# SeaLevel.info

Note: This page is intended to contain a complete list of all significant known or hypoth­e­sized climate feedback mechanisms. If you notice any errors or omissions, please tell me.  -DAB

## What are “feedbacks?”

In Systems Science, a “feedback” or “feedback loop” is a mechanism through which the output of a system loops around or “feeds back,” and affects an input to the same system (which, in turn, affects the output, which affects the input, etc.).

For example, when the thermostat in your house detects that the temperature is getting too cold, it turns on the furnace to raise the temperature. That's a (manmade) feedback system: The temperature causes a change in thermostat & furnace behavior, which, in turn, causes a change in temperature.

Feedback mechanisms (or simply “feedbacks,” for short) are grouped into two categories: positive & negative. That doesn't mean good vs. bad.  It means amplifying (positive) vs. attenuating/reducing/stabilizing (negative).

A positive feedback is one which causes a same-direction response, so it tends to increase (amplify) the effect of a change in input.

A common misconception is that positive feedbacks necessarily “run away,” and make a system unstable. That is incorrect. Positive feedbacks of less than 100% don't make a system unstable.

For example, consider a linear system with a gain G=1 and positive 10% (i.e. +1/10) feedback from the output to the input. An input change of 1.0 will "feed back" +10% to become, effectively, 1.1. The “.1” (additional part) is also then amplified by 10%, becoming .11, etc. The +10% feedback ends up, in the long term, asymptotically approaching 11.11111…% (i.e., +1/9 = ×109) amplification.

Similarly, a +20% (i.e. 1/5) linear feedback causes a +25% (i.e., +1/4 = ×1.25) amplification, a +33⅓% (i.e. 1/3) feedback causes a +50% (i.e. +1/2 = ×1.5) amplification, and a +50% (i.e. 1/2) feedback causes a +100% (i.e. +1 = ×2) amplification.

In general, in a linear system, a feedback ƒ causes a “compounded” net amplification (or attenuation, if ƒ is negative) which multiplies the original effect by 1/(1-ƒ).

For example, if ƒ =+20% then net amplification = 1/(1-ƒ) = 1/(1-0.2) = 1.25×.

(Caveats: In practice, delays in the feedback path may mean that the full amplification effect of a positive feedback isn't immediately seen. Also, these calculations assume linearity, but most systems are not perfectly linear, though many are approximately linear over ranges of interest.)

A negative feedback is something which causes an opposite-direction response, and thereby reduces the magnitude of the effect of the change. (Exception: if there are delays in the feedback path, very strong negative feedback can cause oscillations in the system, but that's beyond the scope of this little primer.)

The thermostat in your home is an example of a negative feedback mechanism (albeit a highly nonlinear one). It reduces the effect on indoor temperature of input changes, like changes in the weather, or someone leaving a window open.

Negative feedbacks abound in nature, especially in biological systems, such as your own body. E.g., if your body overheats, you will sweat in reaction to your elevated body temperature. Evaporation of perspiration cools your body: a negative feedback.

“Course corrections” are another example: When you are driving your car, and it drifts toward the edge of the road, in reaction to that drift you reflexively nudge the steering wheel toward the center of the road: a negative feedback.

As noted above, in general, in a linear system, a feedback ƒ causes a “compounded” net effect on the output of 1/(1-ƒ). So, for example, if ƒ =-25% then net attenuation = 1/(1-ƒ) = 1/(1+0.25) = 0.8×, i.e., a 20% reduction in the effect on the system output.

Of course, natural feedback mechanisms are rarely perfect simple, immediate, linear multipliers of a system output, though that's very often a useful approximation. Analyzing more complex systems is beyond the scope of this primer, but a few rules-of-thumb are worth noting:

• Delays in a feedback path slow the responsiveness of the system to input changes. Delays in exceptionally strong feebacks, whether positive or negative, can cause “overshoot” or oscillatory behavior, though, in practice, this is mainly observed in artificial control systems, not in natural climate systems.
• Integral feedback occurs when the effect of a feedback accumulates over time. Theoretically, positive integral feedback tends to make a system unstable. In practice, non-linearities generally limit integral feedback, causing, instead of instability, “hysteresis” in system response. That is, the system tends to “latch up” in either of two states. Ice / Albedo Feedback is an example, in which the integration function is via the accumulation of ice (ice sheet advance & retreat).
• Derivative feedback occurs when the rate of change of the output, rather than the output itself, feeds back to affect the input. Negative derivative feedback tends to stabilize a system, and prevent oscillation, and is commonly used in artificial control systems for that reason.

