
significant changes north of 29°S (Saavedra et al., 2018). In terms of precipitation, instrumental station data
do not reveal spatially consistent trends in rainfall over the last century (Neukom et al., 2015). However,
a recent analysis of Altiplano summer precipitation between 12°S–20°S does exhibit an increasing signifi-
cant trend since the 1980’s, associated with regional changes in the precipitation driving processes (Segura
et al., 2020). While climate model projections have suggested that summers will get drier and experience
less frequent rainfall events (Minvielle & Garreaud, 2011; Neukom et al., 2015; Thibeault et al., 2012), shift-
ing atmospheric circulation patterns that are known to drive Altiplano precipitation variability have the
potential to influence these trends (Segura et al., 2020). In addition to the impacts of anthropogenic climate
change, large-scale mining operations add to the growing demand for limited water resources in the Alti-
plano (Camacho, 2012). Chile holds over 50% of the world’s lithium reserves and plans to double lithium
extraction in the next several years in conjunction with a significant expansion of copper production (Jofré
et al., 2020; Minería Chilena, 2019). These changes will certainly impact both ecological communities and
water availability in the region. For this reason, it is of critical importance to understand the spatiotemporal
variations in and climate controls on bofedal productivity across the Chilean Altiplano.

Here, we build on a recently developed inventory of bofedales based on 31 years of Landsat data for the
entire Chilean Altiplano from Chávez, Christie et al. (2019). From this inventory we create temporally con-
tinuous monthly 1 km2, 5 km2, and bofedal unit time series of productivity from 1988–2017. We use these
data sets to determine the large-scale patterns of bofedal productivity and their relationships to regional

ANDERSON ET AL. 3 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

Figure 2. Climatology based on the Southern Hemisphere phenological year (July of calendar year n-1 to June of year n). Mean monthly precipitation (mm),
maximum temperature (°C), minimum temperature (°C), and Normalized Difference Vegetation Index (NDVI) are shown in graph. Mean NDVI is calculated
from the bofedal unit time series and climate variables are calculated from the CR2MET product (Boisier, 2019) for all land pixels above 3,500 masl between
17°S–24°S and 71°W–66°W. All metrics used in this paper are shown above the graph and assigned to year n for ease of comparison and to be consistent with
productivity, which is the primary focus of this study. SP refers to snow persistence, the fraction of a year with snow cover (%). The standardized precipitation
index (SPI) is used to represent cumulative precipitation over 26-month period ending in February of year n. Both SP and SPI are annual values and are
therefore not represented in the climatology graph.

climate variability across the high altitude, semiarid region characterized by a moisture gradient. A better
understanding of climate relationships and interannual productivity differences will provide further insight
to current and future changes for this unique ecosystem across the driest sector of the Altiplano plateau.

2. Methods & Materials


2.1. Climate Setting of the Altiplano

With an average elevation of 4,000 masl, the South American Altiplano spans over 1,000 km north to south
(15°S–25°S) through parts of Bolivia, Peru, Argentina, and Chile (Garreaud et al., 2003; Vera et al., 2019).
It is situated between the eastern and western Andes Cordilleras and within a region of limited rainfall
known as the “Arid Diagonal” (Abraham de Vazquez et al., 2000). Located between the Amazon lowlands
to the east and Atacama Desert to the west, the plateau is marked by a northeast to southwest decreasing
moisture gradient (Morales et al., 2012). In our study area, maximum total annual precipitation reaches
roughly 544 mm in the north while the minimum is only 41 mm in the south (Figure 1a). The Altiplano
has a distinct seasonal cycle in precipitation and relatively small seasonal differences in temperatures, re-
sulting in cool-dry austral winters and cool-wet austral summers (Figure 2) (Garreaud et al., 2003; Vuille
& Bradley, 2000). The majority of precipitation ( >70% in Chile) falls during the austral summer (DJF) and

ANDERSON ET AL. 4 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

is locally determined by a combination of the complex topography and large-scale circulation mechanisms
(Garreaud et al., 2003; Segura et al., 2020; Vuille & Keimig, 2004). Interannual precipitation variability
has been linked to the South American monsoon and the position of the Bolivian High, an anticyclonic
system that produces upper level easterlies bringing moisture-laden air from the Amazon to the Andes
highlands (Garreaud et al., 2003; Vera et al., 2006; Vuille & Keimig, 2004). Tropical Pacific variability has
also been associated with climate anomalies across the Altiplano, with El Niño (La Niña) events leading to
strengthened (weakened) westerlies and generally drier (wetter) conditions (Bradley et al., 2003; Garreaud
& Aceituno, 2001). Temperatures have strong diurnal fluctuations and although intraannual differences are
relatively low, early austral summer temperatures (end of November) have been shown to control the begin-
ning of the growing season (Squeo, Warner et al., 2006). In the semi-arid plateau, snow cover also plays an
important role in the hydrological cycle because of the short rainy seasons (Saavedra et al., 2017). Between
14°S–23°S snow is primarily found in high elevation areas ( >5,000 masl) and reaches a maximum extent
toward the end of the austral summer (March) (Saavedra et al., 2017). Both tropical glaciers atop volcanoes
and rock glaciers are found along the Cordilleras in this region (Barcaza et al., 2017). While the majority of
precipitation still falls in the austral summer at the southernmost part of this study’s domain (23°S–24°S),
winter snow events are more frequent and are not constrained to the high elevations (Saavedra et al., 2017;
Vuille & Ammann, 1997; Vuille & Keimig, 2004).

2.2. Western Altiplano Peatland Inventory

We build on the digital inventory of Chilean Altiplano bofedal productivity developed by Chávez, Christie
et al. (2019) to construct the temporally continuous data sets of productivity used for climate comparisons
in this study. The original inventory spans 63,705 km2 between 17.3°S–24.0°S and 70.0°W–66.8°W including
a 30 km buffer beyond the national boundaries of Chile to prevent the artificial division of cross-border
bofedales. The same study area is considered here (Figure 1a). To delineate peatland units, Chávez, Christie
et al. (2019) use 31 years of Landsat Normalized Difference Vegetation Index (NDVI) data with a spatial
resolution of 90 m2 (Figure S1). NDVI is an indicator of vegetation productivity and relative greeness, and
represents the ratio between the near infrared (NIR) and red (R) spectral bands, where NDVI = (NIR − R)/
(NIR + R). It accounts for the absorption of light by chlorophyll in the red wavelengths and the reflectance
of NIR from the plant cellular structure, and is therefore sensitive to photosynthetically active green bio-
mass (Tucker, 1979). While there are various remotely sensed vegetation metrics, NDVI has proven to be
successful in mapping and capturing the dynamics of bofedales across the Altiplano (Baldassini et al., 2012;
Casagranda et al., 2019; Dangles et al., 2017; Garcia & Otto, 2015; Izquierdo et al., 2015; Moreau et al., 2003;
Otto et al., 2011; Polk et al., 2017). Other forms of high-altitude wetlands are found in the Altiplano, includ-
ing wet meadows and tall grasslands, but Oxychloe andina has been previously found to be the dominant
species in northern and southern sectors of the Chilean Altiplano, indicating the primary presence of peat-
land vegetation (Ruiz-Esquide, 2015).

Peatland area is fixed in the Chávez, Christie et al. (2019) inventory and was determined based on years
when a group of pilot study peatlands reached a maximum spatial extent. All individual pixels in the study
area were subsequently classified as peatland if NDVI surpassed a threshold of 0.23 in at least 75% of the
years of maximum extent defined by the pilot peatlands (Chávez, Christie et al., 2019). Only contiguous
peatlands above 3,700 masl and with an area ≥4,500 m2 were included in the final inventory (Chávez, Chris-
tie et al., 2019). An object-based accuracy assessment was used to compare manual and automatic delin-
eations of peatlands in order to validate the classification procedure (Ardila et al., 2012; Chávez, Christie
et al., 2019). The mean delineation agreement is 0.79 (possible range from 0 to 1) and accounts for both un-
derestimated and overestimated area (Chávez, Christie et al., 2019). Further detailed information regarding
the methods used to delineate the bofedales can be found in Chávez, Christie et al. (2019).

For this study, we use the NDVI time series for both the 90 m2 pixels and the individually delineated bofedal
units from the Chávez, Christie et al. (2019) inventory. We evaluate monthly NDVI data based on 30 SH
phenological years beginning in the 1987–1988 growing season and ending in 2016–2017. Phenological
year n begins in July of the calendar year n-1 and ends in June of year n (Figure 2). While the 90 m2 resolu-
tion was critical for delineating bofedales and understanding the productivity dynamics within individual
units, here we build on the Chávez, Christie et al. (2019) data set to develop temporally continuous, coarser

ANDERSON ET AL. 5 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

resolution NDVI data products to focus on large-scale regional patterns and climate relationships across the
entire Chilean Altiplano.

2.3. NDVI Inventory Upscaling at Multiple Spatial Resolutions

Prior to evaluating the large-scale bofedal productivity dynamics and climate relationships, it was necessary
to address the non-random missing data structures within the peatland inventory. Since at least 50% of
pixels are missing NDVI values for many months during the first phenological year (1986–1987) defined by
Chávez, Christie et al. (2019), we analyze SH phenological years beginning in 1987–1988. Cloud or snow/ice
cover and Landsat ETM sensor failure result in contiguous regions of missing data over parts of the domain
across multiple years, substantially limiting the capacity to accurately characterize large-scale relationships
between NDVI and climate. In addition to months with partially missing data, there are four dates with
no data at all for the entire study area (03/1994, 02/1995, 03/1997, and 04/1997). To address these limita-
tions, we infilled missing NDVI values through a multi-step aggregation and prediction procedure to create
the complete, monthly time series for (a) individual contiguous bofedal units, (b) a 990 m2 grid (∼1 km2),
and (c) a 5,010 m2 grid (∼5 km2). Testing different gapfilling methods and sizes ensured the aggregation
procedure and our results are robust across scales and through time. The coarser resolution also reduced
processing challenges associated with infilling multiple missing values for nearly 567,000 individual 90 m2
peatland pixels.

To characterize productivity patterns for individual bofedal units, we used the monthly median NDVI
time series from 07/1987–06/2017 for the 5,665 discrete bofedal polygons identified in Chávez, Christie
et al. (2019). Linear interpolation was applied to infill missing NDVI values for all time series. To ensure
the linear interpolation was appropriate for this data set (Goetz, 1997; T. J. Jackson et al., 2004; Z. Pan
et al., 2015), we carried out a cross-validation procedure by leaving out NDVI values and estimating them
using different interpolation methods. Since all units have at least four missing values due to the months
entirely without data, we randomly selected and removed NDVI values from 4 non-missing dates in the
original time series for each unit. We then imputed those values using both a linear and a spline interpola-
tion. Results from the validation procedure are found in (Table S1). The linear interpolation had an overall
lower validation root mean square error and standard deviation, indicating the predicted NDVI values are
more similar to the actual observations for this method, and is therefore used in all of the following analyses
in this paper.

