Abstract

This study ties increasing climate feedbacks to projected warming consistent with temperatures when Earth last had this much CO₂ in the air. The relationship between CO₂ and temperature in a Vostok ice core is used to extrapolate temperature effects of today’s CO₂ levels. The results suggest long-run equilibrium global surface temperatures (GSTs) 5.1°C warmer than immediately “pre-industrial” (1880). The relationship derived holds well for warmer conditions 4 and 14 million years ago (Mya). Adding CH₄ data from Vostok yields 8.5°C warming due to today’s CO₂ and CH₄ levels. Long-run climate sensitivity to doubled CO₂, given Earth’s current ice state, is estimated to be 8.2°C: 1.8° directly from CO₂ and 6.4° from albedo effects. Based on the Vostok equation using CO₂ only, holding ∆GST to 2°C requires 318 ppm CO₂. This means Earth’s remaining carbon budget for +2°C is estimated to be negative 313 billion tonnes. Meeting this target will require very large-scale CO₂ removal. Lagged warming of 4.0°C (or 7.4°C when CH₄ is included), starting from today’s 1.1°C ∆GST, comes mostly from albedo changes. Their effects are estimated here for ice, snow, sulfates, and cloud cover. This study estimates magnitudes for sulfates and for future snow changes. Magnitudes for ice, cloud cover, and past snow changes are drawn from the literature. Albedo changes, plus their water vapor multiplier, caused an estimated 39% of observed GST warming over 1975-2016. Estimated warming effects on GST by water vapor; ocean heat; and net natural carbon emissions (from permafrost, etc.), all drawn from the literature, are included in projections alongside ice, snow, sulfates, and clouds. Six scenarios embody these effects. Projected ∆GSTs on land by 2400 range from 2.4 to 9.4°C. Phasing out fossil fuels by 2050 yields 7.1°C. Ending fossil fuel use immediately yields 4.9°C, similar to the 5.1°C inferred from paleoclimate studies for current CO₂ levels. Phase-out by 2050 coupled with removing 71% of CO₂ emitted to date yields 2.4°C. At the other extreme, postponing peak fossil fuel use to 2035 yields +9.4°C GST, with more warming after 2400.

Introduction

The December 2015 Paris climate pact set a target of limiting global surface temperature (GST) warming to 2°C above “pre-industrial” (1750 or 1880) levels. However, study of past climates indicates that this will not be feasible, unless greenhouse gas (GHG) levels, led by carbon dioxide (CO₂) and methane (CH₄), are reduced dramatically. Already, global air temperature at the land surface (GLST) has warmed 1.6°C since the 1880 start of NASA’s record [1]. (Temperatures in this study are 5-year moving averages from NASA, Goddard Institute for Space Studies, in °C. Baseline is 1880 unless otherwise noted.) The GST has warmed by 2.5°C per century since 2000. Meanwhile, global sea surface temperature (=(GST – 0.29 * GLST)/0.71) has warmed by 0.9°C since 1880 [2].

The paleoclimate record can inform expectations of future warming from current GHG levels. This study examines conditions during ice ages and during the most recent (warmer) epochs when GHG levels were roughly this high, some lower and some higher. It strives to connect future warming derived from paleoclimate records with physical processes, mostly from albedo changes, that produce the indicated GST and GLST values.

The Temperature Record section examines Earth’s temperature record, over eons. Paleoclimate data from a Vostok ice core covering 430,000 years (430 ky) is examined. The relations among changes in GST relative to 1880, hereafter “∆°C”, and CO₂ and CH₄ levels in this era colder than now are estimated. These relations are quite consistent with the ∆°C to CO₂ relation in eras warmer than now, 4 and 14 Mya. Overall climate sensitivity is estimated based on them. Earth’s remaining carbon budget to keep warming below 2°C is calculated next, based on the equations relating ∆°C to CO₂ and CH₄ levels in the Vostok ice core. That budget is far less than zero. It requires returning to CO₂ levels of 60 years ago.

The Feedback Pathways section discusses the major factors that lead from our present GST to the “equilibrium” GST implied by the paleoclimate data, including a case with no further human carbon emissions. This path is governed by lag effects deriving mainly from albedo changes and their feedbacks. Following an overview, eight major factors are examined and modeled to estimate warming quantities and time scales due to each. These are (1) loss of sulfates (SO₄) from ending coal use; (2) snow cover loss; (3) loss of northern and southern sea ice; (4) loss of land ice in Antarctica, Greenland and elsewhere; (5) cloud cover changes; (6) water vapor increases due to warming; (7) net emissions from permafrost and other natural carbon reservoirs; and (8) warming of the deep ocean.

Particular attention is paid to the role that anthropogenic and other sulfates have played in modulating the GST increase in the thermometer record. Loss of SO₄ and northern sea ice in the daylight season will likely be complete not long after 2050. Losses of snow cover, southern sea ice, land ice grounded below sea level, and permafrost carbon, plus warming the deep oceans, should happen later and/or more slowly. Loss of other polar land ice should happen still more slowly. But changes in cloud cover and atmospheric water vapor can provide immediate feedbacks to warming from any source.

In the Results section, these eight factors, plus anthropogenic CO₂ emissions, are modeled in six emission scenarios. The spreadsheet model has decadal resolution with no spatial resolution. It projects CO₂ levels, GSTs, and sea level rise (SLR) out to 2400. In all scenarios, GLST passes 2°C before 2040. It has already passed 1.5°. The Discussion section lays out the implications of Earth’s GST paths to 2400, implicit both in the paleoclimate data and in the development of specific feedbacks identified for quantity and time-path estimation. These, combined with a carbon emissions budget to hold GST to 2°C, highlights how crucial CO₂ removal (CDR) is. CDR is required to go beyond what emissions reduction alone can achieve. Fifteen CDR methods are enumerated. A short overview of solar radiation management follows. It may be required to supplement ending fossil fuel use and large-scale CDR.

The Temperature Record

In a first approach, temperature records from the past are examined for clues to the future. Like causes (notably CO₂ levels) should produce like effects, even when comparing eras hundreds of thousands or millions of years apart. As shown in Figure 1, Earth’s surface can grow far warmer than now, even 13°C warmer, as occurred some 50 Mya. Over the last 2 million years, with more ice, temperature swings are wider, since albedo changes – from more ice to less ice and back – are larger. For GSTs 8°C or warmer than now, ice is rare. Temperature spikes around 55 and 41 Mya show that the current one is not quite unique.

Figure 1: Temperatures and Ice Levels over 65 Million Years [3].