Feedbacks are at the center of the climate debate. The direct warming effects of anthropogenic greenhouse gas emissions are known to be small, but climate alarmists believe that those slight warming effects will be multiplied dramatically through positive feedbacks, with catastrophic consequences. I find scant evidence of that.

## Climate feedback mechanisms

• Negative (stabilizing/attenuating) climate feedbacks:
1. Planck Feedback. The most fundamental feedback effect is simply that when the Earth's surface gets warmer, it loses heat faster, thereby reducing the increase in temperature.
“It is like pumping air into a tyre with a puncture: the harder you pump the faster the air escapes.” –Clive Best
The simplest and easiest to quantify component of that effect is the radiative component, called “Planck feedback.” Radiative emissions from a warm body are proportional to the 4th power of the body's absolute temperature (temperature in Kelvin). It is calculated that a uniform global temperature increase of 1°C would increase radiant heat loss from the surface or the Earth by about 1.4% (variously estimated to be 3.2 to 3.7 W/m²).
warmer surface → more rapid radiative heat loss → cooler surface

Coincidentally, 3.7 ±0.4 W/m² is also the approximate amount of additional energy usually estimated to be retained (i.e., the “forcing”), due to a doubling of atmospheric CO2 levels. (However, that's probably too high. Atmospheric physicist Will Happer has found evidence that CO2's forcing is commonly overestimated by about 40%. If that is correct, it means CO2's forcing is only about 2.6 W/m² per doubling. Supporting Happer's calculation, the measurements of Feldman et al 2015 indicate that CO2's forcing is only 2.40 ±0.72 W/m per doubling, which is consistent with Happer's result, and inconsistent with the commonly claimed 3.7 W/m² figure.)
Aside: I've been asked why this is customarily called Planck Feedback instead of Stefan-Boltzmann Feedback. I don't know, though I have seen it called “Stefan-Boltzmann response.”
2. Convective Cooling Feedback. Convective heat loss increases with surface temperature, as well. It is roughly proportional to the difference in temperature between the surface and the air.
warmer surface → more rapid convective heat loss → cooler surface

3. Water Cycle / Evaporative Cooling Feedback. Evaporative cooling is expected to increase with higher temperatures, because warmer water evaporates faster, accelerating the Water Cycle.
warmer surface → more rapid evaporative heat loss → cooler surface

The water cycle is a classic phase-change refrigeration cycle, just like the Freon refrigeration cycle in your refrigerator: Water evaporates at the surface, absorbing “Heat Of Evaporation” (evaporative heat loss). Because the molecular weight of water vapor molecules is just 18 (compared to 28 for nitrogen), moist air is lighter than dry air (perhaps contrary to intuition). So the moist air rises to the mid-troposphere, where the water condenses into clouds, releasing the “latent heat” which it had absorbed at the surface.
This process is the most important way in which heat is removed from surface of the Earth.
Warmer temperatures should increase the rate of evaporation, and thereby increase the rate at which heat is transported away from the surface. That decreases the surface temperature changes which result from other forcings.
The importance this feedback is demonstrated by the dramatically lower diurnal temperature swings in humid climates, compared to deserts.
4. Sea-surface temperature / cloud feedback: Ramanathan & Collins 1991; Lindzen 2001; Eschenbach 2015. A number of researchers have investigated an apparent link between tropical sea surface temperatures and clouds, which seems to regulate temperatures in the tropics:
warmer surface water in tropics → changes in clouds → cooler surface water temperature

These mechanisms are controversial. The misleadingly-named “Skeptical Science” (“SkS”) climate blog catalogs the contrary arguments here and here.
(Caveat lector: I hesitated to link to those pages, because SkS is a notoriously untrustworthy web site, which cherry-picks those studies and evidence that confirm their bias, and ignores all others. I apologize for using them as a reference; I wish I had a better source.)
5. CLAW Feedback. There is evidence that increased ocean temperatures and/or sunlight increase the abundance of Pelagibacterales (“SAR11”) bacteria in the oceans, which produce dimethyl sulfide (DMS) via an intermediate compound called DMSP. DMS escapes to the atmosphere and leads to increased sulfate aerosols, which act as cloud condensation nuclei. That “seeds” clouds, increasing cloud cover, which reduces the amount of sunlight reaching the ocean and thus reduces ocean temperature, making it a negative (stabilizing) feedback mechanism.[1][2][3] (But see also Ocean Acidification / Temperature Linkage.)
warmer water and/or more sunlight → more SAR11 bacteria → DMSP → DMS → sulfate aerosols → more clouds → less sunlignt and cooler water temperature

