Table 1 Overview of GTA-identified threats to trees⁵ and their proxies used in this study

Unidirectional change

Crop agriculture (29%)

Cropland expansion

The cropland classification was given to grid cells with land used for annual and perennial herbaceous crops for human consumption, forage (including hay) and biofuel, excluding perennial woody crops, permanent pastures and shifting cultivation. Land abandonment was not considered.

Cropland extent gain was calculated as the sum of areas of grid cells defined as no cropland or no data that, at any time comparison, changed to an agriculture defined grid cell. Note that area already being cropland in 2000-2003 was not considered in this analysis, hence we consider 2003 as the baseline year.

GEE^11,58:

users/potapovpeter/Global_cropland

Time range: 2003–2019

Original resolution: 30 meters

Number of layers:5

Original data values:

 0 – no cropland or no data

 1 – cropland

map_post2003 = map_2004-2007 + map_2008-2011 + map_2012-2015 + map_2016-2019

Binarize map_post2003 so that cells converted at any timestep post 2003 have a value of 1

map_expansion = map_post2003 - map_2000-2003

\({{{{{\rm{Rate}}}}}}=\frac{\sum {{{{{{\rm{cell}}}}}}\; {{{{{\rm{area}}}}}}}_{{{{{{\rm{value}}}}}}==1}}{{{{{{{\rm{EOO}}}}}}}_{{{{{{\rm{area}}}}}}}}/16\)

Overexploitation (27%)

Tree cover decline

Tree cover (TC) is the percentage of horizontal ground in each 30-m pixel covered by woody vegetation greater than 5 meters in height.TC decline was calculated as the sum of areas of grid cells that have seen a decrease (minimum 5%) in TC at any time comparison. Stable or increases in tree cover were not considered. Although using this variable to quantify overexploitation or selective logging has been advised against in the past^97,98, data has improved and the use of tree cover decline will tell a more complete story to forest disturbance and degradation^99,100,101.

GEE¹²: NASA/MEASURES/GFCC/TC/v3

Time range: 2000–2015

Original resolution: 30 meters

Number of layers: 4

Original data values: TC (%)

TC₂₀₀₀ – TC₂₀₀₅

TC₂₀₀₅ – TC₂₀₁₀

TC₂₀₁₀ – TC₂₀₁₅

Combine these three difference maps using maximum, resulting in each cell that underwent a reduction in TC at any timestep having a negative value.

\({{{{{\rm{Rate}}}}}}=\frac{\sum {{{{{{\rm{cell}}}}}}\; {{{{{\rm{area}}}}}}}_{{{{{{\rm{value}}}}}} > 1}}{{{{{{{\rm{EOO}}}}}}}_{{{{{{\rm{area}}}}}}}}/15\)

Urban development (13%)

Built-up area expansion

The built-up classification was given to grid cells with any man-made land surface associated with infrastructure, commercial and residential land uses.

Built-up expansion was calculated as the sum of grid cell area defined as no built-up land in the year 2000 and changed to a built-up defined grid cell in the year 2020. Note that area already being built-up in 2000 was not considered in this analysis.

GEE¹¹:

projects/glad/GLCLU2020/Builtup_type

Time range: 2000–2020

Original resolution: 30 metersNumber of layers: 1

Original data values:

 0 – no built-up area

 1 – stable built-up area

 2 – built-up expansion

\({{{{{\rm{Rate}}}}}}=\frac{\sum {{{{{{\rm{cell}}}}}}\; {{{{{\rm{area}}}}}}}_{{{{{{\rm{value}}}}}}==2}}{{{{{{{\rm{EOO}}}}}}}_{{{{{{\rm{area}}}}}}}}/20\)

Habitat loss

Deforestation

Deforestation was defined as the complete removal of the tree canopy at the Landsat pixel scale between 2000 and 2020. The forest classification was given to grid cells with ≥ 50% of the grid cell area filled with vegetation ≥ 5 m and refers to tree cover and not land use. Deforestation was calculated as the sum of areas of grid cells defined as forest loss between the year 2000 and 2020.

GEE¹³:

UMD/hansen/global_forest_change_2021_v1_9

Time range: 2000 – 2020

Original resolution: 30 meters

Number of layers: 1

Original data values:

 0 – no forest or no forest loss

 1 – forest loss

\({{{{{\rm{Rate}}}}}}=\frac{\sum {{{{{{\rm{cell}}}}}}\; {{{{{\rm{area}}}}}}}_{{{{{{\rm{value}}}}}}==1}}{{{{{{{\rm{EOO}}}}}}}_{{{{{{\rm{area}}}}}}}}/20\)

Bidirectional change

Fire and fire suppression (13%)

Burned area

Using surface reflectance in the Near Infrared band from MODIS, burned pixels are indicated in addition to the estimate date of burn.A change in fire regime was calculated as a significant trend over time in the sum of grid cell area defined as burned in each year. Though other aspects of fire regimes are important for tree mortality, we wanted to keep the amount of detail for each threat similar and selected a variable used by other global studies³⁶. In addition, the relatively short time period is able to capture trends in burned area over time⁵⁹. Please not that this variable is dissimilar to fire return interval.

GEE¹⁵: ESA/CCI/FireCCI/5_1

Time range: 2001 – 2020

Original resolution: 250 meters

Number of layers: 20

Original data values:

  burn date (day of the year)

Per year, burned area within extent: \(\sum {{{{{{\rm{cell}}}}}}\; {{{{{\rm{area}}}}}}}_{{{{{{\rm{value}}}}}} > 0}\)

Fit a linear model of area burned over the years. If the slope is significant, the slope indicates the RRC. If the slope is not significant, the RRC is set to 0.

Climate change (4%)

Temperature and drought

Trees may not be directly affected by small differences in temperature and drought (Vapor Pressure Deficit (VPD) and precipitation), but prolonged periods or extremer extremes can impact the individual and if not the forest around the individual in turn affecting that individual. We considered minimum and maximum temperature, VPD, VPD seasonality, precipitation and precipitation seasonality.

A change in climate variable was calculated as a significant trend in yearly averages of extent over the years.

Temperature values were transformed from K*10 to °C and the precipitation unit equals mm.

Source¹⁴: Brun et al. (2022)

Time range:

Temperature 2000 – 2019

VPD 2000 – 2018

Precipitation 2000 – 2019

Original resolution: 4638.3 meters

Number of layers: 6 variables * 20 years

Original data values:

 Temperature (K*10)

 VPD (Pa)

 Precipitation accumulation (kg.m⁻²)

Per year, climate variable within extent: \(\frac{\sum {{{{{\rm{climate}}}}}}\; {{{{{\rm{variable}}}}}}}{{{{{{{\rm{EOO}}}}}}}_{{{{{{\rm{number}}}}}}\; {{{{{\rm{of}}}}}}\; {{{{{\rm{grid}}}}}}\; {{{{{\rm{cells}}}}}}}}\)

Fit a median-based linear model of yearly averaged climate variables over the years. If the slope is significant, the slope indicates the RRC. If the slope is not significant, the RRC is set to 0.

Note, the unit for this threat is not [% of extent year⁻¹] but [°C year⁻¹] [mm year⁻¹] or [PA year⁻¹]