For the 990 m2 and 5,010 m2 grids, we adopted a more complex gapfilling approach. The NDVI values for the
coarser grids are based on the median of all bofedal pixels within each larger pixel. Missing values are in-
filled using the “gapfill” R package (Gerber et al., 2018) and linear interpolation for the four dates in which
there is no existing satellite data for the region. A complete description of the upscaling and validation
procedures for the gridded data sets can be found in the Supporting Information.

2.4. Climate Data

Because of inconsistent recent climate trends and the complex spatial variability in rainfall across the Alti-
plano (Neukom et al., 2015), we used multiple monthly gridded precipitation products to characterize the
spatiotemporal variability in Altiplano climate. Monthly precipitation data include the Center for Climate
and Resilience Research Meteorological data set (CR2MET, Boisier (2019)) and the Climate Hazards Group
InfraRed Precipitation with Station data version 2.0 (CHIRPS; Funk et al. (2015)). The CHIRPS daily pre-
cipitation product is also used to evaluate extreme rainfall events defined here as the total amount of DJF
precipitation falling in events >95th percentile (R95p) (Funk et al., 2015; Zhang et al., 2011). Monthly maxi-
mum and minimum temperature data are from CR2MET (Boisier, 2019), which includes high elevation me-
teorological station records from the Altiplano unlike other high-resolution temperature data sets such as
the Climatic Research Unit (CRU) Time-series data version 4.03 (Harris & Jones, 2020; Harris et al., 2014).

Due to these limitations, we rely on the CR2MET precipitation and temperature products, which have 0.05°
spatial resolution and cover the time period from 1979–2018. The CR2MET precipitation field is based on sta-
tistical models that translate precipitation data from ERA-Interim (Dee et al., 2011) into better regional esti-
mates for Chile by incorporating topography and by calibrating with local meteorological stations primarily

ANDERSON ET AL. 6 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

maintained by the Chilean Water Authority (Boisier, 2019; Dirección General de Aguas (DGA), 2017).
CR2MET temperature estimates also incorporate local instrumental observations, topography, and large-
scale variables from ERA-Interim, but are further informed by surface temperatures from MODIS LST data
(Boisier, 2019). The CR2MET products are available at http://www.cr2.cl/datos-productos-grillados/. The
original CR2MET product extends 100 km over the Chilean border and is used for this study to account
for the cross-border bofedales [personal communication, JP Boisier]. We also identified CHIRPS as a tar-
get climate data set for precipitation because of its relatively high spatial resolution (0.05° × 0.05°) from
1981–present and good agreement with observations from meteorological stations in the Altiplano (Funk
et al., 2015; Segura et al., 2020; Torres-Batlló & Martí-Cardona, 2020). All climate data analysis is based land
pixels above 3,500 masl between 17.0°S–24.0°S and 70.0°W–66°W to encompass the entire bofedal study
area and to characterize the highland climate dynamics.

Based on initial exploratory correlation analyses between growing season NDVI of individual bofedal units
and climate data, we focused on austral summer (DJF) precipitation totals and minimum and maximum
December temperatures (early summer). In addition to analyzing DJF precipitation totals for CR2MET
and CHIRPS, we calculated the 26-month Standardized Precipitation Index (SPI) using these datasets. SPI
is a common index used to evaluate meteorological droughts and represents how the observed cumulative
precipitation, in our case based on 26-month totals, differs from the long-term climatological average (Gutt-
man, 1999; McKee et al., 1993). The integrative nature of SPI has been related to groundwater storage over
longer timescales (McKee et al., 1993). Here, we focus on the 26-month SPI ending in February to capture
and integrate summer precipitation variability from the current growing season through 2 years prior.

Finally, since some bofedales are in part fed by groundwater from snowmelt of high elevations (Squeo,
Warner et al., 2006), we evaluate relationships between productivity and snow cover. We use gridded annual
snow persistence (SP) data from 2001–2016 developed by Saavedra et al. (2018) for the Chilean Andes re-
gion between 8°S–36°S. SP is originally based on the MODIS 8-day 500 m Collection 5 Level 3 binary snow
cover product (MOD10A2), where each 8-day period is classified by maximum snow cover extent (Riggs
et al., 2006). SP is then calculated as the percentage of time in a given year that each 500 × 500 m pixel is
covered with snow (Saavedra et al., 2018). Saavedra et al. (2018) provides more detail related to this data
set. For climate comparisons, we only evaluated pixels with a mean annual SP greater than 7% and with no
more than 8 zero values (<50%). The 7% threshold is based on previous studies and prevents the inclusion
of MODIS pixels that might be falsely identified as snow (Hall et al., 2002; Saavedra et al., 2018). We masked
additional areas around large salt flats and lakes, which are sometimes misclassified as snow in this region
(Saavedra et al., 2017).

2.5. Climate & Productivity Analyses

We evaluate the spatiotemporal patterns of the bofedal productivity and climate data sets across our study
region using empirical orthogonal function (EOF) analyses (Jolliffe, 2003). EOF analysis is commonly used
in climate and ecological studies to isolate the large-scale spatiotemporal patterns in large but noisy data
sets (Hannachi, 2004). Based on the strongest shared patterns of all time series in a data set, EOF analysis
generates new times series, which we refer to as Principal Components (PCs) (Jolliffe, 2003). In this study,
we regularly use the first PC, the time series that captures the most common variance in the data, to un-
derstand the large-scale productivity and climate relationships. Each PC time series is associated with a
spatial expression (which we refer to as an EOF pattern), or map of loadings, which indicates how similar
the original time series at a given location is to the newly generated PC (Jolliffe, 2003). The leading EOF
patterns and associated time series (the PCs), were calculated for the three aggregated and infilled inventory
products (PC-BOF, PC-1KM, and PC-5KM) and the climate variables of interest including DJF precipitation
(PC-precip), December maximum and minimum temperature (PC-tmax, PC-tmin), February 26-month SPI
(PC-SPI) and annual SP (PC-SP). All data were standardized to have a mean of zero and standard deviation
of 1. For the EOF analysis of productivity, we use mean JFM NDVI to focus on growing season dynamics.
In order to be consistent across data sets, the JFM mean from calendar year n was assigned to phenological
year n despite the SH phenological year beginning in the calendar year n-1 (Figure 2). All grid cells and
bofedal units were included in the EOF analysis. We found that EOF analysis of the bofedal units, 1 km2
grid, and 5 km2 grid all resulted in a similar leading mode of variability, so our analyses proceeded using the

ANDERSON ET AL. 7 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

bofedal units (EOF1-BOF) and the 5 km2 (EOF1-5KM) data sets. When necessary, we also detrended those
PCs that had significant trends before evaluating interannual climate relationships. The presence of secular
trends in the NDVI and climate data can lead to potentially spurious correlation results.

We use a method called “field correlation analysis” (c.f. Livezey & Chen, 1983; Wigley & Santer, 1990) to
quantify the statistical association between a single time series and the time series at each grid point in a
spatiotemporal climate or productivity data set. Note that here “field” specifically refers to the three dimen-
sional (latitude, longitude, and time) data. These statistical analyses allow us to evaluate the relationships
between bofedal productivity and climate variables across space over the last three decades. For climate
data, individual months and seasonal combinations of up to 2 years prior to the average JFM NDVI value
were considered due to the complex and lagged relationships with groundwater and snowmelt. The season-
al combinations with the strongest relationships to JFM mean productivity are included in the results and
their corresponding assigned years are shown in Figure 2. We evaluated the correlation field significance
(c.f. Livezey & Chen, 1983) by repeatedly (n = 1,000) randomizing the temporal order of the single time se-
ries used in the field correlation and calculating the resulting number of significant correlations. This proce-
dure disrupts the temporal relationship while maintaining the spatial covariance structure of the field data.

Finally, the magnitude and significance of trends in mean JFM NDVI were calculated for individual bofedal
units and the 5 km2 gridded data products with the non-parametric Mann-Kendall significance test and the
Theil-Sen linear slope (Burkey, 2006; Sen, 1968; Theil, 1992).

3. Results
3.1. Large-Scale Productivity Patterns

The main pattern of bofedal productivity across the study region is represented by the first EOF pattern,
explaining 34% and 38% of the total variance in the bofedal units (EOF1-BOF) and 5,010 m2 (EOF1-5KM)
grid, respectively (Figures 3a and 3b). For EOF1-BOF, roughly 96% of the total 5,665 bofedales have positive

series are dominated by statistically significant (p ≪ 0.01) upward trends over the last 30 years and have
loadings. At least 96% of grid cells from EOF1-5KM also load positively. The associated leading PC time

similar patterns of interannual variability (r ≥ 0.98, p ≪ 0.01) (Figure 3c). While we conclude that the ag-
gregation and imputation method does not appear to have a significant effect on the leading mode from the
EOF analysis, we use both the detrended bofedal unit and 5 km2 original products in the field correlations
and trend analyses to assess potential differences at more localized scales. The second EOF explains only 9%
and 11% of the total variance, of the bofedal units (EOF2-BOF) and 5,010 m2 (EOF2-5KM) grid, respectively.
EOF2-BOF has concentrations of strong positive loadings in the central part of the domain and toward the
northern tip of the southern cluster of bofedales (Figure 3d). EOF2-5KM is dominated by a north-south gra-

and EOF2-5KM are similar (r = 0.73, p ≪ 0.01) (Figure 3f). While EOF2 may provide additional insight into
dient, with the strongest positive loadings in the south (Figure 3e). However, the time series of EOF2-BOF

sub-regional dynamics, only the first leading mode of bofedal productivity passes all the tests for statistical
significance and thus we focus on interpreting this dominant mode for the following climate analyses.

3.2. Precipitation and Temperature Field Correlations

Field correlations between productivity and precipitation reveal the importance of cumulative rainfall
across previous growing seasons. The detrended PC1-5KM is more highly correlated with CR2MET cu-
mulative precipitation from the current growing season (DJF) through 2 years prior (year n-2) than with
rainfall totals during only the current (year n) or the current and immediately previous (year n + year n-1)
growing seasons (Figure 4). Field significance testing reveals that correlations between the detrended PC1-
5KM and cumulative DJF precipitation from year n to year n-2 are significant (p ≤ 0.05). Figure 4 highlights
that the strength of the correlations greatly increases by incorporating austral summer precipitation from
2 years prior to the current growing season and that the leading mode of variability in JFM productivity
is positively correlated with cumulative austral summer precipitation from year n through year n-2 across
most of the domain.

ANDERSON ET AL. 8 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

Figure 3. Spatial loadings of the leading two empirical orthogonal function (EOF) modes for (a, d) the bofedal units and (b, e) the 5 km2 NDVI products.
Circles in a and d are scaled by the area of each bofedal unit. EOF1 (a, b) explains 34% and 38% of the variance in the original bofedal unit and 5 km2 NDVI data
sets respectively, while EOF2 (d, e) explains only 9% and 11%. Time series expansions of the leading modes for NDVI (Principal Components (PCs)) (c, f) are in
blue (PC-BOF) and black (PC-5KM) and are scaled to facilitate comparison.

a. YR
n b. YR + YR
n n-1 c. YR + YR
n n-1
+ YR
n-2
17°S 0.6 17°S 0.6 17°S 0.6

18°S 0.5 18°S 0.5 18°S 0.5


CORRELATION COEFFICIENT

19°S 0.4 19°S 0.4 19°S 0.4

20°S 0.3 20°S 0.3 20°S 0.3

21°S 0.2 21°S 0.2 21°S 0.2

22°S 0.1 22°S 0.1 22°S 0.1

23°S 0 23°S 0 23°S 0

24°S
71°W 70°W 69°W 68°W 67°W 66°W
-0.1 24°S
71°W 70°W 69°W 68°W 67°W 66°W
-0.1 24°S
71°W 70°W 69°W 68°W 67°W 66°W
-0.1

Figure 4. Field correlations between the detrended PC1-5KM for JFM mean NDVI (year n) and total CR2MET DJF precipitation from (a) year n, (b) year n
through year n-1, and (c) year n through year n-2. Areas within contours indicate significant correlations for p ≤ 0.05.