CLAW Feedback was first hypothesized in 1987 by researchers named Charlson, Lovelock, Andreae & Warren, hence the acronym: “CLAW,” from their initials.
6. Sea Ice / Evaporation Feedback. Decreased polar sea ice coverage (Arctic & Southern Ocean) increases water evaporation, cooling the ocean by evaporative heat loss (but see also “Ice / Albedo (positive) Feedback,” below):
warmer water temp → less sea ice coverage → more evaporation → cooler water temp

Based on Nimbus-5 observations, Zwally, et al. 1983 reported that:
“...the release of heat to the atmosphere from the open water is up to 100 times greater than the heat conducted through the ice.”
It's an important effect, as the NSIDC explains:
“Sea ice regulates exchanges of heat, moisture and salinity in the polar oceans. It insulates the relatively warm ocean water from the cold polar atmosphere except where cracks, or leads, in the ice allow exchange of heat and water vapor from ocean to atmosphere in winter. The number of leads determines where and how much heat and water are lost to the atmosphere, which may affect local cloud cover and precipitation.”
...and in another article:
“Less ice also contributes to higher air temperatures by allowing transfer of heat from the relatively warmer ocean.”
(Note that the Earth's polar regions have net-negative radiation budgets. That is, they radiate more energy than they absorb from sunlight. That is always the case in Antarctica, even in summer. It is nearly always the case in the Arctic, as well, except for a brief period near the summer solstice, when the Sun is at its zenith, and solar radiation absorbed barely exceeds radiation emitted.)
The additional evaporation (due to more open water) also apparently causes additional cloud cover, increasing albedo at altitude, and probably moderating temperature changes at the surface. (See also Cloud Feedback.)
It also increases “lake/ocean-effect snowfall” (LOES) downwind. Some of that snow falls on the ice sheets and glaciers, increasing ice accumulation, and offsetting meltwater losses. Other snow falls on land, increasing albedo and snowpack, decreasing land temperatures, and prolonging winter.
Note that snow accumulation has a large effect on grounded ice mass, which in turn affects sea-level. In both Greenland and Antarctica, snowfall is the most important factor affecting ice sheet mass balance, greater in magnitude than melting, sublimation, or iceberg calving. In fact, in Antarctica, snowfall accumulation is approximately equal to the sum of those other three factors.
Multiple studies have found that snowfall accumulation in Antarctica has, indeed, been increasing.[1][2][3][4][5]
The magnitude of ice accretion from snowfall on ice sheets was illustrated by the amazing story of Glacier Girl. She's a WWII Lockheed P-38 Lightning airplane, which was extracted in pieces from beneath 268 feet(!) of accumulated ice and snow (mostly ice), fifty years after she made an emergency landing on the Greenland Ice Sheet.
That's more than 5 feet of ice per year, which is equivalent to more than seventy (70) feet of annual snowfall, which had piled up on top of the airplane! That snow represents evaporated water, mostly removed from the Arctic and North Atlantic Oceans, which then fell as ocean-effect snow on the Greenland Ice Sheet. http://wattsupwiththat.com/2012/03/29/negative-feedback-declining-sea-ice-to-lead-to-cloudier-arctic/
http://p38assn.org/glacier-girl-recovery.htm
Diagram: Glacier Girl under the ice
http://www.airspacemag.com/history-of-flight/glacier-girl-the-back-story-19218360/?all
A second airplane that landed on the Greenland Ice Sheet in 1942 has been located in 2018.
http://beta.dmi.dk/en/groenland/maalinger/greenland-ice-sheet-surface-mass-budget
Langen et al, Earth Sci., 12 January 2017.  doi:10.3389/feart.2016.00110
A Tale of Two Winters [in Greenland], The Arctic Journal, May 9, 2017.
[Note: I'm grateful to “The VooDude” (an insightful scientist who wishes to remain anonymous), from whom I learned a lot about this feedback mechanism.]
7. Sea Ice / Turbulence Feedback. Decreased polar ice cover increases water turbulence due to wind, “stirring” the water, so that surface heat loss cools the water to greater depths.
warmer seawater → less sea ice coverage → more water turbulence → faster cooling of the water → cooler seawater

8. CO2 Fertilization Feedback (“greening”). Higher CO2 levels increase plant growth rates, which reduces atmospheric CO2 levels.
higher atmospheric CO2 level → accelerated plant growth → faster removal of CO2 from the air → lower CO2 level