Some 93% of global warming goes to heat Earth’s oceans [4]. They show a strong warming trend. Ocean heat absorption has accelerated, from near zero in 1960: 4 zettaJoules (ZJ) per year from 1967 to 1990, 7 from 1991 to 2005, and 10 from 2010 to 2016 [5]. 10 ZJ corresponds to 100 years of US energy use. The oceans now gain 2/3 as much heat per year as cumulative human energy use or enough to supply US energy use for 100 years [6] or the world’s for 17 years. By 2011, Earth was absorbing 0.25% more energy than it emits, a 300 (±75) million MW heat gain [7]. Hansen deduced in 2011 that Earth’s surface must warm enough to emit another 0.6 Wm^-2 heat to balance absorption; the required warming is 0.2°C. The imbalance has probably increased since 2011 and is likely to increase further with more GHG emissions. Over the last 100 years (since 1919), GSTs have risen 1.27°C, including 1.45°C for the land surface (GLST) alone [1]. The GST warming rate from 2000 to 2020 was 0.24°C per decade, but 0.35 over the most recent decade [1,2]. At this rate, warming will exceed 2°C in 2058 for GST and in 2043 for GLST only.

Paleoclimate Analysis

Atmospheric CO₂ levels have risen 47% since 1750, including 40% since 1880 when NASA’s temperature records begin [8]. CH₄ levels have risen 114% since 1880. CO₂ levels of 415 parts per million (ppm) in 2020 are the highest since 14.1 to 14.5 Mya, when they ranged from 430 to 465 ppm [9]. The deep ocean then (over 400 ky) ranged around 5.6°C±1.0°C warmer [10] and seas were 25-40 meters higher [9]. CO₂ levels were almost as high (357 to 405 ppm) 4.0 to 4.2 Mya [11,12]. SSTs then were around 4°C±0.9°C warmer and seas were 20-35 meters higher [11,12].

The higher sea levels in these two earlier eras tell us that ice then was gone from almost all of the Greenland (GIS) and West Antarctic (WAIS) ice sheets. They hold an estimated 10 meters (7 and 3.2 modeled) of SLR between them [13,14]. Other glaciers (chiefly in Arctic islands, the Himalayas, Canada, Alaska, and Siberia) hold perhaps 25 cm of SLR [15]. Ocean thermal expansion (OTE), currently about (~) 1 mm/year [5], is another factor in SLR. This corresponds to the world ocean (to the bottom) currently warming by ~0.002°Kyr^-1. The higher sea levels 4 and 14 Mya indicate 10-30 meters of SLR that could only have come from the East Antarctic ice sheet (EAIS). This is 17-50% of the current EAIS volume. Two-thirds of the WAIS is grounded below sea level, as is 1/3 in the EAIS [16]. Those very areas (which are larger in the EAIS than the WAIS) include the part of East Antarctica most likely to be subject to ice loss over the next few centuries [17]. Sediments from millions of years ago show that the EAIS then had retreated hundreds of kilometers inland [18].

CO₂ levels now are somewhat higher than they were 4 Mya, based on the current 415 ppm. This raises the possibility that current CO₂ levels will warm Earth’s surface 4.5 to 5.0°C, best estimate 4.9°, over 1880 levels. (This is 3.4 to 3.9°C warmer than the current 1.1°C.) Consider Vostok ice core data that covers 430 ky [19]. Removing the time variable and scatter-plotting ∆°C against CO₂ levels as blue dots (the same can be done for CH₄), gives Figure 2. Its observations span the last 430 ky, at 10 ky resolution starting 10 kya.

Figure 2: Temperature to Greenhouse Gas Relationship in the Past.

Superimposed on Figure 2 are trend lines from two linear regression equations, using logarithms, for temperatures at Vostok (left-hand scale): one for CO₂ (in ppm) alone and one for both CO₂ and CH₄ (ppb). The purple trend line in Figure 2, from Equation (1) for Vostok, uses only CO₂. 95% confidence intervals in this study are shown in parentheses with ±.

(1) ∆°C = -107.1 (±17.7) + 19.1054 (±3.26) ln(CO₂).

The t-ratios are -11.21 and 11.83 for the intercept and CO₂ concentration, while R² is 0.773 and adjusted R² is 0.768. The F statistic is 139.9. All are highly significant. This corresponds to a climate sensitivity of 13.2°C at Vostok [19.1054 * ln (2)] for doubled CO₂, within the range of 180 to 465 ppm CO₂. As shown below, most of this is due to albedo changes and other amplifying feedbacks. Therefore, climate sensitivity will decline as ice and snow become scarce and Earth’s albedo stabilizes. The green trend line in Figure 2, from Equation (2) for Vostok, adds a CH₄ variable.

(2) ∆°C = -110.7 (±14.8) +11.23 (±4.55) ln(CO₂) + 7.504 (±3.48) ln(CH₄).

The t-ratios are -15.05, 4.98, and 4.36 for the intercept, CO₂, and CH₄. R² is 0.846 and adjusted R² is 0.839. The F statistic of 110.2 is highly significant. To translate temperature changes at the Vostok surface (left-hand axis) over 430 ky to changes in GST (right-hand axis), the ratio of polar change to global over the past 2 million years is used, from Snyder [20]. Snyder examined temperature data from many sedimentary sites around the world over 2 My. Her results yield a ratio for polar to global warming: 0.618. This relates the left- and right-hand scales in Figure 2. The GST equations, global instead of Vostok local, corresponding to Equations (1) and (2) for Vostok, but using the right-hand scale for global temperature, are:

(3) ∆°C = -66.19 + 11.807 ln(CO₂) and

(4) ∆°C = -68.42 + 6.94 ln(CO₂) + 4.637 ln(CH₄).

Both equations yield good fits for 14.1 to 14.5 Mya and 4.0 to 4.2 Mya. Equation 3 yields a GST climate sensitivity estimate of 8.2° (±1.4) for doubled CO₂. Table 1 below shows the corresponding GSTs for various CO₂ and CH₄ levels. CO₂ levels range from 180 ppm, the lowest recorded during the past four ice ages, to twice the immediately “pre-industrial” level of 280 ppm. Columns D, I and N add 0.13°C to their preceding columns, the difference the 1880 GST and the 1951-80 mean GST used for the ice cores. Rows are included for CO₂ levels corresponding to 1.5 and 2°C warmer than 1880, using the two equations, and for the 2020 CO₂ level of 415 ppm. The CH₄ levels (in ppb) in column F are taken from observations or extrapolated. The CH₄ levels in column K are approximations of the CH₄ levels about 1880, before human activity raised CH₄ levels much – from some mixture of fossil fuel extraction and leaks, landfills, flooded rice paddies, and large herds of cattle.