ANDERSON ET AL. 9 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

a. CR2MET & 5KM GRID b. CHIRPS & 5KM GRID


17°S 17°S
0.5 0.5

18°S 0.4 18°S 0.4

CORRELATION COEFFICIENT
0.3 0.3
19°S 19°S
0.2 0.2

20°S 0.1 20°S 0.1

0 0
21°S 21°S
-0.1 -0.1

-0.2 -0.2
22°S 22°S

-0.3 -0.3

23°S -0.4 23°S -0.4

-0.5 -0.5
24°S 24°S
71°W 70°W 69°W 68°W 67°W 66°W 71°W 70°W 69°W 68°W 67°W 66°W
c. CR2MET & BOFEDALES d. CHIRPS & BOFEDALES
17°S 17°S
0.5 0.5

18°S 0.4 18°S 0.4

CORRELATION COEFFICIENT
0.3 0.3
19°S 19°S
0.2 0.2

20°S 0.1 20°S 0.1

0 0
21°S 21°S
-0.1 -0.1

-0.2 -0.2
22°S 22°S

-0.3 -0.3

23°S -0.4 23°S -0.4

-0.5 -0.5
24°S 24°S
71°W 70°W 69°W 68°W 67°W 66°W 71°W 70°W 69°W 68°W 67°W 66°W

Figure 5. Field correlations between February 26-month PC1-SPI (standardized precipitation index) for the CR2MET
and CHIRPS precipitation data and detrended JFM mean NDVI (year n) for the 5 km2 grid and bofedal units.
Correlations are based on the full time period (1988–2017). Panels show correlations between (a) PC1-SPI CR2MET and
5 km2 gridded NDVI data, (b) PC1-SPI CHIRPS and 5 km2 gridded NDVI data, (c) PC1-SPI CR2MET and bofedal units,
and (d) PC1-SPI CHIRPS and bofedal units. Significance contours are not included because of the fine grid scale and
discontinuous nature of the field; correlations higher than ±0.36 are locally significant at the p ≤ 0.05 level.

While the relationship between JFM productivity and cumulative precipitation through 2 years prior is
prominent when considering the leading mode of variability in bofedal productivity, field correlations be-
tween February 26-month PC1-SPI for CR2MET and CHIRPS and detrended JFM mean NDVI (year n) for
the 5 km2 grid and bofedal units reveal varying strengths and relationships locally (Figure 5). For correla-
tions between CR2MET PC1-SPI and the 5 km2 grid, 32% grids have significant positive correlations ranging
from 0.36-0.76. These are primarily concentrated in the northern part of the domain. In this case, only 1% of
grids have significant negative correlations (Figure 5a). For CHIRPS PC1-SPI and the 5 km2 grid, less than
1% of grids have significant negative correlation and only 3% of grids have significant positive correlations
that range from 0.37-0.55 (Figure 5b). When considering bofedal units instead of the 5 km2 grid, we find
similar percentages of significant correlations, but more variable locations. 30% of bofedal units are signif-
icantly positively correlated with CR2MET PC1-SPI, with correlations ranging from 0.36-0.79. Significant

ANDERSON ET AL. 10 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

Figure 6. Field correlations between the detrended PC1-5KM for JFM mean NDVI (year n) and CR2MET (a) maximum and (b) minimum December
temperatures (year n), for the 1988–2017 period areas with contours indicate significant correlations for p ≤ 0.05.

correlations are more scattered across the domain, but with concentrations in the northern and south-
ernmost sectors. Similar to the 5 km2 grid, only 1% of bofedal units are significantly negatively correlated
with CR2MET PC1-SPI (Figure 5c). For CHIRPS PC1-SPI and the bofedal units, only 5% are significantly
positively correlated (r = 0.36–0.58) and less than 1% of bofedales were significantly negatively correlated
(Figure 5d). Interestingly, the largest bofedal in the Chávez, Christie et al. (2019) inventory and the sur-
rounding cluster, found at the bases of the Sajama and Parinacota volcanoes, is negatively correlated with
the 26-month February SPI. However, most of these correlations are not significant (p ≥ 0.05).

The bofedal units and 5 km2 grids show the strength and localization of the significant field correlations chang-
es depending on both the productivity and precipitation product. While SPI time series from CR2MET and
CHIRPS are consistent in terms of recording the timing and magnitude of wet periods, SPI departures vary in
sign and strength during other years. The combination of these differences leads to stronger (r = 0.36 − 0.79)
and more widespread correlations with the CR2MET PC1-SPI time series and the detrended bofedal and 5 km2
products. The overall lack of significant correlations in the southern part of the domain when considering the
5 km2 grid could be a result of the smoothing of variability across many small peatlands. Correlations between
the CR2MET PC1-SPI and detrended JFM NDVI are field significant for both bofedal units and 5 km2 grid
fields (p ≤ 0.01). While CHIRPS has relatively strong positive correlations in the northern part of the domain
with the 5 km2 grid, neither field correlation using CHIRPS passes our field significance test.

Field correlations between the detrended PC1-5KM and detrended maximum December temperatures from
year n are non-significant across almost the entirety of the domain (Figure 6a). However, December mini-
mum temperatures are significantly correlated with PC1-5KM across a much larger portion of the domain
(Figure 6b). While temperatures are not as strong of a control on large-scale interannual peatland varia-
bility as the variables associated with precipitation, 13% of the bofedal units and 14.6% of the 5 km2 grids
are significantly correlated with PC1-tmin. The PC1-tmin and PC1-tmax time series are not significantly
correlated and reflect distinct patterns of interannual variability.

3.3. Snow

Although the length of the MODIS SP record is only about half that of the Landsat NDVI time series, field
correlations between annual PC1-SP from year n-2 are positively correlated with detrended JFM productiv-

ANDERSON ET AL. 11 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

a. SP YR-2 & 5KM GRID b. SP YR-2 & BOFEDALES


17°S 17°S

0.6 0.6

18°S 18°S

0.4 0.4

CORRELATION COEFFICIENT
19°S 19°S

0.2 0.2

20°S 20°S

0 0

21°S 21°S

-0.2 -0.2

22°S 22°S

-0.4 -0.4

23°S 23°S

-0.6 -0.6

24°S 24°S
71°W 70°W 69°W 68°W 67°W 66°W 71°W 70°W 69°W 68°W 67°W 66°W

Figure 7. Field correlations between snow persistence (PC1-SP) from year n-2 and detrended mean JFM NDVI (year n) for the (a) 5 km2 gridded data and (b)
bofedal units. Correlations higher than ±0.51 are locally significant at the p ≤ 0.05 level.

ity in year n. Significant positive correlation coefficients range from r = 0.52–0.86 for the 5 km2 grid (33% of
all grids) and from r = 0.51 − 0.89 for the bofedal units (27% of all units). Both pass a field significance test
(p ≤ 0.05) (Figure 7). Significant correlations are scattered across the domain, but with high concentrations
in the southern cluster. Since SP is a measure of snow cover from January–December, SP from 2 years pri-

influences growing season productivity in year n. Snow cover is more extensive south of ∼22°S, but is also
or indicates that contributions of cumulative snow from two summers previous through one spring prior

concentrated at high-elevation peaks in the northern sector (Saavedra et al., 2017). Figure 8 shows the time
series for the detrended PC1-5KM overlapping with the PC1-SP, which is shifted 2 years forward. Both PC1-
5KM and PC1-SP record a decreasing pattern between 2006–2011, followed by anomalously high values
from 2013–2015.

3.4. Long-Term Productivity Trends or a Step Change?

The number of bofedales and grids with significant trends in JFM NDVI varies strongly depending on the
period of analysis. When the full 30 growing seasons are considered (1988–2017), we find positive trends
in 83% of bofedal units. When considering only the significant trends (p ≤ 0.05), 2,766 are positive (93% of
all significant trends or 49% of the total number of bofedales) (Figure 9c). For the full period and the 5 km2
grids, 51% of the total grids are marked by significant positive trends and only 5% show significant negative
trends (p ≤ 0.05) (Figure 9a). When the recent period of enhanced productivity is removed beginning in
2012, only 1,601 bofedal units (28% of all units) have significant trends in productivity (Figure 9d). Of these,
1,123 trends are positive (20% of all 5,665 bofedales). Similarly, fewer 5 km2 grids show significant changes
when considering the shorter time period; 13% of all grids show significant increasing trends and 10% show
significant negative trends (Figure 9b). This demonstrates that the recent and relatively abrupt increase in
productivity is strongly influencing the trend analyses and may reflect only a temporary greening instead of
a long-term secular trend. Continual updating of the NDVI peatland database (Chávez, Christie et al., 2019)
will be needed to assess whether or not this increase is sustained over a longer period of time.

3.5. Shifts in Extreme Precipitation Events

Using the daily CHIRPS product, we find a clear dipole between the north and south regions of the Chilean
Altiplano when evaluating changes in extreme precipitation (R95p) events (Figure 10a). Significant increas-

ANDERSON ET AL. 12 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

2
NDVI PC1-5KM DETRENDED
1.5 SP YR+2

0.5
ZSCORE
0

-0.5

-1

-1.5

-2
r=0.71, p=0.003
-2.5
2002 2004 2006 2008 2010 2012 2014 2016 2018
YEAR

Figure 8. Z-scores for the first principal components of detrended mean JFM NDVI from the 5 km2 gridded data (PC1-
5KM) and snow persistence (PC1-SP) shifted two years forward.

es in extreme austral summer rainfall events are concentrated in the north and decreases are found in the
central and southern sectors. The averaged time series of all pixels with significant increasing trends shows
that years following 2010 are marked by anomalously high cumulative precipitation totals for days that are
over the 95th percentile (Figure 10b).