AR5 estimates that this effect currently removes about 29% of anthropogenic CO2 emissions from the atmosphere (Fig. 6.1), but that's a very rough estimate. (Note: on p. 6-3 they give slightly numbers: 2.5 / 9.2 = 27% went into the biosphere.) Other sources give considerably different estimates, but there's general agreement that this is an important climate feedback mechanism, apparently increasingly so (see paper and discussion).
The Earth is getting greener, and studies indicate that the primary cause is anthropogenic CO2, and the secondary cause is climate change.
However, since higher CO2 levels make plants more water efficient, the increase in green plants has not been accompanied by much increase in water usage (transpiration).
(Note: the sum of the amounts of CO2 taken up by fertilization/greening and water absorption is estimated by AR5 to be 55% of anthropogenic CO2 emissions, and that estimate is not as rough as either of the two addends.)
9. CO2 Absorption By Water Feedback. Higher atmospheric CO2 levels increase CO2 absorption by water bodies (mainly the oceans), removing CO2 from the atmosphere.
higher CO2 level → faster absorption of CO2 from atmosphere by oceans → lower CO2 level

AR5 estimates that this effect currently removes about 26% of anthropogenic CO2 emissions from the atmosphere (Fig. 6.1), but that's a very rough estimate.
(Note: Since the oceans contain about 50 times as much CO2 as the atmosphere, the absorption of atmospheric CO2 by the oceans affects the oceans much less than it affects the atmosphere.)
10. CO2 Coccolithophore Feedback. Increased CO2 levels dramatically increase growth of calcifying coccolithophores, removing carbon from the upper ocean. (This effect seems to be much greater than can presently be explained.)
Calcium carbonate (CaCO3) has density about 2.6 times that of seawater, so when coccolithophores die their exoskeletons sink. Along with other biological processes, this moves carbon (and calcium) from surface waters to the ocean depths, and it does so much more rapidly than thermohaline circulation does.  https://www.youtube.com/watch?v=ZS3MH79-AY8
doi: 10.1073/pnas.1117508109
Toggweiler, J R., 1990: Bombs and ocean carbon cycles. Nature, 347, 122-123.
11. Greenland Ice Melt / Ocean Iron Fertilization Feedback. Increased Greenland ice melt fertilizes the ocean via iron in the runoff water, increasing absorption of CO2 by photosynthesis in the oceans.
http://www.nature.com/ncomms/2014/140521/ncomms4929/full/ncomms4929.html
http://www.nature.com/ngeo/journal/v6/n4/full/ngeo1746.html
12. Lapse Rate Feedback. The “lapse rate” is the rate at which air temperature decreases with altitude within the troposphere. (It's why mountains have snowcaps.) At any given time and place, the lapse rate varies considerably, but, on average, it is about 6.5°C per km of altitude (greater for dry air, less for very humid air).
Greenhouse warming is generally expected to slightly reduce the average temperature vs. altitude lapse rate, disproportionately warming the atmosphere at higher altitudes, especially in the tropics. That should increase radiative energy losses to space, thus reducing overall warming.
(Exception: The interior of Antarctica is a special case. Above the Antarctic Plateau the lapse rate is inverted, causing a “negative greenhouse effect.” In most places, most of the time, the surface heats the air, and the surface is warmer than the air above it. So, near the surface the lapse rate is positive, meaning that as altitude increases air temperature decreases. But the Antarctic Plateau is different. It is the coldest place on Earth. The surface is so cold there, and receives and absorbs so little solar radiation, that the temperature relationship between air and surface is reversed. Near the surface, the air heats the surface, and the surface is colder than the air above it. So, near the surface the lapse rate is negative, meaning that as altitude increases air temperature also increases, causing a “negative greenhouse effect.”[1][2][3])
Physicist Clive Best discusses lapse rate feedback here, and Nick Lutsko discusses it here.
AR5 considers Lapse Rate and Water Vapor feedbacks together; see the discussion under Water Vapor Feedback
13. Rock Weathering Feedback. Atmospheric CO2 dissolves in raindrops, forming weak carbonic acid, which causes chemical weathering of wollastonite and similar silicate rocks. That chemical process removes CO2 from the rainwater, and hence from the atmosphere, and the process accelerates at warmer temperatures and higher carbonic acid levels. Higher atmospheric CO2 accelerates this process two ways: it increases carbonic acid content in rainwater, slightly lowering the water's pH, and it also causes slightly warmer temperatures through global warming.