Other GHGs (e.g., N₂O and some not present in the Vostok ice cores, such as CFCs) are omitted in this discussion and in modeling future changes. Implicitly, this simplifying assumption is that the weighted rate of change of other GHGs averages the same as CO₂.

Implications

Applying Equation (3) using only CO₂, now at 415 ppm, yields a future GST 4.99°C warmer than the 1951-80 baseline. This translates to 5.12°C warmer than 1880, or 3.99°C warmer than 2018-2020 (2). This is consistent not only with the Vostok ice core records, but also with warmer Pliocene and Miocene records using ocean sediments from 4 and 14 Mya. However, when today’s CH₄ levels, ~ 1870 ppb, are used in Equation (4), indicated equilibrium GST is 8.5°C warmer than 1880. Earth’s GST is currently far from equilibrium.

Consider the levels of CO₂ and CH₄ required to meet Paris goals. To hold GST warming to 2°C requires reducing atmospheric CO₂ levels to 318 ppm, using Equation (3), as shown in Table 1. This requires CO₂ removal (CDR), at first cut, of (415-318)/(415-280) = 72% of human CO₂ emissions to date, plus any future ones. Equation (3) also indicates that holding warming to 1.5°C requires reducing CO₂ levels to 305 ppm, equivalent to 81% CDR. Using Equation (4) with pre-industrial CH₄ levels of 700 ppb, consistent with 1750, yields 2°C GST warming for CO₂ at 314 ppm and 1.5°C for 292 ppm CO₂. Human carbon emissions from fossil fuels from 1900 through 2020 were about 1600 gigatonnes (GT) of CO₂, or about 435 GT of carbon [21]. Thus, using Equation (3) yields an estimated remaining carbon budget, to hold GST warming to 2°C, of negative 313 (±54) GT of carbon, or ~72% of fossil fuel CO₂ emissions to date. This is only the minimum CDR required. First, removal of other GHGs may be required. Second, any further human emissions make the remaining carbon budget even more negative and require even more CDR. Natural carbon emissions, led by permafrost ones, will increase. Albedo feedbacks will continue, warming Earth farther. Both will require still more CDR. So, the true remaining carbon budget may actually be in the negative 400-500 GT range, and most certainly not hundreds of GT greater than zero.

Table 1: Projected Equilibrium Warming across Earth’s Surface from Vostok Ice Core Analysis (1951-80 Baseline).

The difference between current GSTs and equilibrium GSTs of 5.1 and 8.5°C stem from lag effects. The lag effects come mostly from albedo changes and their feedbacks. Most albedo changes and feedbacks happen over days to decades to centuries. Ones due to land ice and vegetation changes can continue over longer timescales. However, cloud cover and water vapor changes happen over minutes to hours. The specifics (except vegetation, not examined or modelled) are detailed in the Feedback Pathways section below.

However, the bottom two lines of Table 1 probably overestimate the temperature effects of 500 and 560 ppm of CO₂, as discussed further below. This is because albedo feedbacks from ice and snow, which in large measure underlie the derivations from the ice core, decline with higher temperatures outside the CO₂ range (180-465 ppm) used to derive and validate Equations (1) through (4).

Feedback Pathways to Warming Indicated by Paleoclimate Analysis

To hold warming to 2°C or even 1.5°, large-scale CDR is required, in addition to rapid reductions of CO₂ and CH₄ emissions to almost zero. As we consider the speed of our required response, this study examines: (1) the physical factors that account for this much warming and (2) the possible speed of the warming. As the following sections show, continued emissions speed up amplifying feedback processes, making “equilibrium” GSTs still higher. So, rapid emission reductions are the necessary foundation. But even an immediate end to human carbon emissions will be far from enough to hold warming to 2°C.

The first approach to projecting our climate future, in the Temperature Record section above, drew lessons from the past. The second approach, in the Feedback Pathways section here and below, examines the physical factors that account for the warming. Albedo effects, where Earth reflects less sunlight, will grow more important over the coming decades, in part because human emissions will decline. The albedo effects include sulfate loss from ending coal burning, plus reduced extent of snow, sea ice, land-based ice, and cloud cover. Another key factor is added water vapor, a powerful GHG, as the air heats up from albedo changes. Another factor is lagged surface warming, since the deeper ocean heats up more slowly than the surface. It will slowly release heat to the atmosphere, as El Niños do.

A second group of physical factors, more prominent late this century and beyond, are natural carbon emissions due to more warming. Unlike albedo changes, they alter CO₂ levels in the atmosphere. The most prominent is from permafrost. Other major sources are increased microbial respiration in soils currently not frozen; carbon evolved from warmer seas; release of seabed CH₄ hydrates; and any net decreased biomass in forests, oceans, and elsewhere.

This study estimates rough magnitudes and speeds of 13 factors: 9 albedo changes (including two for sea ice and four for land ice); changes in atmospheric water vapor and other ocean-warming effects; human carbon emissions; and natural emissions – from permafrost, plus a multiplier for the other natural carbon emissions. Characteristic time scales for these changes to play out range from decades for sulfates, northern and southern sea ice, human carbon emissions, and non-polar land ice; to centuries for snow, permafrost, ocean heat content, and land ice grounded below sea level; to millennia for other land ice. Cloud cover and water vapor respond in hours to days, but never disappear. The model also includes normal rock weathering, which removes about 1 GT of CO₂ per year [22], or about 3% of human emissions.

Anthropogenic sulfur loss and northern sea ice loss will be complete by 2100 and likely more than half so by 2050, depending on future coal use. Snow cover and cloud cover feedbacks, which respond quickly to temperature change, will continue. Emissions from permafrost are modeled as ramping up in an S-curve through 2300, with small amounts thereafter. Those from seabed CH₄ hydrates and other natural sources are assumed to ramp up proportionately with permafrost: jointly, by half as much. Ice loss from the GIS and WAIS grounded below sea level is expected to span many decades in the hottest scenarios, to a few centuries in the coolest ones. Partial ice loss from the EAIS, led by the 1/3 that is grounded below sea level, will happen a bit more slowly. Other polar ice loss should happen still more slowly. Warming the deep oceans, to reestablish equilibrium at the top of the atmosphere, should continue for at least a millennium, the time for a circuit of the world thermohaline ocean circulation.