4. Discussion
4.1. Climate Influences on Productivity

The field correlation results demonstrate the complexities of lagged relationships between climate and
growing season productivity of the bofedales. Cumulative growing season precipitation from the two
previous years has a much stronger relationship with the large-scale pattern in bofedal productivity
than current growing season precipitation alone (Figure 4). The majority of previous studies that have
evaluated relationships between climate and bofedal NDVI have not explored lags much beyond a year
(Casagranda et al., 2019; Dangles et al., 2017; Garcia & Otto, 2015), but have noted the importance of
precipitation in the form of both rain and snow. Total bofedal area in Bolivia was significantly corre-
lated with cumulative precipitation from the prior 3 months based on data from one meteorological
station (Dangles et al., 2017). Because wetland area revealed an increasing trend over the study period
while total precipitation did not, Dangles et al. (2017) hypothesized that bofedal area increases could
be a result of an increasing frequency in extreme heavy rainfall events. Alternatively, the documented
decrease in glacial extent in certain locations could be providing a short-term increases in water availa-
bility in high elevation areas like the Altiplano (Dangles et al., 2017). In Peru’s Cordillera Blanca, Polk
et al. (2017) found an increase in wetland area when comparing differences between 2 years (1987
and 1995), which was associated with a loss in glacial cover and an increase in mean annual stream
discharge from the previous 12 months. Although a few tropical glaciers exist in the northern sector
of our study area, only relatively small rock glaciers can be found between 21°S–24°S in the Chilean
Andes (Barcaza et al., 2017). Total CRU precipitation from the previous 12 months was also found to be
a significant predictor variable for bofedal area in Peru (Polk et al., 2017). Recognizing the importance
of longer lags, Polk et al. (2017) considered a 6 year lag based on the availability of data, but it was not
significant. In the Argentine Altiplano adjacent to the administrative borders with Chile and Bolivia,
including some of the southernmost part of our study area, Casagranda et al. (2019) found no signif-

ANDERSON ET AL. 13 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

a. 5KM GRID (1988-2017) 10 -3 b. 5KM GRID (1988-2011) 10 -3


17°S 8 17°S 8

18°S
6 18°S
6

4 4
19°S 19°S

2 2
20°S 20°S

SLOPE
0 0

21°S 21°S

-2 -2

22°S 22°S
-4 -4

23°S 23°S
-6 -6

24°S
71°W 70°W 69°W 68°W 67°W 66°W
-8 24°S
71°W 70°W 69°W 68°W 67°W 66°W
-8
-3 -3
c. BOFEDALES (1988-2017) 10 d. BOFEDALES (1988-2011) 10
17°S 8 17°S 8

18°S
6 18°S
6

4 4
19°S 19°S

2 2
20°S 20°S

SLOPE
0 0

21°S 21°S

-2 -2

22°S 22°S
-4 -4

23°S 23°S
-6 -6

24°S
71°W 70°W 69°W 68°W 67°W 66°W
-8 24°S
71°W 70°W 69°W 68°W 67°W 66°W
-8

Figure 9. Trends in mean JFM NDVI for the 5 km2 gridded data and individual bofedal units from (a, c) 1988–2017 and (b, d) 1988–2011. Only grids and
bofedales with statistically significant trends (Mann-Kendall, p ≤ 0.05) are shown.

icant correlations between NDVI and accumulated precipitation from November–May of the current
year based on data from the Tropical Rainfall Measuring Mission. This is consistent with our results
when only considering total DJF precipitation in year n (Figure 4a).

ANDERSON ET AL. 14 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

a.
17°S 1 b. 80 500

0.8 400

ANNUAL P
18°S 60

R95p
0.6 300

19°S 40
0.4 200

0.2
20 100

MM/YEAR
20°S
1980 1985 1990 1995 2000 2005 2010 2015 2020
0 YEAR
21°S
-0.2
c. 80 500

-0.4 400
22°S

ANNUAL P
60

R95p
-0.6 300
23°S 40
-0.8 200

24°S
71°W 70°W 69°W 68°W 67°W 66°W
-1 20 100
1980 1985 1990 1995 2000 2005 2010 2015 2020
YEAR

Figure 10. (a) Trends (Theil-Sen slope) in DJF R95p from 1982–2018. Statistically significant trends (Mann-Kendall, p ≤ 0.05) are outlined in black. (b) Average
R95p (blue; mm) and annual July–June precipitation (gray; mm) time series for all grids with statistically significant positive trends. (c) Average R95p (red; mm)
and annual July–June precipitation (gray; mm) time series for all grids with statistically significant negative trends.

In this study, we find a direct relationship between large-scale patterns in bofedal productivity in the driest
part of the Altiplano and cumulative growing season precipitation from year n through year n-2. However,
we show that the strength of this general relationship varies locally, depends on the precipitation product
used, as well as the aggregated NDVI product to a degree. Correlations between JFM NDVI and CR2MET
precipitation are consistently stronger and more widespread across the entire region, showing that the
CR2MET interannual variability is more tightly linked to productivity than CHIRPS. When looking at the
cumulative 26-month February SPI for both products, the relationship to productivity is consistently signif-
icant and positive in the northern part of the domain for both bofedal units and the 5 km2 grids. However, a
number of the bofedal units are significantly correlated with cumulative SPI in the southern sector, which
does not appear in the 5 km2 gridded product. This suggests that the smoothing processes across bofedal
units in the southern region reduces important information and that certain local relationships are lost
with the coarsest grid, particularly where the bofedales are smaller in size and number (Chávez, Christie
et al., 2019).

Of all the climate variables, SP from year n-2 shows the strongest relationship with large-scale productivity
(r = 0.71 between PC1-SP and the detrended PC1-5KM). The sign of the signal is coherent regionally and

(south of ∼22°S). Given that total annual precipitation decreases toward the southern sector of our study
the southern region is dominated by significant positive correlations, where snow extent is much greater

area, the higher frequency of austral winter snowfall events in this area is a particularly important water
supply source (Vuille & Ammann, 1997). In the northern sector, snowfall is more constrained to higher
elevations (Saavedra et al., 2017) and clean-ice glaciers are relatively scarce, but melt contributions from
those along the Bolivia–Chile border including Nevado Sajama, Parinacota, Pomerape, and Guallatiri could
contribute to some of the significant correlations with SP in this region (Barcaza et al., 2017; Veettil &
Kamp, 2017). Rock glaciers represent an additional water source to certain bofedales in this region, but the
lack of data prevents characterizing their relative hydrological contributions (Barcaza et al., 2017; Schaffer
et al., 2019; Valois et al., 2021). While SP is a result of combined temperature and precipitation patterns,
annual precipitation has been previously found to be the most important driving factor in SP in this region

ANDERSON ET AL. 15 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

and is positively correlated with SP in our study area (Saavedra et al., 2018). Additionally, PC1-SP and
PC1-precip from CR2MET are significantly correlated for our study area (r = 0.55, p ≤ 0.05). The coupled
relationship between precipitation and snow explains the consistent 2-year lagged relationships we see
between productivity and these climate variables. These results suggest the total precipitation in the form
of rain and snow in year n-2 have an important relationship with productivity with varying contributions
across the study area.

Other research has also shown that distinct levels of photosynthetically active vegetation (PAV), as expressed
by NDVI, in Peruvian bofedales have varying relationships between different forms of precipitation (Garcia
& Otto, 2015). Garcia and Otto (2015) found that bofedales with consistently higher NDVI values through-
out the year were associated with snow (71% of variance explained in the spatial extent) and bofedales with
lower fractions of PAV were connected to summer rains (68% of the variance in spatial extent explained).
However, it is unclear if the opposite comparisons were made and if connections between rainfall and snow
were considered for both the categories (Garcia & Otto, 2015). While we do not distinguish between types of
bofedales based on NDVI thresholds, our large-scale analysis reveals that both forms of precipitation appear
to be relevant sources of water subsequently stored and used by peatland vegetation. In our case, significant
correlations between rainfall for both precipitation products are concentrated in the northern sector of the
study area and the strongest correlations with SP are in the south, where snow cover is primarily ephemeral
(Saavedra et al., 2017). This indicates that the different sources of precipitation are of varying importance in
water supply to peatlands across the study area, but that the two-year lag is regionally coherent.

Similar to Dangles et al. (2017) and Polk et al. (2017), our study does not include in-situ measurements of
groundwater fluxes or residence times that might explain the hydrogeological mechanism for lagged con-
nections between precipitation, snow, and peatland productivity. While Cooper et al. (2019) describe a fast
response of peatland water tables to rainfall events in Peru, they also found that groundwater levels were
generally stable over the 3 year monitoring period, indicating a relatively consistent groundwater supply
throughout the year. Several studies in the Altiplano have demonstrated that significant groundwater and
spring recharge occurs in high-elevation areas of catchments, originating from sources including precipi-
tation in the wet season and snow and glacier melt in the dry season, depending on the individual basin
(Herrera et al., 2016; Houston, 2009; Valois et al., 2021). Toward the southern sector of our study region,
stable isotopes from spring water that discharges into a laguna and multiple bofedales indicate that its re-
charge primarily originates from snow at higher elevations (Herrera et al., 2016). Gordon et al. (2015) also
show that high elevation recharge moves across steep elevational gradients through groundwater flow cells
and can pass through bofedales. In a basin located in the southern Cordillera Blanca of Peru, groundwater
was found to be the most significant contributor to basin outflow during the dry season and its temporal
variability was linked to precipitation from 3 to 36 months prior (Baraer et al., 2009). Such groundwater
contributions to the surface water system—of up to four years in this region—suggests the importance of
hydrological memory and lagged responses due to the presence of both short and long groundwater flow
paths (Alvarez-Garreton et al., 2021; Baraer et al., 2009). Although some studies show little to no input to
peatlands from streams originating at higher elevations (Cooper et al., 2019), others demonstrate significant
streamflow loss to groundwater reservoirs in peatlands (Valois et al., 2021). Our large-scale analysis appears
to capture the response of plant productivity to water availability across longer timescales, but future mul-
ti-year field based studies would be needed to explore productivity and vegetation composition responses
to differences in lags in water availability associated with processes across elevational gradients in distinct
catchments.

While it is known that bofedales store large amounts of precipitation from the wet season and inflow from
perennial groundwater, the groundwater residence times remain largely unknown and are an area of future
research (Glas et al., 2018; Valois et al., 2021; Wunderlich et al., 2019). One recent study, characterizing the
water storage capacity of a bofedal lower in elevation and to the south of our study region (30°S), found
that the peatland storage capacity was around 2,000 mm water per square meter of bofedal and that vertical
fluxes are very slow as a result of thin clay and silt layers (Valois et al., 2020). The deep rooting of cushion
plants also indicates that bofedales can access deep soil moisture below the water table and that they may
be able to buffer dry periods as a result (Chimner et al., 2019). At the same time as receiving inflows from
various sources, bofedales are also known to contribute to the recharge of aquifers and the local regulation

ANDERSON ET AL. 16 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

of streamflow (Gordon et al., 2015). The combination of these factors represents the complex hydrogeo-
logical dynamics at play that likely influence water fluxes and plant productivity responses across various
timescales (Alvarez-Garreton et al., 2021; Seimon et al., 2017). Additionally, little is known about the con-
tributions of fossil versus modern groundwater recharge, the relative inputs of the deep and more surface
level water sources, and the subsequent time it takes bofedal vegetation to respond to changes in the water
table, particularly in the drier Chilean Altiplano region (Cooper et al., 2019; Mosquera et al., 2016). How-
ever, other studies indicate that single- to multi-year lags in the productivity responses to changes in water
availability are common in semi-arid regions (Sala et al., 2012) and that they tend to occur at longer time
scales in comparison to humid and extremely arid regions (Vicente-Serrano et al., 2013).