On the face of it, this seems to be a very minor feedback, on less than multi-millennial time scales. Not only are the feedback mechanisms weak, AR5 estimates that rock weathering removes just 0.3 PgC/yr, which is only 3.4% of the estimated anthropogenic emissions of 8.9 PgC/yr. (NASA claims it's only 0.01 to 0.1 PgC/yr; nobody really knows, obviously.) However, a 2018 study (see discussion) found that Arctic warming seems to have caused a significant acceleration in this process, as indicated by increases in Arctic river carbonate alkalinity.
https://www.e-education.psu.edu/earth104/sites/www.e-education.psu.edu.earth104/files/Unit1/Mod4/L4P1-rock-weathering-3.jpg
http://www.columbia.edu/~vjd1/carbon.htm
● AR5 section 6.5.2.3 (p.550), and Fig. 6.1.
https://www.nature.com/articles/ncomms15423, and discussion here
14. Thermohaline Circulation Feedback. Thermohaline Circulation or Meridional Overturning Circulation (MOC) refers to the “Atlantic Conveyor” and related currents which carry warm surface water from the tropics toward the poles, and currents deep in the ocean which carry cold water back toward the tropics. (It's a slow process: the Atlantic Conveyor is estimated to have a cycle time of around 1000 years.) Thermohaline Circulation is driven, in part, by density differences between colder/saltier water and warmer/less-salty water.
Some scientists have speculated that global warming could reduce the density differences which drive thermohaline circulation, by reducing the temperature difference between low and high latitudes, and by adding fresh meltwater to the ocean near the poles.
It doesn't seem to be happening, so far:
Atlantic 'conveyor belt' not slowing, NASA study finds
Rossby: Gulf Stream is Not Slowing
But if thermohaline circulation were to actually slow, the slowed circulation should reduce warming at high latitudes, and thus increase the temperature difference between lower and upper latitudes, making it a negative feedback, and mitigating the effect.
15. Arctic Summer OLR Feedback. Swedish researchers Joseph Sedlar & Michael Tjernström report that they have “identified a relationship between large-scale circulation variability and changing cloud properties permitting LW [Long-Wave] radiation at both the surface and top of the atmosphere to respond to variability in atmospheric thermodynamics. Driven by anomalous advection of warm air, the corresponding enhanced OLR [Outgoing Longwave Radiation] cooling signal on monthly time scales represents an important buffer to regional Arctic warming.”
Ref: Sedlar, J., and M. Tjernström (2017). Clouds, warm air, and a climate cooling signal over the summer Arctic, Geophys. Res. Lett., 44, 10951103. doi:10.1002/2016GL071959.
(Note: This is a new addition to the sealevel.info Feedbacks page, and I've not yet read the paper. -DB)
16. Organic Aerosol Temperature Feedback. Dr. Catherine Scott of The University of Leeds, and her colleagues, conducted a model-based study of the hypothesis that, though a couple of natural mechanisms, warmer temperatures cause higher concentrations of large (>100 nm) natural organic aerosols, which reflect sunlight, cooling the surface: a negative feedback. The paper is here and (as usual) WUWT has a discussion of it.
(Note: This is a new addition to the sealevel.info Feedbacks page, and I've not yet read the paper. -DB)
17. CO2 / Evapotranspiration Linkage. Higher atmospheric CO2 levels cause plants to release less water. That improves drought resistence in crops, which is an important benefit for agriculture.
But it also reduces humidity. Like CO2, water vapor is a greenhouse gas, so reduced release of water vapor by evapotranspiration from plants should also reduce greenhouse warming, by reducing water vapor feedback.
Here's an article about it, with references to a couple of papers.
This is not a true feedback mechanism, but, like negative feedbacks, it might attenuate the warming effect of additional atmospheric CO2, so I've included it in this list for the sake of completeness.
• Positive (amplifying) climate feedbacks:
1. Water Vapor Feedback. It is generally expected that warmer temperatures should increase the amount of water vapor in the atmosphere, because warmer air holds more moisture (roughly 7% more for each 1°C of warming). This effect is usually approximated in climate calculations by assuming stable relative humidity as temperatures change. Under that assumption, warmer temperatures cause greater amounts of water vapor in the atmosphere, and since water vapor is a greenhouse gas, increased water vapor in the atmosphere should increase greenhouse warming: a positive feedback.
warmer air  → increased absolute humidity → more “greenhouse effect” → warmer surface → warmer air