This analysis and model do not include changes in (a) black carbon; (b) mean vegetation color, as albedo effects of grass replacing forests at lower latitudes may outweigh forests replacing tundra and ice at higher latitudes; (c) oceanic and atmospheric circulation; (d) anthropogenic land use; (e) Earth’s orbit and tilt; or (f) solar output.

Sulfate Effects

SO₄ in the air intercepts incoming sunlight before it arrives at Earth’s surface, both directly and indirectly via formation of cloud condensation nuclei. It then re-radiates some of that energy upward, for a net cooling effect at Earth’s surface. Mostly, sulfur impurities in coal are oxidized to SO₂ in burning. SO₂ is converted to SO₄ by chemical reactions in the troposphere. Residence times are measured in days. Including cooling from atmospheric SO₄ concentrations explains a great deal of the variation between the steady rise in CO₂ concentrations and the variability of GLST rise since 1880. Human SO₂ emissions rose from 8 Megatonnes (MT) in 1880 to 36 MT in 1920, 49 in 1940, and 91 in 1960. They peaked at 134 MT in 1973 and 1979, before falling to 103-110 during 2009-16 [23]. Corresponding estimated atmospheric SO₄ concentrations rose from 41 parts per billion (ppb) in 1880 (and a modestly lower amount before then), to 90 in 1920, 85 in 1940, and 119 in 1960, before reaching peaks of 172-178 during 1973-80 [24] and falling to 130-136 over 2009-16. Some atmospheric SO₄ is from natural sources, notably dimethyl sulfides from some ocean plankton, some 30 ppb. Volcanoes are also an important source of atmospheric sulfates, but only episodically (mean 8 ppb) and chiefly in the stratosphere (from large eruptions), with a typical residence time there of many months.

Figure 3 shows the results of a linear regression analysis, in blue, of ∆°C from the thermometer record and concentrations of CO₂, CH₄, and SO₄. SO₄ concentrations between the dates referenced above are interpolated from human emissions, added to SO₄ levels when human emissions were very small (1880). All variables shown are 5-year moving averages and SO₄ is lagged by 1 year. CO₂, CH₄, and SO₄ are measured in ppm, ppb and ppb, respectively. The near absence of an upward trend in GST from 1940 to 1975 happened at a time when human SO₂ emissions rose 170% from 1940 to 1973 [23]. This large SO₄ cooling effect offset the increased GHG warming effect, as shown in Figure 3. The analysis shown in Equation (5) excludes the years influenced by the substantial volcanic eruptions shown. It also excludes the 2 years before and 2-4 years after the years of volcanic eruptions that reached the stratosphere, since 5-year moving temperature averages are used. In particular, it excludes data from the years surrounding eruptions labeled in Figure 3, plus smaller but substantial eruptions in 1886, 1901-02, 1913, 1932-33, 1957, 1979-80, 1991 and 2011. This leaves 70 observations in all.

Figure 3: Land Surface Temperatures, Influenced by Sulfate Cooling.

Equation (5)’s predicted GLSTs are shown in blue, next to actual GLSTs in red.

(5) ∆°C = -20.48 (±1.57) + 09 (±0.65) ln(CO₂) + 1.25 (±0.33) ln(CH₄) – 0.00393 (±0.00091) SO₄

R² is 0.9835 and adjusted R² 0.9828. The F-statistic is 1,312, highly significant. T-ratios for CO₂, CH₄, and SO₄ respectively are 7.10, 7.68, and -8.68. This indicates that CO₂, CH₄, and SO₄ are all important determinants of GLSTs. The coefficient for SO₄ indicates that reducing SO₄ by 1 ppb will increase GLST by 0.00393°C. Deleting the remaining human 95 ppb of SO₄ added since 1880, as coal for power is phased out, would raise GLST by 0.37°C.

Snow

Some 99% of Earth’s snow cover, outside of Greenland and Antarctica, is in the northern hemisphere (NH). This study estimates the current albedo effect of snow cover in three steps: area, albedo effect to date, and future rate of snow shrinkage with rising temperatures. NH snow cover averages some 25 million km² annually [25,26]. 82% of month-km² coverage is during November through April. 25 million km² is 2.5 times the 10 million km² mean annual NH sea ice cover [27]. Estimated NH snow cover declined about 9%, about 2.2 million km², from 1967 to 2018 [26]. Chen et al. [28] estimated that NH snow cover decreased by 890,000 km² per decade for May to August over 1982 to 2013, but increased by 650,000 km² per decade for November to February. Annual mean snow cover fell 9% over this period, as snow cover began earlier but also ended earlier: 1.91 days per decade [28]. These changes resulted in weakened snow radiative forcing of 0.12 (±0.003) W m^-2 [28]. Chen estimated the NH snow timing feedback as 0.21 (±0.005) W m^-2 K^-1 in melting season, from 1982 to 2013 [28].

Future Snow Shrinkage

However, as GST warms further, annual mean snow cover will decline substantially with GST 5°C warmer and almost vanish with 10°. This study considers analog cities for snow cover in warmer places and analyzes data for them. It follows with three latitude and precipitation adjustments. The effects of changes in the timing of when snow is on the ground (Chen) are much smaller than from how many days snow is on the ground (see analog cities analysis, below). So, Chen’s analysis is of modest use for longer time horizons.

NH snow-covered area is not as concentrated near the pole as sea ice. Thus, sun angle leads to a larger effect by snow on Earth’s reflectivity. The mean latitude of northern snow cover, weighted over the year, is about 57°N [29], while the corresponding mean latitude of NH sea ice is 77 to 78°N. The sine of the mean sun angle (33°) on snow, 0.5454, is 2.52 times that for NH sea ice (12.5° and 0.2164). The area coverage (2.5) times the sun angle effect (2.52) suggests a cooling effect of NH snow cover (outside Greenland) about 6.3 times that for NH sea ice. [At high sun angles, water under ice is darker (~95% absorbed or 5% reflected when the sun is overhead, 0°) than rock, grass, shrubs, and trees under snow. This suggests a greater albedo contrast for losing sea ice than for losing snow. However, at the low sun angles that characterize snow latitudes, water reflects more sunlight (40% at 77° and 20% at 57°), leaving much less albedo contrast – with white snow or ice – than rocks and vegetation. So, no darkness adjustment is modeled in this study]. Using Hudson’s 2011 estimate [30] for Arctic sea ice (see below) of 0.6 W m^-2 in future radiative forcing, compared to 0.1 to date for the NH sea ice’s current cooling effect, indicates that the current cooling effect of northern snow cover is about 6.3 times 0.6 W m^-2 = 3.8 W m^-2. This is 31 times the effect of snow cover timing changes, from Chen’s analysis.