Our results point to the need for additional multi-year and ground-based studies in the Altiplano to better
understand belowground hydrological processes and groundwater residence times to determine the mech-
anistic linkages between precipitation variability and bofedal productivity across different timescales. Since
both peatland and water table depths have been shown to be important factors related to the resiliency
and vulnerability of these ecosystems to future changes (Bridgham et al., 2008; Waddington, 2018), it is of
interest to pair the extensive temporal Landsat record with higher spatial and temporal resolution sensors
like Sentinel 2 or with active radar systems to gain additional insight into these relationships. Such pair-
ings could be used to evaluate features related to biomass (Alshammari et al., 2018), plant composition
(McPartland, Falkowski et al., 2019), water table levels (Bechtold et al., 2018), and peatland depth (Va-
lois et al., 2020). These analyses could reduce uncertainties associated with the limitations of this study
and provide an integrative approach to understanding differences in peatland dynamics and belowground
relationships.

In terms of temperature, longer growing seasons have been previously recorded as important drivers of
biomass accumulation over a 10-year period in the Cordillera de Doña Ana in central Chile (south of our
study region), but with a specific focus on only one bofedal (Squeo, Ibacache et al., 2006). More recent
analysis of vegetation dynamics across four Andean summits in NW Argentina showed that plant cover
is positively associated with minimum soil temperature (Carilla et al., 2018). Our results also indicate that
minimum early summer temperatures appear to have a influence on certain sectors of the Chilean peatland
network, but that the response to moisture remains stronger and more cohesive. While minimum temper-
atures emerge as a secondary influence, the lack of relationship to maximum temperatures may be partly
explained by the differences in trends and interannual variability in the observations. Upward trends in
maximum temperatures are stronger than those of minimum temperatures in both our study area and in
the Peruvian Altiplano (Imfeld et al., 2021). Imfeld et al. (2021) also find that Altiplano maximum tempera-
tures show less seasonal variation than minimum temperatures. Despite the weaker relationship with tem-
perature, projected warming in the Altiplano could result in more significant decreases in soil moisture and
increases in evapotranspiration, which can have significant effects on water table variability, peatland plant
community composition, and ecosystem structure, and in turn productivity (McPartland, Kane et al., 2019;
Seimon et al., 2017).

4.2. The Recent Green Period

The recent jump in productivity from 2013–2015 in both PC1-BOF and PC1-5KM follows three anomalously
high years of SP from 2011–2013. In July 2011, a rare snow event spanning the breadth of the entire study
area lasted several days and resulted in accumulation of up to 32 inches of snow in some places (Riebeek &
Carlowicz, 2011). Chávez, Moreira-Muñoz et al. (2019) also found a rare blooming desert event in the north-
eastern Atacama desert in the subsequent year; beginning in February and lasting through August 2012,
GIMMS NDVI data registered an extended anomalous green period. Such an event in this sector of the Ata-
cama had not been previously mentioned in the literature and was found to be correlated with precipitation
totals from 2–12 months in advance (Chávez, Moreira-Muñoz et al., 2019). PC1-precip for our region also
records anomalously high precipitation during that growing season (DJF 2012). Our R95p analysis further
shows that 2012 was marked by high amounts of rainfall concentrated in relatively few events and overall
higher annual rainfall, particularly in the northern part of our study area. While we do not explore the
climate mechanisms driving these precipitation extremes, Segura et al. (2020) shows recent increases in Al-
tiplano DJF precipitation are linked to enhanced connections to convective activity in the western Amazon.

ANDERSON ET AL. 17 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

The combination of these precipitation events reveals the significant accumulation of snow and anomalous
rainfall during this period, and could explain this more recent widespread greening, when considering the
lagged response of peatland vegetation across the Chilean Altiplano. The jump in productivity dominates
trend calculations over the entire study period and results in a much higher number of bofedales and grids
with a significant greening trend than when these years are not included (Figure 9). Because the greening
trend is strongly dependent on the time period used, it should be interpreted with caution or considered a
step change (Chávez, Christie et al., 2019).

Despite this recent green period described here and overall increases in peatland area in other sectors of the
Altiplano (Dangles et al., 2017; Polk et al., 2017), Aymara communities in Sajama National Park in Bolivia
have reported less healthy bofedales overall, dominated by shifts in vegetation cover from the long-standing
cushion plants to more seasonally green dry grasses (Yager et al., 2019). Ecological succession characterized
by the decline and mortality of cushion plants coupled with the colonization by grasses has also been ob-
served locally over a 10-year period in Peru where measurements show that decreases in the water table are
becoming more variable (Seimon et al., 2017). Similar decreases in peat-forming species and increases in
grasses are noted in a drained Andean peatland in Colombia (Benavides, 2014). While it is beyond the scope
of this study to verify species composition changes at the large scale, Dangles et al. (2017) also suggests that
the rapid growth of new wetlands is likely a result of increased Poaceae vegetation, rather than the estab-
lishment of cushion plants. Interviews conducted by Yager et al. (2019) indicate this decrease in healthy
area has accompanied climate changes including a decrease in glaciers, severe snow events, a “strong sun”
(which refers to warming temperatures), and reduced rains. While neither of the precipitation products
used in this study reveal significant decreasing trends in precipitation, we confirm the presence of signifi-
cant increasing maximum temperatures and snow accumulation in recent years and show that rainfall ex-
tremes are increasing in the northern part of the domain. Other studies indicate that interannual variability
in total DJF precipitation has decreased in recent years (Segura et al., 2020). The change in the distribution
in rainfall could challenge the ability to maintain the current network of bofedales (Yager et al., 2019). In
addition, the median values used for each bofedal and grid time series in this study are also likely influenced
by the presence of different vegetation types and possible changes to the peatland community composition
in response to changing water availability over longer timescales (Dangles et al., 2017; McPartland, Kane
et al., 2019; Polk et al., 2017; Seimon et al., 2017). As shown in previous studies, individual bofedales may
also be experiencing strong jumps in productivity in certain parts and decreases in others (Chávez, Christie
et al., 2019; García et al., 2019; Navarro et al., 2020). While the abrupt increase dominates the PC1-BOF
and PC1-5KM, some bofedales and grids show negative trends. Some studies have attributed localized de-
creases in productivity to mining and intensive cattle grazing (García et al., 2019), while others have found
no evidence of decreases due to livestock (Navarro et al., 2020). Additional case studies focused on specific
bofedales could be used to explore the differences between those that show overall increasing or decreasing
patterns in productivity and the influences of distinct vegetation types. Further study of these local dynam-
ics would benefit from additional field-based measurements and observations.

4.3. Implications for the Future

Tropical Andes temperatures have warmed by 0.13°C/decade from 1950–2010 with stronger warming
trends at higher elevations (Vuille et al., 2015). Concurrently, tropical glaciers have continued to retreat
(Rabatel et al., 2013). While this can lead to a temporary increase in water supply in some sectors of the
Altiplano that rely more heavily on glacial melt (Bradley et al., 2006), studies have indicated that dry season
discharge in high Andean regions of Peru has already passed this temporary threshold and shows decreas-
ing trends (Baraer et al., 2012). Because the 0°C isotherm is rising and less precipitation is expected to fall
as snow, the future of precipitation in the form of rain is critically important for these ecosystems (Barnett
et al., 2005). This is particularly true in the southern sector of the Chilean Altiplano, where snow cover is
greater. Differences between precipitation products are an additional source of uncertainty in identifying
how interannual climate variability drives changes in productivity. Here, we show that the strength and area
of significant field correlations can vary depending on the climate data set, particularly for precipitation.
The combination of complex topography and the lack of complete and widespread station data extending
beyond 1960 challenges our ability to accurately characterize current trends, to place them in a longer-term
context, and to be able to draw strong conclusions on how it does and will influence bofedal productivity.

ANDERSON ET AL. 18 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

However, assuming the lagged relationship with precipitation in the form of rain and snow we identify here
holds, it could be useful in projecting future changes in productivity in advance.

In addition to the questions surrounding current trends, a recent study evaluating mechanisms driving
Altiplano rainfall challenges assumptions of current climate projections for the region (Minvielle & Gar-
reaud, 2011; Neukom et al., 2015; Segura et al., 2020; Thibeault et al., 2012). While precipitation variability
has been linked to the strength of 200 hPa zonal winds, Segura et al. (2020) reveal that precipitation over
the last two decades appears to be more strongly related to upward motion in the western Amazon and that
the relationship with upper-level easterlies is no longer significant. Since positive precipitation anomalies
have been previously associated with strong upper-level easterlies in relation to the more southerly dis-
placement of the Bolivian High, the potential mechanistic changes suggested by Segura et al. (2020) driving
rainfall variability in the Altiplano have important implications surrounding the integrity of the network
of bofedales. Precipitation variability across interannual, decadal, multidecadal, and even longer timescales
modulates water availability across the semi-arid region, linking bofedal productivity and the future of
these ecosystems to these changes (Ledru et al., 2013; Schittek et al., 2016; Vera et al., 2019). While recent
increases in precipitation would appear to benefit this network of bofedales, Yager et al. (2019) and García
et al. (2019) show that other factors including the vegetation composition and the timing of the rainfall in
relation to extremes need to be taken into consideration.

5. Conclusion
This study evaluates peatland productivity over 30 growing seasons based on NDVI of varying resolutions
from Landsat data in the western Altiplano. While previous research has assessed these relationships in
other sectors of the Altiplano in Bolivia, Peru, and Argentina, this study is the first to characterize these
relationships across such a broad spatiotemporal scale in Chile. We find a coherent regional signal and
show that cumulative precipitation over the 2 years prior appears to be a key driver modulating productivity
across the domain. Snow cover from 2 years prior also is an important water supply for bofedales, demon-
strating the complexities associated with hydrological processes and water supply to the bofedal network
in our study region. This highlights the need to better understand groundwater fluxes, contributions and
recharge times, as well as how they affect different types of peatland vegetation across the region. We also
show that recent increases in productivity are not necessarily characterized by a long-term trend, and may
be due to a combination of high precipitation years and extreme rainfall events due to regional hydroclimat-
ic changes in our study area.

While this study focuses on climate influences, humans also play an important role in modulating this net-
work of bofedales. Indigenous communities have managed these ecosystems, particularly forage resources
like bofedales, for thousands of years (Yager et al., 2019). Linking knowledge sources to address the social
and ecological factors that influence bofedales can be useful in determining which bofedales are most im-
portant locally in terms of water resources, which are most at risk, and how this considerable carbon sink
can be most effectively managed in the face of a changing climate.

Acknowledgments
Data Availability Statement
Our research was supported by
The infilled NDVI datasets (bofedal units, 1 km2, and 5 km2) can be found at http://doi.org/10.5281/zeno-
Fondecyt grants 1201411, 11171046,
and 1211924 and ANID/FONDAP do.3966437. The CR2MET data (Boisier, 2019) are publicly available from http://www.cr2.cl/datos-produc-
grant 15110009. T.G. Anderson was tos-grillados/. The CHIRPS data (Funk et al., 2015) are publicly available from https://www.chc.ucsb.edu/
supported by a Fulbright Study and
data/chirps. Data for snow persistence are from Saavedra et al. (2018).
Research Grant in Chile. We thank Dr.
Freddy Saavedra for providing the snow
persistence data as well as Dr. Juan
Pablo Boisier for sharing the original References
CR2MET precipitation and temperature
Abraham de Vazquez, E. M., Garleff, K., Liebricht, H., Regairaz, A. C., Schäbitz, F., Squeo, F. A., et al. (2000). Geomorphology and pal-
data sets. We would also like to thank
eoecology of the arid diagonal in southern South America. Zeitschrift für Angewandte Geologie, 55–61.
the Chilean National Forest Service
Alshammari, L., Large, D. J., Boyd, D. S., Sowter, A., Anderson, R., Andersen, R., & Marsh, S. (2018). Long-term peatland condition assess-
(CONAF) for supporting field trips
ment via surface motion monitoring using the ISBAS DInSAR technique over the flow country, Scotland. Remote Sensing, 10(7), 1103.
within the National Protected Areas
https://doi.org/10.3390/rs10071103
network.