This is generally believed to be the most important positive climate feedback mechanism. Quantifying it is difficult, though.
The 2015 version of the U. of Chicago's online MODTRAN interface calculated that for a Tropical Atmosphere water vapor feedback should increase the warming effect of CO2 in the tropics by only about 8% to 9%. That's probably incorrect: most other sources give much higher estimates, generally between 60% and 100% (i.e., up to doubling).
AR4 section 8.6.3.1.1 cites Forster and Collins (2004)'s estimate that water vapor feedback adds 0.9 to 2.5 W/m² of radiative forcing per 1°C of warming (best estimate 1.6 W/m²). For comparison, a doubling of CO2 is estimated to cause a radiative forcing increase of about about +3.7 W/m² (generally estimated to add a little over 1°C to average temperatures), before feedbacks (though Prof. Will Happer has found evidence that CO2's forcing is commonly overestimated by about 40%):
http://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch8s8-6-3-1.html
Such a wide range of values tells us little about the magnitidude of the amplification from water vapor feedback. E.g., if Forster & Collins' best estimate of +1.6 W/m² per °C of warming is assumed, and if it is also assumed that 2.8 W/m² causes 1°C of warming, and if other feedbacks are ignored, that would imply ƒ = 1.6/2.8 = 0.571, i.e., a very strong 57% positive feedback, which, with “compounding,” would result in a net amplification of 1/(1-ƒ) = 1/(1-(1.6/2.8)) = 2.33×, adding 133% to the original warming. But if 1°C of warming increases radiative forcing from water vapor by only 0.9 W/m², and it takes a +3.4 W/m² forcing to increase temperature by 1°C, then ƒ = 0.9/3.4 = 0.265 (26.5%), and, with compounding, the net amplification would be only 1/(1-ƒ) = 1/(1-(0.9/3.4)) = 1.36×, adding just 36% to the original warming.
Much more usefully, AR5 drops the Foster & Collins (2004) reference, and instead considers Water Vapor and Lapse Rate feedbacks together, with a much narrower estimated range (section 7.2.5, p.587) of +0.96 to +1.22 W/m² per 1°C, for the net effect of the two feedbacks, combined. If we also assume that 2.8 to 3.4 W/m² forcing causes 1°C of warming, that would imply a 0.96/3.4=28% to 1.22/2.8=43% positive net combined feedback from water vapor & lapse rate feedbacks, which, with “compounding,” would result in a net amplification of 1/(1-ƒ) = 1/(1-(0.96/3.4)) to 1/(1-(1.22/2.8)) = 1.39× to 1.77×, adding 39% to 77% (best estimate 54%) to the original warming.
The highest estimate (by far!) that I've ever seen was in a 2013 paper by Lacis & Hansen, et al, which claims (without support) that the “feedback contribution to the greenhouse effect by water vapour and clouds” effectively quadruples (adds 3× to) the warming effect of CO2 and other GHGs.
Steve Carson's “Science Of Doom” blog has fairly in-depth discussions, here:
http://scienceofdoom.com/2015/04/30/clouds-water-vapor-part-eight-clear-sky-comparison-of-models-with-erbe-and-ceres/
However, atmospheric water vapor levels do not appear to be increasing as expected:
http://wattsupwiththat.com/2012/07/18/new-paper-on-global-water-vapor-puts-climate-modelers-in-a-bind/
http://wattsupwiththat.com/2012/12/14/another-ipcc-ar5-reviewer-speaks-out-no-trend-in-global-water-vapor/
http://wattsupwiththat.com/2013/03/06/nasa-satellite-data-shows-a-decline-in-water-vapor/
http://wattsupwiththat.com/2014/07/28/new-study-claims-to-confirms-water-vapor-as-global-warming-amplifier-but-other-data-says-no/
A further complication is that, although water vapor feedback certainly amplifies temperature increases, it probably amplifies temperature increases due to rising CO2 by less than it amplifies temperature increases due to most other forcings, because of the overlap between the absorption bands of H2O vapor and CO2, on the long (>15 µm) side of CO2's LWIR absorption band.
Note that some scientists use the term “water vapor feedback” in a broader sense than I'm using it, to encompass not only the direct greenhouse warming effect of atmospheric water vapor, but also water cycle (evaporative) cooling, lapse rate cooling, and/or perhaps clouds. For example:
http://www.drroyspencer.com/2014/09/water-vapor-feedback-and-the-global-warming-pause/
http://www.drroyspencer.com/2015/05/new-satellite-upper-troposphere-product-still-no-tropical-hotspot/
http://www.drroyspencer.com/2015/08/new-evidence-regarding-tropical-water-vapor-feedback-lindzens-iris-effect-and-the-missing-hotspot/
2. Ice / Albedo Feedback. If warmer climate reduces ice and snow cover, reduced ice cover (on water) and snow cover (on land) will decrease average albedo (reflectivity), and thus increase absorption of sunlight during daytime, a positive feedback. (Ice has a low microwave albedo, but snow is a very effective insulator which also reduces heat loss at night, making ice and snow cover a negative feedback mechanism at night.)
The net ice / albedo feedback effect is thought to be modestly positive in the temperate zones, and at summer solstice in the Arctic (but see also “Sea Ice / Evaporation (negative) Feedback,” above).