To model evolution of future snow cover as the NH warms, analog locations are used for changes in snow cover’s cooling effect as Earth’s surface warms. This cross-sectional approach uses longitudinal transects: days of snow cover at different latitudes along roughly the same longitude. For the NH, in general (especially as adjusted for altitude and distance from the ocean), temperatures increase as one proceeds southward, while annual days of snow cover decrease. Three transects in the northern US and southern Canada are especially useful, because the increases in annual precipitation with warmer January temperatures somewhat approximate the 7% more water vapor in the air per 1°C of warming (see “In the Air” section for water vapor). The transects shown in Table 2 are (1) Winnipeg, Fargo, Sioux Falls, Omaha, Kansas City; (2) Toronto, Buffalo, Pittsburgh, Charleston WV, Knoxville; and (3) Lansing, Detroit, Cincinnati, Nashville. Pooled data from these 3 transects, shown at the bottom of Table 2, indicate 61% as many days as now with snow cover ≥ 1 inch [31] with 3°C local warming, 42% with 5°C, and 24% with 7°C. However, these degrees of local warming correspond to less GST warming, since Earth’s land surface has warmed faster than the sea surface and observed warming is generally greater as one proceeds from the equator toward the poles; [1,2,32] the gradient is 1.5 times the global mean for 44-64°N and 2.0 times for 64-90°N [32]. These latitude adjustments for local to global warming pair 61% as many snow cover days with 2°C GLST warming, 42% with 3°C, and 24% with 4°C. This translates to approximately a 19% decrease in days of snow cover per 1°C warming.

Table 2: Snow Cover Days for Transects with ~7% More Precipitation per °C. Annual Mean # of Days with ≥ 1 inch of Snow on Ground.

This study makes three adjustments to the 19%. First, the three transects feature precipitation increasing only 4.43% (1.58°C) per 1°C warming. This is 63% of the 7% increase in global precipitation per 1°C warming. So, warming may bring more snowfall than the analogs indicate directly. Therefore the 19% decrease in days of snow cover per 1°C warming of GLST is multiplied by 63%, for a preliminary 12% decrease in global snow cover for each 1°C GLST warming. Second, transects (4) Edmonton to Albuquerque and (5) Quebec to Wilmington NC, not shown, lack clear precipitation increases with warming. But they yield similar 62%, 42%, and 26% as many days of snow cover for 2, 3, and 4°C increases in GST. Since the global mean latitude of NH snow cover is about 57°, the southern Canada figure should be more globally representative than the 19% figure derived from the more southern US analysis. Use of Canadian cities only (Edmonton, Calgary, Winnipeg, Sault Ste. Marie, Toronto, and Quebec, with mean latitude 48.6°N) yields 73%, 58%, and 41% of current snow cover with roughly 2, 3, and 4°C warming. This translates to a 15% decrease in days of snow cover in southern Canada per 1°C warming of GLST. 63% of this, for the precipitation adjustment, yields 9.5% fewer days of snow cover per 1°C warming of GLST. Third, the southern Canada (48.6°N) figure of 9.5% warrants a further adjustment to represent an average Canadian and snow latitude (57°N). Multiplying by sin(48.6°)/sin(57°) yields 8.5%. The story is likely similar in Siberia, Russia, north China, and Scandinavia. So, final modeled snow cover decreases by 8.5% (not 19, 12 or 9.5%) of current amounts for each 1°C rise in GLST. In this way, modeled snow cover vanishes completely at 11.8°C warmer than 1880, similar to the Paleocene-Eocene Thermal Maximum (PETM) GSTs 55 Mya [3].

Ice

Six ice albedo changes are calculated separately: for NH and Antarctic (SH) sea ice, and for land ice in the GIS, WAIS, EAIS, and elsewhere (e.g., Himalayas). Ice loss in the latter four leads to SLR. This study considers each in turn.

Sea Ice

Arctic sea ice area has shown a shrinking trend since satellite coverage began in 1979. Annual minimum ice area fell 53% over the most recent 37 years [33]. However, annual minimum ice volume shrank faster, as the ice also thinned. Estimated annual minimum ice volume fell 73% over the same 37 years, including 51% in the most recent 10 years [34]. Trends in Arctic sea ice volume [34] are shown in Figure 4, with their corresponding R², for four months. One set of trend lines (small dots) is based on data since 1980, while a second, steeper set (large dots) uses data since 2000. (Only four months are shown, since July ice volume is like November’s and June ice volume is like January’s). The graph suggests sea ice will vanish from the Arctic from June through December by 2050. Moreover, NH sea ice may vanish totally by 2085 in April, the minimum ice volume month. That is, current volume trends yield an ice-free Arctic Ocean about 2085.

Figure 4: Arctic Sea Ice Volume by Month and Year, Past and Future.

Hudson estimated that loss of Arctic sea ice would increase radiative forcing in the Arctic by an amount equivalent to 0.7 W m^-2, spread over the entire planet, of which 0.1 W m^-2 had already occurred [30]. That leaves 0.6 W m^-2 of radiative forcing still to come, as of 2011. This translates to 0.31°C warming yet to come (as of 2011) from NH sea ice loss. Trends in Antarctic sea ice are unclear. After three record high winter sea ice years in 2013-15, record low Antarctic sea ice was recorded in 2017-19 and 2020 is below average [27]. If GSTs rise enough, eventually Antarctic land ice and sea ice areas should shrink. Roughly 2/3 of Antarctic sea ice is associated with West Antarctica [35]. Therefore, 2/3 of modeled SH sea ice loss corresponds to WAIS ice volume loss and 1/3 to EAIS. However, to estimate sea ice area, change in estimated ice volume is raised to the 1.5 power (using the ratio of 3 dimensions of volume to 2 of area). This recognizes that sea ice area will diminish more quickly than the adjacent land ice volume of the far thicker WAIS (including the Antarctic Peninsula) and the EAIS.