ANDERSON ET AL. 19 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

Alvarez-Garreton, C., Boisier, J. P., Garreaud, R., Seibert, J., & Vis, M. (2021). Progressive water deficits during multiyear droughts in basins
with long hydrological memory in Chile. Hydrology and Earth System Sciences, 25(1), 429–446. https://doi.org/10.5194/hess-25-429-2021
Ardila, J. P., Bijker, W., Tolpekin, V. A., & Stein, A. (2012). Context-sensitive extraction of tree crown objects in urban areas using VHR satel-
lite images. International Journal of Applied Earth Observation and Geoinformation, 15, 57–69. https://doi.org/10.1016/j.jag.2011.06.005
Baldassini, P., Volante, J. N., Califano, L. M., & Paruelo, J. M. (2012). Regional characterization of the structure and productivity of the
vegetation of the Puna using MODIS images. Ecología Austral, 22(1), 22–32.
Baraer, M., Mark, B. G., McKenzie, J. M., Condom, T., Bury, J., Huh, K.-I., et al. (2012). Glacier recession and water resources in Peru’s
Cordillera Blanca. Journal of Glaciology, 58(207), 134–150. https://doi.org/10.3189/2012jog11j186
Baraer, M., McKenzie, J. M., Mark, B. G., Bury, J., & Knox, S. (2009). Characterizing contributions of glacier melt and groundwater
during the dry season in a poorly gauged catchment of the Cordillera Blanca (Peru). Advances in Geosciences, 22, 41–49. https://doi.
org/10.5194/adgeo-22-41-2009
Barcaza, G., Nussbaumer, S. U., Tapia, G., Valdés, J., García, J.-L., Videla, Y., et al. (2017). Glacier inventory and recent glacier variations in
the Andes of Chile, South America. Annals of Glaciology, 58(75pt2), 166–180. https://doi.org/10.1017/aog.2017.28
Barnett, T. P., Adam, J. C., & Lettenmaier, D. P. (2005). Potential impacts of a warming climate on water availability in snow-dominated
regions. Nature, 438(7066), 303–309. https://doi.org/10.1038/nature04141
Bechtold, M., Schlaffer, S., Tiemeyer, B., & De Lannoy, G. (2018). Inferring water table depth dynamics from ENVISAT-ASAR c-band
backscatter over a range of peatlands from deeply-drained to natural conditions. Remote Sensing, 10(4), 536. https://doi.org/10.3390/
rs10040536
Benavides, J. (2014). The effect of drainage on organic matter accumulation and plant communities of high-altitude peatlands in the Co-
lombian tropical Andes. Mires and Peat, 15.
Boisier, J. P. (2019). CR2MET. Retrieved from http://www.cr2.cl/datos-productos-grillados/
Bradley, R. S., Vuille, M., Diaz, H. F., & Vergara, W. (2006). Threats to water supplies in the tropical Andes. Science, 312(5781), 1755–1756.
https://doi.org/10.1126/science.1128087
Bradley, R. S., Vuille, M., Hardy, D., & Thompson, L. (2003). Low latitude ice cores record Pacific Sea surface temperatures. Geophysical
Research Letters, 30(4), 1174. https://doi.org/10.1029/2002gl016546
Bridgham, S. D., Pastor, J., Dewey, B., Weltzin, J. F., & Updegraff, K. (2008). Rapid carbon response of peatlands to climate change. Ecology,
89(11), 3041–3048. https://doi.org/10.1890/08-0279.1
Burkey, J. (2006). A non-parametric monotonic trend test computing Mann-Kendall Tau, Tau-b, and Sen’s Slope written in Math-
works-MATLAB implemented using matrix rotations.
Camacho, F. M. (2012). Competing rationalities in water conflict: Mining and the indigenous community in Chiu Chiu, El Loa Province,
Northern Chile. Singapore Journal of Tropical Geography, 33(1), 93–107. https://doi.org/10.1111/j.1467-9493.2012.00451.x
Carilla, J., Halloy, S., Cuello, S., Grau, A., Malizia, A., & Cuesta, F. (2018). Vegetation trends over eleven years on mountain summits in NW
Argentina. Ecology and Evolution, 8(23), 11554–11567. https://doi.org/10.1002/ece3.4602
Casagranda, E., Navarro, C., Grau, H. R., & Izquierdo, A. E. (2019). Interannual lake fluctuations in the Argentine Puna: Relation-
ships with its associated peatlands and climate change. Regional Environmental Change, 19, 1737–1750. https://doi.org/10.1007/
s10113-019-01514-7
Charman, D. J., Amesbury, M. J., Hinchliffe, W., Hughes, P. D., Mallon, G., Blake, W. H., et al. (2015). Drivers of Holocene peatland car-
bon accumulation across a climate gradient in northeastern North America. Quaternary Science Reviews, 121, 110–119. https://doi.
org/10.1016/j.quascirev.2015.05.012
Chávez, R. O., Christie, D. A., Olea, M., & Anderson, T. G. (2019). A multiscale productivity assessment of high Andean peatlands across
the Chilean Altiplano using 31 years of Landsat imagery. Remote Sensing, 11(24), 2955. https://doi.org/10.3390/rs11242955
Chávez, R. O., Moreira-Muñoz, A., Galleguillos, M., Olea, M., Aguayo, J., Latin, A., et al. (2019). GIMMS NDVI time series reveal the extent,
duration, and intensity of “blooming desert” events in the hyper-arid Atacama Desert, Northern Chile. International Journal of Applied
Earth Observation and Geoinformation, 76, 193–203. https://doi.org/10.1016/j.jag.2018.11.013
Chimner, R. A., Bourgeau-Chavez, L., Grelik, S., Hribljan, J. A., Clarke, A. M. P., Polk, M. H., et al. (2019). Mapping mountain peatlands
and wet meadows using multi-date, multi-sensor remote sensing in the Cordillera Blanca, Peru. Wetlands, 39, 1057–1067. https://doi.
org/10.1007/s13157-019-01134-1
Cooper, D. J., Kaczynski, K., Slayback, D., & Yager, K. (2015). Growth and organic carbon production in peatlands dominated by Distichia
muscoides, Bolivia, South America. Arctic and Alpine Research, 47(3), 505–510. https://doi.org/10.1657/aaar0014-060
Cooper, D. J., Sueltenfuss, J., Oyague, E., Yager, K., Slayback, D., Caballero, E. M. C., et al. (2019). Drivers of peatland water table dynamics
in the central Andes, Bolivia and Peru. Hydrological Processes, 33(13), 1913–1925.
Cooper, D. J., Wolf, E. C., Colson, C., Vering, W., Granda, A., & Meyer, M. (2010). Alpine peatlands of the Andes, Cajamarca, Peru. Arctic
and Alpine Research, 42(1), 19–33. https://doi.org/10.1657/1938-4246-42.1.19
Dangles, O., Rabatel, A., Kraemer, M., Zeballos, G., Soruco, A., Jacobsen, D., & Anthelme, F. (2017). Ecosystem sentinels for climate
change? Evidence of wetland cover changes over the last 30 years in the tropical Andes. PloS One, 12(5), e0175814. https://doi.
org/10.1371/journal.pone.0175814
Dee, D. P., Uppala, S., Simmons, A., Berrisford, P., Poli, P., Kobayashi, S., et al. (2011). The ERA-Interim reanalysis: Configuration and
performance of the data assimilation system. Quarterly Journal of the Royal Meteorological Society, 137(656), 553–597. https://doi.
org/10.1002/qj.828
Dirección General de Aguas (DGA) (2017), Actualización del balance hídrico nacional, Tech. rep., SIT No. 417, Ministerio de Obras Públi-
cas, Dirección General de Aguas, División de Estudios y Planifcación, Santiago, Chile. Realizado por Universidad de Chile and Pontifcia
Universidad Católica de Chile.
Earle, L. R., Warner, B. G., & Aravena, R. (2003). Rapid development of an unusual peat-accumulating ecosystem in the Chilean Altiplano.
Quaternary Research, 59(1), 2–11. https://doi.org/10.1016/s0033-5894(02)00011-x
Engel, Z., Skrzypek, G., Chuman, T., Šefrna, L., & Mihaljevič, M. (2014). Climate in the Western Cordillera of the central Andes over the
last 4300 years. Quaternary Science Reviews, 99, 60–77. https://doi.org/10.1016/j.quascirev.2014.06.019
Funk, C., Peterson, P., Landsfeld, M., Pedreros, D., Verdin, J., Shukla, S., et al. (2015). The climate hazards infrared precipitation with sta-
tions—a new environmental record for monitoring extremes. Scientific Data, 2(1), 1–21. https://doi.org/10.1038/sdata.2015.66
Gandarillas, V., Jiang, Y., & Irvine, K. (2016). Assessing the services of high mountain wetlands in tropical Andes: A case study of Caripe
wetlands at Bolivian Altiplano. Ecosystem Services, 19, 51–64. https://doi.org/10.1016/j.ecoser.2016.04.006