However, in the context of North American and Eurasian ice sheets, this is a form of “positive integral feedback,” which means the effect accumulates/increases as the ice sheets grow or shrink, and over extremely long timespans this feedback mechanism becomes very important. It is believed to be a key factor in Milankovitch-cycle-driven glaciation/deglaciation cycles, which occur over (very roughly) 100,000 year periods.
A common misconception is that the dramatic glaciation/deglaciation cycles are evidence that the Earth's climate is very unstable, suggesting that climate sensitivity is high. They aren't. There are two apparent reasons that Milankovitch cycles have a large effect on glaciation, and neither of them implies that climate sensitivity to GHG forcings must be high.
The two apparent reasons that Milankovitch cycles have a large effect on glaciation are:
1. Milankovitch cycles change the breadth of seasonal temperature swings at the latitudes where it matters for glaciation: mainly northern North America and northern Eurasia.
When those seasonal swings increase it increases ice melt in the summers and decreases snowfall in the winters. (When temperatures are very low snowfall is greatly reduced, because the cold air cannot carry much moisture; it is said to be “too cold to snow.”) That causes ice sheets to shrink.
When the seasonal swings are reduced it decreases ice melt in the summers and increases snowfall in the winters. That causes ice sheets to grow.
2. Milankovitch cycles last for tens of thousands of years, which enables them to overcome the extreme damping of the feedback mechanism.
When, due to Milankovitch cycles, the Earth's seasonal swings are reduced to the point that the great northern ice sheets grow a little bit each year, by adding more snow in the winters than they lose in the summers, after thousands of years of ice sheet growth it adds up to a very large increase in planetary albedo. Conversely, when the Earth's seasonal swings are increased to the point that the great northern ice sheets shrink a little bit each year, as the ice sheets dwindle it eventually adds up to a large reduction in planetary albedo.
That is an important positive integral feedback mechanism, but only on timescales of thousands of years, and it is highly non-linear, because it only works when the extent of the great northern ice sheets can significantly increase or decrease.
Those ice sheets are gone, now. The only remnant is Greenland, so ice sheet albedo feedback could only become important in a warming climate if enough of the Greenland Ice Sheet were to melt to significantly increase the exposed land area, which is not in prospect.
So Milankovitch cycle-driven glaciation cycles are evidence of very large natural long-period climate variability, but they are not evidence of high climate sensitivity to anthropogenic GHG forcings.
3. CO2 / Water Temperature Feedback. The solubility of gases like CO2 (and CH4) in water decreases as the water gets warmer (per the temperature dependence of Henry's law), so as the oceans warm they outgas CO2 (or, if they're absorbing CO2, as is currently the case in most places other than the tropics, they absorb it more slowly). The CO2, in turn, works as a GHG to cause warming. (There are almost certainly also biological [2] and/or ice sheet burial mechanisms at work, which increase the magnitude of glacial-interglacial CO2 swings.)
This is a slight positive feedback mechanism.
That positive feedback loop is undoubtedly one of the causes for the apparent hysteresis [2] in the temperature and CO2 records: Over the last million years, the Earth's climate has tended to be either mild, as in our current interglacial (the Holocene), or, more of the time, heavily glaciated and cold, with relatively brief, unstable transitions between. (But see also: Deglaciation / Volcanism / CO2 Feedback.)
In paleoclimate reconstructions from ice cores, CO2 level changes generally lag temperature changes by hundreds of years, which is consistent with the fact that higher CO2 levels not only cause higher temperatures, but are also caused by higher ocean temperatures, and ocean temperature is slow to respond to air temperature changes.
4. Permafrost & Clathrate / Methane Feedback. If the climate warms, it could melt some of the Arctic permafrost, and/or underwater methane clathrate (hydrate) deposits, causing the release of methane (and some CO2) into the atmosphere. Methane (CH4) is a greenhouse gas, so this should increase warming, making it a positive feedback mechanism. However, according to the latest research, this effect is currently small, and is likely to remain so; here's a good article
5. Methane / OH-Radical Feedback. Prof. Lyatt Jaeglé of U. Washington reports that as atmospheric methane (CH4) levels increase, OH radicals are depleted, which reduces OH levels. That reduces the rate at which CH4 is removed from the atmosphere by oxidation, which increases the atmospheric lifetime of CH4, making it a positive feedback mechanism. Prof. Jaegle calculates that this feedback effectively increases the atmospheric lifetime of additional CH4 by about 50%, from about 8 years to about 12 years.