Land Ice

Paleoclimate studies have estimated that global sea levels were 20 to 35 meters higher than today from 4.0 to 4.2 Mya [13,14]. This indicates that a large fraction of Earth’s polar ice had vanished then. Earth’s GST then was estimated to be 3.3 to 5.0°C above the 1951-80 mean, for CO₂ levels of 357-405 ppm. Another study estimated that global sea levels were 25-40 meters higher than today’s from 14.1 to 14.5 Mya [11]. This suggests 5 meters more of SLR from vanished polar ice. The deep ocean then was estimated to be 5.6±1.0°C warmer than in 1951-80, in response to still higher CO₂ levels of 430-465 ppm CO₂ [11,12]. Analysis of sediment cores by Cook [20] shows that East Antarctic ice retreated hundreds of kilometers inland in that time period. Together, these data indicate large polar ice volume losses and SLR in response to temperatures expected before 2400. This tells us about total amounts, but not about rates of ice loss.

This study estimates the albedo effect of Antarctic ice loss as follows. The area covered by Antarctic land ice is 1.4 times the annual mean area covered by NH sea ice: 1.15 for the EAIS and 0.25 for the WAIS. The mean latitudes are not very different. Thus, the effect of total Antarctic land ice area loss on Earth’s albedo should be about 1.4 times that 0.7 Wm^-2 calculated by Hudson for NH sea ice, or about 1.0 Wm^-2. The model partitions this into 0.82 Wm^-2 for the EAIS and 0.18 Wm^-2 for the WAIS. Modeled ice mass loss proceeds more quickly (in % and GT) for the WAIS than for the EAIS. Shepherd et al. [36] calculated that Antarctica’s net ice volume loss rate almost doubled, from the period centered on 1996 to that on 2007. That came from the WAIS, with a compound ice mass loss of 12% per year from 1996 to 2007, as ice volume was estimated to grow slightly in the EAIS [36,37] over this period. From 1997 to 2012, Antarctic land ice loss tripled [36]. Since then, Antarctic land ice loss has continued to increase by a compound rate of 12% per year [37]. This study models Antarctic land ice losses over time using S-curves. The curve for the WAIS starts rising at 12% per year, consistent with the rate observed over the past 15 years, starting from 0.4 mm per year in 2010, and peaks in the 2100s. Except in CDR scenarios, remaining WAIS ice is negligible by 2400. Modeled EAIS ice loss increases from a base of 0.002 mm per year in 2010. It is under 0.1% in all scenarios until after 2100, peaks from 2145 to 2365 depending on scenario, and remains under 10% by 2400 in the three slowest-warming scenarios.

The GIS area is 17.4% of the annual average NH sea ice coverage [27,38], but Greenland experiences (on average) a higher sun angle than the Arctic Ocean. This suggests that total GIS ice loss could have an albedo effect of 0.174 * cos (72°)/cos (77.5°) = 0.248 times that of total NH sea ice loss. This is the initial albedo ratio in the model. The modeled GIS ice mass loss rate decreases from 12% per year too, based on Shepherd’s GIS findings for 1996 to 2017 [37]. Robinson’s [39] analysis indicated that the GIS cannot be sustained at temperatures warmer than 1.6°C above baseline. That threshold has already been exceeded locally for Greenland. So it is reasonable to expect near total ice loss in the GIS if temperatures stay high enough for long enough. Modeled GIS ice loss peaks in the 2100s. It exceeds 80% by 2400 in scenarios lacking CDR and is near total by then if fossil fuel use continues past 2050.

The albedo effects of land ice loss, as for Antarctic sea ice, are modeled as proportional to the 1.5 power of ice loss volume. This assumes that the relative area suffering ice loss will be more around the thin edges than where the ice is thickest, far from the edges. That is, modeled ice-coved area declines faster than ice volume for the GIS, WAIS, and EAIS. Ice loss from other glaciers, chiefly in Arctic islands, Canada, Alaska, Russia, and the Himalayas, is also modeled by S-curves. Modeled “other glaciers” ice volume loss in the 6 scenarios ranges from almost half to almost total, depending on the scenario. Corresponding SLR rise by 2400 ranges from 12 to 25 cm, 89% or more of it by 2100.

In the Air: Clouds and Water Vapor

As calculated by Equation (5), using 70 years without significant volcanic eruptions, GLST will rise about 0.37°C as human sulfur emissions are phased out. Clouds cover roughly half of Earth’s surface and reflect about 20% [40] of incoming solar radiation (341 W m^–2 mean for Earth’s surface). This yields mean reflection of about 68 W m^–2, or 20 times the combined warming effect of GHGs [41]. Thus, small changes in cloud cover can have large effects. Detecting cloud cover trends is difficult, so the error bar around estimates for forcing from cloud cover changes is large: 0.6±0.8 Wm^–2K^–1 [42]. This includes zero as a possibility. Nevertheless, the estimated cloud feedback is “likely positive”. Zelinka [42] estimates the total cloud effect at 0.46 (±0.26) W m^–2K ^–1. This comprises 0.33 for less cloud cover area, 0.20 from more high-altitude ones and fewer low-altitude ones, -0.09 for increased opacity (thicker or darker clouds with warming), and 0.02 for other factors. His overall cloud feedback estimate is used for modeling the 6 scenarios shown in the Results section. This cloud effect applies both to albedo changes from less ice and snow and to relative changes in GHG (CO₂) concentrations. It is already implicit in estimates for SO₄ effects. 1°C warmer air contains 7% more water vapor, on average [43]. That increases radiative forcing by 1.5 W m^–2 [43]. This feedback is 89% as much as from CO₂ emitted from 1750 to 2011 [41]. Water vapor acts as a warming multiplier, whether from human GHG emissions, natural emissions, or albedo changes. The model treats water vapor and cloud feedbacks as multipliers. This is also done in Table 3 below.

Table 3: Observed GST Warming from Albedo Changes, 1975-2016.

Albedo Feedback Warming, 1975-2016, Informs Climate Sensitivities

Amplifying feedbacks, from albedo changes and natural carbon emissions, are more prominent in future warming than direct GHG effects. Albedo feedbacks to date, summarized in Table 3, produced an estimated 39% of GST warming from 1975 to 2016. This came chiefly from SO₄ reductions, plus some from snow cover changes and Arctic sea ice loss, with their multipliers from added water vapor and cloud cover changes. On the top line of Table 3 below, the SO₄ decrease, from 177.3 ppb in 1975 to 130.1 in 2016, is multiplied by 0.00393°C/ppb SO₄ from Equation (5). On the second line, in the second column, Arctic sea ice loss is from Hudson [30], updated from 0.10 to 0.11 W m^–2 to cover NH sea ice loss from 2010 to 2016. The snow cover timing change effect of 0.12 W m^–2 over 1982-2013 is from Chen [28]. But the snow cover data is adjusted to 1975-2016, for another 0.08 W m^-2 in snow timing forcing, using Chen’s formula for W m^-2 per °C warming [28] and extra 0.36°C warming over 1975-82 plus 2013-16. The amount of the land ice area loss effect is based on SLR to date from the GIS, WAIS, and non-polar glaciers. It corresponds to about 10,000 km², less than 0.1% of the land ice area.

