Move the cursor over land. Click to pin a basin.
Rainfall –
Basin area –
Rain from land –
From its own basin –
From ocean –
Closes its water balance at –
UTrack cells it spans –
Supplied one step further back –
Supplied two steps further back –
Forested and cleared
Over forest –
Over cleared land –
Difference –
Cells compared –
Supplying basins
share of this basin's total rainfall
Water risk
Water stress here –
Drought risk here –
Drought risk of the rainshed –
Driver of deforestation in the rainshed
highlight
no highlight
very land dependent (over 80% from land)
most land dependent (over 90% from land)
very ocean dependent (over 80% from ocean)
most ocean dependent (over 90% from ocean)
land dependent, most deforested rainsheds
risk exposure
off
deforested rainshed × people
deforested rainshed × cropland
forest loss in the rainshed itself
none of it 80% or more
how much of the land supplying the basin lies
in a cell that lost forest, 2001–2025
The two views above multiply this by how much of the
basin's rain comes from land at all. This is the deforestation of the
rainshed on its own, so a basin reads pale either because its
rainshed is intact or, above, because little of its rain came from
land in the first place.
more less
less more
up,
This is one dimension of water risk exposure, not
all. This is deforestation risk of land-based precipitationsheds.
For example, India reads as orange with many people and croplands and
little of its precipitationshed impacted by forest loss. However, much of the North
is severely water stressed.
dryness over cleared land
off
dry season
wet season
all year
0
Inside one basin: how thirsty the air is over the
land that lost its forest, minus the same over the land that still
has it. Matched by basin, by month and by hundred meter height band,
so those three are not carrying the difference. Soil, slope and what
is grown on the cleared ground are not matched, so this is a
comparison and not the effect of clearing.
Thirst here is vapor pressure deficit, in
kilopascals: how much more water the air could hold before it is
full. The higher it runs, the harder it pulls water out of leaves and
soil.
Clearing takes the low ground: across the basins
holding both kinds of ground, the cleared half sits a median of 52
meters below the forest, and in these basins lower ground reads drier
whatever grows on it, at 0.021 kPa per 100 meters. Matching inside
height bands takes that gradient out - 0.002 kPa per 100 meters is
left - and halves the average difference.
The average is small and the map is uneven: at
level 5 in the dry season the median basin reads +0.001 kPa and 72 of
every 100 fall within 0.05 of zero, but the 160 tropical basins
average +0.060 while the 239 boreal ones average nothing, their air
being near saturation year round. Read a quiet basin as not detected
rather than not there.
The deficit splits into a temperature half and a
moisture half, and both were pulled from the same product on the same
days to see which one moves. Four fifths of the tropical difference is
the moisture half, which is the lost transpiration. The temperature
half returns what this same method returns comparing forest with
forest, so it carries no signal. The humidity comes from a quarter
degree reanalysis, so this works between distant cells and says
nothing about a clearing smaller than that.
where to trust this map
off
water balance closure
basin against UTrack's grid
both
Rainfall accounting and limitations
Two additional limitations apply across the
dataset. First, the choice of tracking scheme moves the answer:
UTrack's own authors ran Lagrangian and Eulerian schemes on the same
moisture and found the continental recycling ratio shifting by more
than ten percentage points between them, with transport distances
differing by several degrees (Tuinenburg and Staal, 2020). How far a
basin's water came, and how much of it was continental at all, is
partly a property of the model rather than of the weather. Second,
the proportion of rainfall supplied from within a basin increases
with basin size, even without any change in atmospheric transport. On
this build self-supply scales as basin area to the power 0.8 and area
alone accounts for most of the spread, which is why the median runs
from 0.3% at level 6 to 4.7% at level 3. Self-supply compares between
basins at the same HydroBASINS level, never across them (van der Ent
and Savenije, 2011).
One independent check. Theeuwen et al. (2023)
publish a local moisture recycling ratio on this same half-degree
grid: the share of a cell's evaporation that rains out inside a
three-by-three block of cells, about 50 km across. Theirs is an
evaporation-side ratio and this map's self-supply is a
precipitation-side one, so the two should differ by roughly the land
evaporation-to-rainfall ratio, which ERA5 puts at 0.65 over this
grid. Matched for footprint size, this build comes out at 0.47 to
0.51 times theirs, and the two rank basins together at a Spearman
correlation of 0.36. The same order of magnitude and the same sign,
loose on the detail.
The hatched areas do not represent UTrack's
uncertainty. UTrack provides a single climatological estimate for
2008–2017, without uncertainty intervals. Hatching identifies
basins where tracked moisture differs substantially from recorded
rainfall or where the basin is too small relative to the 0.5°
grid to resolve its individual moisture sources. Hatching appears on
the flat map only; the alternative projections show the basin markers
without it.
how far back to follow the water
where it last evaporated
one step further back
two steps further back
both further steps
where this
basin's rain last evaporated