6. Sahara Dust Feedback. It has been hypothesized that dust from the Sahara could play a significant role in the Earth's cycling between glacial maxima and interglacials. Researchers have found evidence that as the Earth was warming after the last glacial maximum, much less dust was being blown from the Sahara into the atmosphere. That could have helped warm the oceans, which, in turn, may have increased monsoon rains in North Africa, helping vegetation to grow, and further reducing windblown dust.
7. Tree Line Feedback. Boreal forests are dominated by evergreens. Since they remain green even when there is snow on the ground, boreal forests have lower albedo (i.e., they are darker) than unforested ground, in snowy conditions, hence they absorb more sunlight. If a warmer climate causes tree lines to advance toward higher latitudes, this should be a positive feedback mechanism in those regions:
warmer temperature → increased forestation → lower albedo → warmer temperature
(This feedback is mentioned in the 7th paragraph of p.4 of Grantham's “Briefing paper No 12, June 2015, Biosphere feedbacks and climate change.”)
8. Deglaciation / Volcanism / CO2 Feedback. Huybers & Langmuir (2009) reported finding that deglaciations cause increased volcanic activity (perhaps due to mantle decompression), which releases CO2 to the atmosphere. That raises atmospheric CO2 levels, which causes warming, which accelerates deglaciation, making it a positive feedback loop.
They say, “Such a positive feedback may contribute to the rapid passage from glacial to interglacial periods.” (But see also: CO2 / Water Temperature Feedback.)
9. Ocean Acidification / Temperature Linkage. Higher CO2 levels in the atmosphere cause increased absorption of CO2 in the oceans, which slightly reduces seawater pH, which some researchers believe might impede the biogenic production of DMS. If so, that would reduce the production of sulfate aerosols, which, in turn, would presumably reduce cloud cover, and thereby increase ocean temperatures. (See also CLAW Feedback.)
This is not a feedback mechanism. However, if real, it would be a mechanism which increases (amplifies) the very long-term warming effect of anthropogenic CO2, so I've included it in this list, for the sake of completeness.
• Unknown-sign climate feedbacks:
1. Cloud Feedback (Overall). Clouds are the elephant in the living room. They're obviously extremely important, but they are very poorly understood. High, wispy cirrus clouds have a warming effect, because they are made of ice crystals, which makes them much more nearly opaque to outgoing longwave infrared than to incoming visible and near-IR solar radiation. Lower clouds, which are made of liquid water droplets, have a strong cooling effect in daytime, but a warming effect at night. How clouds are affected by warming or cooling climate is very complex. See: [1] [2] [3] [4] [5] [6] [7] [8] [9] [and AR5 sections 7.2.5.1 - 7.2.5.7
2. Ice Topography Feedback, a/k/a Ice-Elevation feedback. Temperatures decrease with elevation, so if an ice sheet diminishes in height, the temperature at the surface will increase, which could accelerate melting, a positive feedback. On the other hand, snowfall rates increase as temperatures approach 0°C, and lower elevations reduce gravitationally-driven ice flows and iceberg calving, both negative feedbacks. Additionally, melting at the edges of the Greenland ice sheet and ice accumulation at the center changes the topography, which might change wind patterns and hence snowfall patterns, which might change the topography, a feedback of uncertain sign:
http://blogs.plos.org/models/open-to-positive-feedback/
On centennial and shorter time scales, ice topography feedback is probably of minor importance.
3. Methane-Flora feedback. Most plants produce mostly CO2 when they decompose, but some can produce a lot of methane (CH4). An example of the latter is forage crops eaten by ruminants. Another is apparently cattails.
According to a 2018 paper, when trees rot in ponds and lakes, they produce mostly CO2, but when cattails rot in ponds and lakes they produce more methane. In either case, the carbon released represents CO2 that was removed from the atmosphere when the plant was alive, but since methane is a more potent greenhouse gas releasing more methane would have a warming effect.
The paper's authors speculated that if global warming caused an increase in cattails and decrease in trees, decaying in northern lakes, the result would be an increase is methane production, which would cause more warming: a positive feedback.
This feedback is obviously negligible (so of course BBC and dozens of other outlets hyped it as “damaging”), but it could be either net-positive-but-negligible or net-negative-but-negligible, depending on the details of the flora changes.

More resources:

Here's another (much shorter) list of climate feedbacks, with some useful discussion and references:
https://www.azimuthproject.org/azimuth/show/Climate+forcing+and+feedback#list_of_feedbacks