Falling water level drove about half the Dead Sea's deep-water warming in a GFZ-led model

Hotter air and a shrinking lake arrived over the same decades ^[2], and human diversion of incoming water contributed to the shrinking ^[3]. The temperature record alone cannot say which one warmed the water. Guillerm's group ^[1] separated them inside the model, running it once with air temperature held constant and once with lake level held constant ^[5]. By the 2010s, air warming accounted for about 1.3C of the deep-water rise in those runs ^[6]. Add the 1.5C from the falling level and the total is 2.8C, the whole observed deep warming, with the level supplying about 54% of it ^[22][23]. Single-factor runs need not sum to the full record when the factors interact. In this model they do.

I like the validation best. The model was calibrated on measurements from 2012 to 2015, four years of data, and then reproduced the warming trend across the four decades from 1979 to 2019, along with the seasonal temperature cycle ^[4][24]. Fitting a short window and matching a long record is a fair test. The thing it does not tell you is whether the split would survive a different model. The 1.5C is a counterfactual computed by this model, and no measurement of the real lake can check it directly.

Surface heat also reaches the depths. In summer, warmer and saltier water sits over cooler, slightly fresher water. Heat diffuses faster than salt, so sinking streams carry both downward in a process called salt fingering, and winter mixing gives air warming a second route ^[9]. "This part of water warming is thus due to water diversion, and adds on top of water warming due to anthropogenic air warming," Guillerm told ScienceAlert ^[8].

Projections to 2100 rest on simplifications that push in opposite directions. Leaving out inflow tends to overstate decline and warming. Leaving out industrial extraction tends to understate water loss ^[11]. Guillerm thinks the two balance, and points to the absence of any break in the simulated decline at the 2019 transition to projections ^[11]. Groundwater inflow remains uncertain ^[11]. Humidity, wind and sunlight repeat their historical seasonal patterns while air temperature is prescribed to rise ^[12]. The projected fall works out to about 1.05 metres a year, against roughly 0.85 metres a year from 1979 to 2019 ^[26][25].

This loop runs only while the lake keeps losing water. Restoring historical freshwater inflow would stop the decline, Guillerm said ^[14]. "Warming due to increasing salinity would stop," he said ^[15]. A dilute surface layer would then insulate the deeper water from atmospheric warming, though the surface itself would keep warming ^[16].

For the Mediterranean, the assumptions carry more weight. The run treats the eastern basin as completely disconnected from the ocean, and whether that happened during the Mediterranean's ancient loss of water is still debated ^[17]. It starts with the surface 400 metres below sea level and the brine already concentrated enough for rock salt to crystallise ^[18]. According to ScienceAlert's account, the same loop could explain how ancient seas laid down vast salt deposits without an exceptionally hot, dry climate around them ^[21]. Guillerm is careful about the isolation question. "Our work does not provide the solution," he said ^[20].