ANDERSON ET AL. 20 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

García, C. L., Teich, I., Gonzalez-Roglich, M., Kindgard, A. F., Ravelo, A. C., & Liniger, H. (2019). Land degradation assessment in the
Argentinean Puna: Comparing expert knowledge with satellite-derived information. Environmental Science & Policy, 91, 70–80. https://
doi.org/10.1016/j.envsci.2018.10.018
Garcia, E., & Otto, M. (2015). Caracterización ecohidrológica de humedales alto andinos usando imágenes de satélite multitemporales en
la cabecera de cuenca del Río Santa, Ancash, Perú. Ecología Aplicada, 14(2), 115–125. https://doi.org/10.21704/rea.v14i1-2.88
Garreaud, R., Vuille, M., & Clement, A. C. (2003). The climate of the Altiplano: Observed current conditions and mechanisms of past
changes. Palaeogeography, Palaeoclimatology, Palaeoecology, 194(1–3), 5–22. https://doi.org/10.1016/s0031-0182(03)00269-4
Garreaud, R. D., & Aceituno, P. (2001). Interannual rainfall variability over the South American Altiplano. Journal of Climate, 14(12),
2779–2789. https://doi.org/10.1175/1520-0442(2001)014<2779:irvots>2.0.co;2
Gerber, F., de Jong, R., Schaepman, M. E., Schaepman-Strub, G., & Furrer, R. (2018). Predicting missing values in spatio-temporal remote
sensing data. IEEE Transactions on Geoscience and Remote Sensing, 56(5), 2841–2853. https://doi.org/10.1109/tgrs.2017.2785240
Glas, R., Lautz, L., McKenzie, J., Mark, B., Baraer, M., Chavez, D., & Maharaj, L. (2018). A review of the current state of knowledge of
proglacial hydrogeology in the Cordillera Blanca, Peru. Wiley Interdisciplinary Reviews: Water, 5(5), e1299. https://doi.org/10.1002/
wat2.1299
Goetz, S. (1997). Multi-sensor analysis of NDVI, surface temperature and biophysical variables at a mixed grassland site. International
Journal of Remote Sensing, 18(1), 71–94. https://doi.org/10.1080/014311697219286
Gordon, R. P., Lautz, L. K., McKenzie, J. M., Mark, B. G., Chávez, D., & Baraer, M. (2015). Sources and pathways of stream generation in trop-
ical proglacial valleys of the Cordillera Blanca, Peru. Journal of Hydrology, 522, 628–644. https://doi.org/10.1016/j.jhydrol.2015.01.013
Guttman, N. B. (1999). Accepting the standardized precipitation index: A calculation algorithm 1. JAWRA Journal of the American Water
Resources Association, 35(2), 311–322. https://doi.org/10.1111/j.1752-1688.1999.tb03592.x
Hall, D. K., Riggs, G. A., Salomonson, V. V., DiGirolamo, N. E., & Bayr, K. J. (2002). MODIS snow-cover products. Remote Sensing of Envi-
ronment, 83(1–2), 181–194. https://doi.org/10.1016/s0034-4257(02)00095-0
Hannachi, A. (2004). A primer for EOF analysis of climate data. Department of Meteorology, University of Reading (pp. 1–33).
Harris, I., & Jones, P. (2020). CRU TS4.03: Climatic Research Unit (CRU) Time-Series (TS) version 4.03 of high-resolution gridded data of
month-by-month variation in climate (Jan. 1901- Dec. 2018). Centre for Environmental Data Analysis.
Harris, I., Jones, P. D., Osborn, T. J., & Lister, D. H. (2014). Updated high-resolution grids of monthly climatic observations–the CRU TS3.
10 Dataset. International Journal of Climatology, 34(3), 623–642. https://doi.org/10.1002/joc.3711
Herrera, C., Custodio, E., Chong, G., Lambán, L. J., Riquelme, R., Wilke, H., et al. (2016). Groundwater flow in a closed basin with a saline
shallow lake in a volcanic area: Laguna Tuyajto, northern Chilean Altiplano of the Andes. The Science of the Total Environment, 541,
303–318. https://doi.org/10.1016/j.scitotenv.2015.09.060
Houston, J. (2009). A recharge model for high altitude, arid, Andean aquifers. Hydrological Processes: International Journal, 23(16), 2383–
2393. https://doi.org/10.1002/hyp.7350
Hribljan, J., Cooper, D., Sueltenfuss, J., Wolf, E., Heckman, K., Lilleskov, E., & Chimner, R. (2015). Carbon storage and long-term rate of
accumulation in high-altitude Andean peatlands of Bolivia. Mires and Peat, 15(12), 1–14.
Imfeld, N., Sedlmeier, K., Gubler, S., Correa Marrou, K., Davila, C. P., Huerta, A., et al. (2021). A combined view on precipitation and
temperature climatology and trends in the southern Andes of Peru. International Journal of Climatology, 41(1), 679–698. https://doi.
org/10.1002/joc.6645
Izquierdo, A. E., Foguet, J., & Grau, H. R. (2015). Mapping and spatial characterization of Argentine High Andean peatbogs. Wetlands
Ecology and Management, 23(5), 963–976. https://doi.org/10.1007/s11273-015-9433-3
Jackson, R. B., Lajtha, K., Crow, S. E., Hugelius, G., Kramer, M. G., & Piñeiro, G. (2017). The ecology of soil carbon: Pools, vulnera-
bilities, and biotic and abiotic controls. Annual Review of Ecology, Evolution, and Systematics, 48, 419–445. https://doi.org/10.1146/
annurev-ecolsys-112414-054234
Jackson, T. J., Chen, D., Cosh, M., Li, F., Anderson, M., Walthall, C., et al. (2004). Vegetation water content mapping using Landsat
data derived normalized difference water index for corn and soybeans. Remote Sensing of Environment, 92(4), 475–482. https://doi.
org/10.1016/j.rse.2003.10.021
Jofré, F., Espinosa, B., & Jara, J., (2020). Chile: Leader in metals that facilitate the future, Tech. rep. Chile Ministry of Mining.
Jolliffe, I. T. (2003). A cautionary note on artificial examples of EOFs. Journal of Climate, 16(7), 1084–1086. https://doi.org/10.1175/1520-
0442(2003)016<1084:acnoae>2.0.co;2
Köchy, M., Hiederer, R., & Freibauer, A. (2015). Global distribution of soil organic carbon–Part 1: Masses and frequency distributions of
SOC stocks for the tropics, permafrost regions, wetlands, and the world. Soil, 1(1), 351–365. https://doi.org/10.5194/soil-1-351-2015
Ledru, M.-P., Jomelli, V., Bremond, L., Ortuño, T., Cruz, P., Bentaleb, I., et al. (2013). Evidence of moist niches in the Bolivian Andes during
the mid-Holocene arid period. The Holocene, 23(11), 1547–1559. https://doi.org/10.1177/0959683613496288
Leifeld, J., & Menichetti, L. (2018). The underappreciated potential of peatlands in global climate change mitigation strategies. Nature
Communications, 9(1), 1071. https://doi.org/10.1038/s41467-018-03406-6
Livezey, R. E., & Chen, W. (1983). Statistical field significance and its determination by Monte Carlo techniques. Monthly Weather Review,
111(1), 46–59. https://doi.org/10.1175/1520-0493(1983)111<0046:sfsaid>2.0.co;2
McKee, T. B., Doesken, N. J., & Kleist, J. (1993). The relationship of drought frequency and duration to time scales Proceedings of the 8th
conference on Applied Climatology (Vol. 17, pp. 179–183). Boston.
McPartland, M. Y., Falkowski, M. J., Reinhardt, J. R., Kane, E. S., Kolka, R., Turetsky, M. R., et al. (2019). Characterizing boreal peat-
land plant composition and species diversity with hyperspectral remote sensing. Remote Sensing, 11(14), 1685. https://doi.org/10.3390/
rs11141685
McPartland, M. Y., Kane, E. S., Falkowski, M. J., Kolka, R., Turetsky, M. R., Palik, B., & Montgomery, R. A. (2019). The response of boreal
peatland community composition and NDVI to hydrologic change, warming, and elevated carbon dioxide. Global Change Biology, 25(1),
93–107. https://doi.org/10.1111/gcb.14465
Minería Chilena (2019).Chile aumentará más del doble su producción de litio a 2023. Retrieved from https://www.mch.cl/reportajes/
chile-aumentara-mas-del-doble-su-produccion-de-litio-a-2023/
Minvielle, M., & Garreaud, R. D. (2011). Projecting rainfall changes over the South American Altiplano. Journal of Climate, 24(17), 4577–
4583. https://doi.org/10.1175/jcli-d-11-00051.1
Moomaw, W. R., Chmura, G., Davies, G. T., Finlayson, C., Middleton, B. A., Natali, S. M., et al. (2018). Wetlands in a changing climate:
Science, policy and management. Wetlands, 38(2), 183–205. https://doi.org/10.1007/s13157-018-1023-8
Morales, M. S., Christie, D., Villalba, R., Argollo, J., Pacajes, J., Silva, J., et al. (2012). Precipitation changes in the South American Altiplano
since 1300 AD reconstructed by tree-rings. Climate of the Past, 8, 653–666. https://doi.org/10.5194/cp-8-653-2012