For the third column of Table 3, cloud feedback is taken from Zelinka [42] as 0.46 W m^–2K^–1. Water-vapor feedback is taken from Wadhams [43], as 1.5 W m^–2K^–1. The combined cloud and water-vapor feedback of 1.96 W m^–2K^–1 modeled here amounts to 68.8% of the 2.85 total forcing from GHGs as of 2011 [41]. Multiplying column 2 by 68.8% yields the numbers in column 3. Conversion to ∆°C in column 4 divides the 0.774°C warming from 1880 to 2011 [2] by the total forcing of 2.85 W m^-2 from 1880 to 2011 [41]. This yields a conversion factor of 0.2716°C W^-1m², applied to the sum of columns 2 and 3, to calculate column 4. Error bars are shown in column 5. In summary, estimated GST warming over 1975-2016 from albedo changes, both direct (from sulfate, ice, and snow changes) and indirect (from cloud and water-vapor changes due to direct ones), totals 0.330°C. Total GST warming then was 0.839°C [2]. (This is more than the 0.774°C (2) warming from 1880 to 2011, because the increase from 2011 to 2016 was greater than the increase from 1880 to 1975.) So, the ∆GST estimated for albedo changes over 1975-2016, direct and indirect, comes to 0.330/0.839 = 39.3% of the observed warming.

1975-2016 Warming Not from Albedo Effects

The remaining 0.509°C warming over 1975-2016 corresponds to an atmospheric CO₂ increase from 331 to 404 ppm [44], or 22%. This 0.509°C warming is attributed in the model to CO₂, consistent with Equations (3) and (1), using the simplification that the sum total effect of other GHGs changes as the same rate as for CO₂. It includes feedbacks from H₂O vapor and cloud cover changes, estimated, per above, as 0.686/(1+1.686) of 0.509°C, which is 0.207°C or 24.7% of the total 0.839°C warming over 1975-2016. This leaves 0.302°C warming for the estimated direct effect of CO₂ and other factors, including other GHGs and factors not modeled, such as black carbon and vegetation changes, over this period.

Partitioning Climate Sensitivity

With the 22% increase in CO₂ over 1975-2016, we can estimate the change due to a doubling of CO₂ by noting that 1.22 [= 404/331] raised to the power 3.5 yields 2.0. This suggests that a doubling of CO₂ levels – apart from surface albedo changes and their feedbacks – leads to about 3.5 times 0.509°C = 1.78°C of warming due to CO₂ (and other GHGs and other factors, with their H₂O and cloud feedbacks), starting from a range of 331-404 ppm CO₂. In the model, for projected temperature changes for a particular year, 0.509°C is multiplied by the natural logarithm of (the CO₂ concentration/331 ppm in 1975) and divided by the natural logarithm of (404 ppm/331 ppm), that is divided by 0.1993. This yields estimated warming due to CO₂ (plus, implicitly, other non-H₂O GHGs) in any particular year, again apart from surface albedo changes and their feedbacks, including the factors noted that are not modelled in this study.

Using Equation (3), warming associated with doubled CO₂ over the past 14.5 million years is 11.807 x ln(2.00), or 8.184°C per CO₂ doubling. The difference between 8.18°C and 1.78°C, from CO₂ and non-H₂O GHGs, is 6.40°C. This 6.40°C climate sensitivity includes the effect of albedo changes and the consequent H₂O vapor concentration. Loss of tropospheric SO₄ and Arctic sea ice are the first of these to occur, with immediate water vapor and cloud feedbacks. Loss of snow and Antarctic sea ice follow over centuries to decades. Loss of much land ice, especially where grounded above sea level, happens more slowly.

Stated another way, there are two climate sensitivities: one for the direct effect of GHGs and one for amplifying feedbacks, led by albedo changes. The first is estimated as 1.8°C. The second is estimated as 6.4°C in epochs, like ours, when snow and ice are abundant. In periods with little or no ice and snow, this latter sensitivity shrinks to near zero, except for clouds. As a result, climate is much more stable to perturbations (notably cyclic changes in Earth’s tilt and orbit) when there is little snow or ice. However, climate is subject to wide temperature swings when there is lots of snow and ice (notably the past 2 million years, as seen in Figure 1).

In the Oceans

Ocean Heat Gain: In 2011, Hansen [7] estimated that Earth is absorbing 0.65 Wm^-2 more than it emits. As noted above, ocean heat gain averaged 4 ZJ per year over 1967 to 1990, 7 over 1991-2005, and 10 over 2006-16. Ocean heat gain accelerated while GSTs increased. Therefore, ocean heat gain and Earth’s energy imbalance seem likely to continue rising as GSTs increase. This study models the situation that way. Oceans would need to warm up enough to regain thermal equilibrium with the air above. While oceans are gaining heat (now ~ 2 times cumulative human energy use every 3 years), they are out of equilibrium. The ocean thermohaline circuit takes about 1,000 years. So, if human GHG emissions ended today, this study assumes that it could take Earth’s oceans 1,000 years to thermally re-equilibrate heat with the atmosphere. The model spreads the bulk of that over 400 years, in an exponential decay shape. The rate peaks during 2130 to 2170, depending on the scenario. The modeled effect is about 5% of total GST warming. Ocean thermal expansion (OTE), currently about 0.8 mm/year [5], is another factor in SLR. Changes to its future values are modeled as proportional to future temperature change.

