The energy transition doesn't add up the way the spreadsheet says

Non-additive by default

Three preprints posted October 25 and 26, from Earth and environmental archives, measure the same kind of surprise in three different systems: a power grid, a dammed river basin, and Arctic catchments. In each case, the combined or long-run effect is not what a simple sum would predict. The details differ, but the planning lesson is shared: averages and additions miss the effect, and place and timing carry much of the signal. m:…3882 m:…5712 m:…7675

The grid

The first preprint models bulk grid peak demand in the California ISO system under combined residential technology adoption. Adopting electrification, building efficiency, and solar-plus-storage together creates non-additive peak demand profiles: the triplet does not behave like the sum of its parts. Technology adoption pathways account for 89 percent of peak demand variance in the system. But within solar-plus-storage pathways specifically, climate uncertainty accounts for 23 percent of peak variance, more than the 11 percent attributed to technology adoption in those scenarios. The driver of uncertainty flips depending on which pathway you are in, which means a single planning assumption will be wrong for some plausible futures. m:…3882

Rivers and dams

The second preprint studies 404 hydroelectric plants in Yunnan, China, between 2000 and 2020. The mean habitat index across the studied plants declined from 0.48 to 0.46 over that period, a modest average that hides an era effect: plants built in 2000 show a median species decline of 0.007, while plants built after 2015 show near-zero change. Construction peaked between 2005 and 2010 with 138 plants and fell to 16 in the 2015-2020 period. The damage, in other words, is concentrated in an older build-out, and newer plants look ecologically cheaper, a distinction that matters for any country still building dams, since the average over all plants understates what the recent ones achieve and overstates what the old ones cost. m:…5712

Rivers and carbon

The third preprint tracks inorganic chemical weathering in the North American Arctic and its effect on carbon dioxide. In the Mackenzie River catchment, increases in carbonate and silicate weathering since 1980 have fully compensated for increases in sulfide oxidation. In the Peel and Yukon catchments, the balance runs the other way: enhanced sulfide oxidation has outpaced weathering, reducing carbon dioxide drawdown. The responses are non-linear, the authors report, which suggests future reductions in drawdown as permafrost degradation continues. Two neighboring river systems thus tell opposite stories about whether Arctic weathering is currently helping or hurting the carbon balance, and the difference comes down to which chemical process is winning locally. m:…7675

What planning should assume

The shared implication is to model pathways and places, not just totals. A grid plan that assumes adoption effects add up will misjudge peak demand; a dam assessment that averages across eras will misjudge new construction; a carbon budget that treats Arctic weathering as a uniform sink will misjudge individual catchments. All three results are preprints, and each is specific to its system: California's grid, Yunnan's dams, three Arctic catchments, so none of them generalizes directly. But the pattern of non-additivity showing up independently in three systems is a reason to distrust any energy or climate plan whose arithmetic is only addition. m:…3882 m:…5712 m:…7675

There is also a timing lesson. The Yunnan data show the worst ecological effects concentrated in a 2005-2010 building boom, and the Arctic data show weathering balances shifting since 1980 as permafrost degrades. Both are reminders that the system being planned for is itself moving: the baseline at the start of a project is not the baseline at the end of it. m:…5712 m:…7675