ANDERSON ET AL. 21 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

Moreau, S., Bosseno, R., Gu, X. F., & Baret, F. (2003). Assessing the biomass dynamics of Andean bofedal and totora high-protein wetland
grasses from NOAA/AVHRR. Remote Sensing of Environment, 85(4), 516–529. https://doi.org/10.1016/s0034-4257(03)00053-1
Mosquera, G. M., Célleri, R., Lazo, P. X., Vaché, K. B., Perakis, S. S., & Crespo, P. (2016). Combined use of isotopic and hydrometric data
to conceptualize ecohydrological processes in a high-elevation tropical ecosystem. Hydrological Processes, 30(17), 2930–2947. https://
doi.org/10.1002/hyp.10927
Navarro, C. J., Izquierdo, A. E., Aráoz, E., Foguet, J., & Grau, H. R. (2020). Rewilding of large herbivore communities in high elevation
Puna: Geographic segregation and no evidence of positive effects on peatland productivity. Regional Environmental Change, 20(4), 1–11.
https://doi.org/10.1007/s10113-020-01704-8
Neukom, R., Rohrer, M., Calanca, P., Salzmann, N., Huggel, C., Acuña, D., et al. (2015). Facing unprecedented drying of the Cen-
tral Andes? Precipitation variability over the period AD 1000–2100. Environmental Research Letters, 10(8), 084017. https://doi.
org/10.1088/1748-9326/10/8/084017
Otto, M., Scherer, D., & Richters, J. (2011). Hydrological differentiation and spatial distribution of high altitude wetlands in a semi-ar-
id Andean region derived from satellite data. Hydrology and Earth System Sciences, 15(5), 1713–1727. https://doi.org/10.5194/
hess-15-1713-2011
Pan, Y., Birdsey, R. A., Fang, J., Houghton, R., Kauppi, P. E., Kurz, W. A., et al. (2011). A large and persistent carbon sink in the world’s
forests. Science, 333(6045), 988–993. https://doi.org/10.1126/science.1201609
Pan, Z., Huang, J., Zhou, Q., Wang, L., Cheng, Y., Zhang, H., et al. (2015). Mapping crop phenology using NDVI time-series derived
from HJ-1 A/B data. International Journal of Applied Earth Observation and Geoinformation, 34, 188–197. https://doi.org/10.1016/j.
jag.2014.08.011
Pepin, N., Bradley, R. S., Diaz, H., Baraër, M., Caceres, E., Forsythe, N., et al. (2015). Elevation-dependent warming in mountain regions of
the world. Nature Climate Change, 5(5), 424–430. https://doi.org/10.1038/nclimate2563
Polk, M. H., Young, K. R., Baraer, M., Mark, B. G., McKenzie, J. M., Bury, J., & Carey, M. (2017). Exploring hydrologic connections between
tropical mountain wetlands and glacier recession in Peru’s Cordillera Blanca. Applied Geography, 78, 94–103. https://doi.org/10.1016/j.
apgeog.2016.11.004
Rabatel, A., Francou, B., Soruco, A., Gomez, J., Ceballos, J., Basantes, R., et al. (2013). Current state of glaciers in the tropical Andes: A
multi-century perspective on glacier evolution and climate change. The Cryosphere, 7, 81–102. https://doi.org/10.5194/tc-7-81-2013
Riebeek, H., & Carlowicz, M. (2011). Rare Snow in Atacama Desert, Chile. Retrieved from https://earthobservatory.nasa.gov/images/51312/
rare-snow-in-atacama-desert-chile
Riggs, G. A., Hall, D. K., & Salomonson, V. V. (2006). MODIS sea ice products user guide to collection 5 (p. 49). Greenbelt, MD: NASA God-
dard Space Flight Center.
Ruiz-Esquide, M. J. (2015). Assessing high altitude Andean wetlands using plant community structure: A multivariate analysis and remote
sensing approach. Master’s thesis. The University of British Columbia.
Russell, A. M., Gnanadesikan, A., & Zaitchik, B. (2017). Are the Central Andes mountains a warming hot spot? Journal of Climate, 30(10),
3589–3608. https://doi.org/10.1175/jcli-d-16-0268.1
Ruthsatz, B. (2012). Vegetación y ecología de los bofedales altoandinos de Bolivia. Phytocoenologia, 42(3–4), 133–179. https://doi.
org/10.1127/0340-269x/2012/0042-0535
Saavedra, F. A., Kampf, S. K., Fassnacht, S. R., & Sibold, J. S. (2017). A snow climatology of the Andes Mountains from MODIS snow cover
data. International Journal of Climatology, 37(3), 1526–1539. https://doi.org/10.1002/joc.4795
Saavedra, F. A., Kampf, S. K., Fassnacht, S. R., & Sibold, J. S. (2018). Changes in Andes snow cover from MODIS data, 2000–2016. The
Cryosphere, 12(3), 1027–1046. https://doi.org/10.5194/tc-12-1027-2018
Sala, O. E., Gherardi, L. A., Reichmann, L., Jobbagy, E., & Peters, D. (2012). Legacies of precipitation fluctuations on primary production:
Theory and data synthesis. Philosophical Transactions of the Royal Society B: Biological Sciences, 367(1606), 3135–3144. https://doi.
org/10.1098/rstb.2011.0347
Schaffer, N., MacDonell, S., Réveillet, M., Yáñez, E., & Valois, R. (2019). Rock glaciers as a water resource in a changing climate in the
semiarid Chilean Andes. Regional Environmental Change, 19, 1263–1279. https://doi.org/10.1007/s10113-018-01459-3
Schittek, K., Kock, S. T., Lücke, A., Hense, J., Ohlendorf, C., Kulemeyer, J. J., et al. (2016). A high-altitude peatland record of environmen-
tal changes in the NW Argentine Andes (24°S) over the last 2100 years. Climate of the Past, 12, 1165–1180. https://doi.org/10.5194/
cp-12-1165-2016
Segura, H., Espinoza, J. C., Junquas, C., Lebel, T., Vuille, M., & Garreaud, R. (2020). Recent changes in the precipitation-driving process-
es over the southern tropical Andes/western Amazon. Climate Dynamics, 54, 2613–2631. https://doi.org/10.1007/s00382-020-05132-6
Seimon, T. A., Seimon, A., Yager, K., Reider, K., Delgado, A., Sowell, P., et al. (2017). Long-term monitoring of tropical alpine habitat
change, Andean anurans, and chytrid fungus in the Cordillera Vilcanota, Peru: Results from a decade of study. Ecology and Evolution,
7(5), 1527–1540. https://doi.org/10.1002/ece3.2779
Sen, P. K. (1968). Estimates of the regression coefficient based on Kendall’s tau. Journal of the American Statistical Association, 63(324),
1379–1389. https://doi.org/10.1080/01621459.1968.10480934
Squeo, F., Ibacache, E., Warner, B., Espinoza, D., Aravena, R., & Gutiérrez, J. (2006). Productividad y diversidad florística de la Vega Los
Tambos, Cordillera de Doña Ana: Variabilidad interanual, herbivoría y nivel freático. In J. Cepeda (Ed.), Geoecología de la Alta Montaña
del Valle del Elqui (pp. 333–362). La Serena, Chile: Ediciones Universidad de La Serena.
Squeo, F., Warner, B., Aravena, R., & Espinoza, D. (2006). Bofedales: High altitude peatlands of the central Andes. Revista Chilena de His-
toria Natural, 79, 245–255. https://doi.org/10.4067/s0716-078x2006000200010
Theil, H. (1992). A rank-invariant method of linear and polynomial regression analysis. In Henri Theil’s contributions to economics and
econometrics (pp. 345–381). Springer. https://doi.org/10.1007/978-94-011-2546-8_20
Thibeault, J., Seth, A., & Wang, G. (2012). Mechanisms of summertime precipitation variability in the Bolivian Altiplano: Present and
future. International Journal of Climatology, 32(13), 2033–2041. https://doi.org/10.1002/joc.2424
Torres-Batlló, J., & Martí-Cardona, B. (2020). Precipitation trends over the southern Andean Altiplano from 1981 to 2018. Journal of Hy-
drology, 590(125), 485. https://doi.org/10.1016/j.jhydrol.2020.125485
Tucker, C. J. (1979). Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment, 8(2),
127–150. https://doi.org/10.1016/0034-4257(79)90013-0
Valois, R., Schaffer, N., Figueroa, R., Maldonado, A., Yáñez, E., Hevia, A., et al. (2020). Characterizing the water storage capacity and hy-
drological role of mountain peatlands in the arid Andes of North-Central Chile. Water, 12(4), 1071. https://doi.org/10.3390/w12041071

ANDERSON ET AL. 22 of 23
Journal of Geophysical Research: Biogeosciences 10.1029/2020JG005994

Valois, R., Vargas, J. A., MacDonell, S., Pinones, C. G., Fernandoy, F., Carrizo, G. Y., et al. (2021). Improving the underground structural
characterization and hydrological functioning of an Andean peatland using geoelectrics and water stable isotopes in semi-arid Chile.
Environmental Earth Sciences, 80(1), 1–14. https://doi.org/10.1007/s12665-020-09331-6
Veettil, B. K., & Kamp, U. (2017). Remote sensing of glaciers in the tropical Andes: A review. International Journal of Remote Sensing,
38(23), 7101–7137. https://doi.org/10.1080/01431161.2017.1371868
Vera, C., Higgins, W., Amador, J., Ambrizzi, T., Garreaud, R., Gochis, D., et al. (2006). Toward a unified view of the American monsoon
systems. Journal of Climate, 19(20), 4977–5000. https://doi.org/10.1175/jcli3896.1
Vera, C. S., Díaz, L. B., & Saurral, R. I. (2019). Influence of anthropogenically-forced global warming and natural climate variability in
the rainfall changes observed over the South American Altiplano. Frontiers in Environmental Science, 7, 87. https://doi.org/10.3389/
fenvs.2019.00087
Vicente-Serrano, S. M., Gouveia, C., Camarero, J. J., Beguería, S., Trigo, R., López-Moreno, J. I., et al. (2013). Response of vegetation to
drought time-scales across global land biomes. Proceedings of the National Academy of Sciences of the United States of America, 110(1),
52–57. https://doi.org/10.1073/pnas.1207068110
Vuille, M., & Ammann, C. (1997). Regional snowfall patterns in the high, arid Andes. Climatic Change, 36(3–4), 413–423. https://doi.
org/10.1023/a:1005330802974
Vuille, M., & Bradley, R. S. (2000). Mean annual temperature trends and their vertical structure in the tropical Andes. Geophysical Research
Letters, 27(23), 3885–3888. https://doi.org/10.1029/2000gl011871
Vuille, M., Bradley, R. S., Werner, M., & Keimig, F. (2003). 20th century climate change in the tropical Andes: Observations and mod-
el results. In Climate variability and change in high elevation regions: Past, present & future (pp. 75–99). Springer. https://doi.
org/10.1007/978-94-015-1252-7_5
Vuille, M., Franquist, E., Garreaud, R., Lavado Casimiro, W. S., & Cáceres, B. (2015). Impact of the global warming hiatus on Andean tem-
perature. Journal of Geophysical Research: Atmospheres, 120(9), 3745–3757. https://doi.org/10.1002/2015jd023126
Vuille, M., & Keimig, F. (2004). Interannual variability of summertime convective cloudiness and precipitation in the central Andes de-
rived from ISCCP-B3 data. Journal of Climate, 17(17), 3334–3348. https://doi.org/10.1175/1520-0442(2004)017<3334:ivoscc>2.0.co;2
Waddington, J. (2018). Survival of the thickest? peat depth as a control of peatland ecohydrological resilience to drought in EGU General
Assembly Conference Abstracts (Vol. 20, p. 2285).
Wigley, T., & Santer, B. (1990). Statistical comparison of spatial fields in model validation, perturbation, and predictability experiments.
Journal of Geophysical Research, 95(D1), 851–865. https://doi.org/10.1029/jd095id01p00851
Wunderlich, W., Oshun, J., Keating, K., Lang, M. M., Saloma, R. P., Schmidt, L., et al. (2019). Quantifying water storage capacity in and dry
season water yield from Bofedales, Andean Wetlands. American Geophysical Union Fall Meeting Abstracts, pp. A21U–2668.
Yager, K., Valdivia, C., Slayback, D., Jimenez, E., Meneses, R. I., Palabral, A., et al. (2019). Socio-ecological dimensions of Andean pastoral
landscape change: Bridging traditional ecological knowledge and satellite image analysis in Sajama National Park, Bolivia. Regional
Environmental Change, 19(5), 1353–1369. https://doi.org/10.1007/s10113-019-01466-y
Yu, Z., Loisel, J., Brosseau, D. P., Beilman, D. W., & Hunt, S. J. (2010). Global peatland dynamics since the Last Glacial Maximum. Geophys-
ical Research Letters, 37(13), L13402. https://doi.org/10.1029/2010gl043584
Zhang, X., Alexander, L., Hegerl, G. C., Jones, P., Tank, A. K., Peterson, T. C., et al. (2011). Indices for monitoring changes in extremes based
on daily temperature and precipitation data. Wiley Interdisciplinary Reviews: Climate Change, 2(6), 851–870. https://doi.org/10.1002/
wcc.147

References From the Supporting Information


Hannachi, A. (2007). Pattern hunting in climate: A new method for finding trends in gridded climate data. International Journal of Clima-
tology, 27(1), 1–15. https://doi.org/10.1002/joc.1375
Kaiser, H. F. (1958). The varimax criterion for analytic rotation in factor analysis. Psychometrika, 23(3), 187–200. https://doi.org/10.1007/
bf02289233
Monahan, A. H., Fyfe, J. C., Ambaum, M. H., Stephenson, D. B., & North, G. R. (2009). Empirical orthogonal functions: The medium is the
message. Journal of Climate, 22(24), 6501–6514. https://doi.org/10.1175/2009jcli3062.1
North, G. R., Bell, T. L., Cahalan, R. F., & Moeng, F. J. (1982). Sampling errors in the estimation of empirical orthogonal functions. Monthly
Weather Review, 110(7), 699–706. https://doi.org/10.1175/1520-0493(1982)110<0699:seiteo>2.0.co;2
Preisendorfer, R. W., & Mobley, C. D. (1988). Principal component analysis in meteorology and oceanography. Developments in Atmospheric
Science, 17.
Richman, M. B. (1986). Rotation of principal components. Journal of Climatology, 6(3), 293–335. https://doi.org/10.1002/joc.3370060305

ANDERSON ET AL. 23 of 23
View publication stats

You might also like