Where are the trees? Extent, configuration, and drivers of poor forest recovery 30 years after the 1988 Yellowstone Fires.

Postfire recovery of fire-adapted forests remains uncertain as climate and fire
regimes continue to change. Areas of poor postfire tree regeneration following
late-20th-century fires may reveal characteristics associated with increased vulnerability to
forest decline. However, sufficient time must have elapsed and pre- and postfire forest cover
must be compared to distinguish areas that have not recovered. We used remotely sensed data
and the Normalized Difference Vegetation Index (NDVI) to detect areas of poor forest recovery
across >250,000 ha of area burned as stand-replacing fire 30 years after the 1988
Yellowstone fires. We asked three questions: (1) What is the extent and configuration of
sparse and reduced forest recovery? (2) How do vegetation characteristics compare between
areas of sparse and reduced recovery vs. recovered forest? (3) What environmental
characteristics explain the distribution and patch size of sparse and reduced recovery? We
related postfire (2013-14) NDVI to field-measured stem density to establish an NDVI threshold
of sparse tree regeneration, and we contrasted pre- (1986-87) and postfire (2018-19) NDVI as a
proxy for pre- and postfire forest cover. Sparse and reduced forest recovery occupied ~41,000
ha across the burned area, about half of which was ≥150 m from ex situ seed sources. Patches
of poor recovery were generally large, with ~13,400 ha in patches ≥50 ha and an area-weighted
mean patch size of 97 ha. Vegetation was short (<2 m) in areas of sparse and reduced
recovery and non-evergreen biomass was three times greater than in recovered forest. Sparse
and reduced recovery was more likely at high elevations, on steep slopes, and far from ex situ
seed sources, and patches were larger at high elevations and far from seed sources. It took 20
years for sparse and reduced recovery to be distinguishable from recovered forest using NDVI,
suggesting a time lag before remotely sensed data can detect alternative pathways of postfire
forest recovery. Now, 30 years after the 1988 fires, ~16% of the forest burned as
stand-replacing fire has failed to recover. The extent and configuration of these areas
suggest some may persist as sparse or non-forest for the foreseeable future, with implications
for biodiversity and ecosystem processes.