Land Ice Mass Loss, Its Albedo Effect, and Sea Level Rise: Modeled SLR derives mostly from modeled ice sheet losses. Their S-curves were introduced above. The amount and rate parameters are informed by past SLR. Sea levels have varied by almost 200 meters over the past 65 My. They were almost 125 meters lower than now during recent Ice Ages [3]. SLR reached some 70 meters higher in ice-free warm periods more than 10 Mya, especially more than 35 Mya [3]. From Figure 1, Earth was largely ice-free when deep ocean temperature (DOT) was 7°C or more, for SLR of about 73 meters from current levels, when DOT is < 2°C. This yields a SLR estimate of 15 meters/°C of DOT in warm eras. Over the most recent 110-120 ky, 110 meters of SLR is associated with 4 to 6°C GST warming (Figure 2), or 19-28 meters/°C GST in a cold era. The 15:28 warm/cold era ratio for SLR rate shows that the amount of remaining ice is a key SLR variable. However, this study projects only 1.5 to 4 meters rate of SLR by 2400 per °C of GST warming, but still rising. The WAIS and GIS together hold 10-12 meters of SLR [15,16]. So, 25-40 meter SLR during 14.1-14.5 Mya suggests that the EAIS lost about 1/3 to 1/2 of its current ice volume (20 to 30 meters of SLR, out of almost 60 today in the EAIS [45]) when CO₂ levels were last at 430-465 ppm and DOTs were 5.6±1.0°C [11,12]. This is consistent with this study’s two scenarios with human CO₂ emissions after 2050 and even 2100: 13 and 21 meters of SLR from the EAIS by 2400, with Δ GLSTs of 8.2 and 9.4°C. DeConto [19] suggested that sections of the EAIS grounded below sea level would lose all ice if we continue emissions at the current rate, for 13.6 or even 15 meters of SLR by 2500. This model’s two scenarios with intermediate GLST rise yield SLR closest to his projections. SLR is even higher in the two warmest scenarios. Modeled SLR rates are informed by the most recent 19,000 years of data ([46,47], chart by Robert A. Rohde). They include a SLR rate of 3 meters/century during Meltwater Pulse 1A for 8 centuries around 14 ky ago. They also include 1.5 meters/century over the 70 centuries from 15 kya to 8 kya. The DOT rose 3.3°C over 10,000 years, for an average rate of 0.033°C per century. However, the current SST warming rate is 2.0°C per century [1,2], about 60 times as great. Although only 33-40% as much ice (73 meters SLR/(73+125)) is left to melt, this suggests that rates of SLR will be substantially higher, at current rates of warming, than the 1.5 to 3 meters per century coming out of the most recent ice age. In four scenarios without CDR, mean rates of modeled SLR from 2100 to 2400 range from 4 to 11 meters per century.

Summary of Factors in Warming to 2400

Table 4 summarizes the expected future warming effects from feedbacks (to 2400), based on the analyses above.

Table 4: Projected GST Warming from Feedbacks, to 2400.

The 3.5°C warming indicated, added to 1.1°C warming since 1880, or 4.6°C, is 0.5°C less than the 5.1°C warming based on Equation (4) from the paleoclimate analysis. This gap suggests four overlapping possibilities. First, underestimations (perhaps sea ice and clouds) may exceed overestimations (perhaps snow) for the processes shown in Table 4. Underestimation of cloud feedbacks, and their consequent warming, is quite possible. Using Zelinka’s 0.46 Wm^–2K^–1 in this study, instead of the IPCC central estimate of 0.6, is one possibility. Moreover, recent research suggests that cloud feedbacks may be appreciably stronger than 0.6 Wm^–2K^–1 [48]. Second, change in the eight factors not modelled (black carbon, vegetation and land use, ocean and air circulation, Earth’s orbit and tilt, and solar output) may provide feedbacks that, on balance, are more warming than cooling. Third, temperatures used here for 4 and 14 Mya may be overestimated or should not be used unadjusted. Notably, the joining of North and South America about 3 Mya rearranged ocean circulation and may have resulted in cooling that led to ice periodically covering much of North America [49]. Globally, Figure 1 above suggests this cooling effect may be 1.0-1.6°C. In contrast, solar output increases as our sun ages, by 7% per billion years [50], so that solar forcing is now 1.4 W m^–2 more than 14 Mya and 0.4 more than 4 Mya. A brighter sun now indicates that, for the same GHG levels and albedo levels, GST would be 0.7°C warmer than it would have been 14 Mya and 0.2°C warmer than 4 Mya. Fourth, nothing (net) may be amiss. Underestimated warming (perhaps permafrost, clouds, sea ice, black carbon) may balance overestimated warming (perhaps snow, land ice, vegetation). The gap would then be due to a lower albedo climate sensitivity than 6.4°C, as discussed above using data for 1975-2016, because all sea ice and much snow vanish by 2400.

Natural Carbon Emissions

Permafrost: One estimate of the amount of carbon stored in permafrost is 1,894 GT of carbon [51]. This is about 4 x carbon that humans have emitted by burning fossil fuels. It is also 2 x as much as in Earth’s atmosphere. More permafrost may lie under Antarctic ice and the GIS. DeConto [52] proposed that the PETM’s large carbon and temperature (5-6°C) excursions 55 Mya are explained by “orbitally triggered decomposition of soil organic carbon in circum-Arctic and Antarctic terrestrial permafrost. This massive carbon reservoir had the potential to repeatedly release thousands of [GT] of carbon to the atmosphere-ocean system”. Permafrost area in the Northern Hemisphere shrank 7% from 1900 to 2000 [53]. It may shrink 75-88% more by 2100 [54]. Carbon emissions from permafrost are expected to accelerate, as the ground in which they are embedded warms up. In general, near-surface air temperatures have been warming twice as fast in the Arctic as across the globe as a whole [32]. More research is needed to estimate rates of permafrost warming at depth and consequent carbon emissions. Already in 2010, Arctic permafrost emitted about as carbon as all US vehicles [55]. Part of the carbon emerges as CH₄, where surface water prevents carbon under it being oxidized. That CH₄ changes to CO₂ in the air over several years. This study accounts for the effects of CO₂ derived from permafrost. MacDougall et al. estimated that thawing permafrost can add up to ~100 ppm of CO₂ to the air by 2100 and up to 300 more by 2300, depending on the four RCP emissions scenarios [56]. This is 200 GT of carbon by 2100 plus 600 GT more by 2300. The direct driver of such emissions is local temperatures near the air-soil interface, not human carbon emissions. Since warming is driven not just by emissions, but also by albedo changes and their multipliers, permafrost carbon losses from thawing may proceed faster than MacDougall estimated. Moreover, MacDougall estimated only 1,000 GT of carbon in permafrost [56], less than more recent estimates. On the other hand, a larger fraction of carbon may stay in permafrost soil in than MacDougall assumed, leaving deep soil rich in carbon, similar to that left by “recent” glaciers in Iowa.

Other Natural Carbon Emissions

Seabed CH₄ hydrates may hold a similar amount of carbon to permafrost or somewhat less, but the total amount is very difficult to measure. By 2011, subsea CH₄ hydrates were releasing 20-30% as much carbon as permafrost was [57]. This all suggests that eventual carbon emissions from permafrost and CH₄ hydrates may be half to four times what MacDougall estimated. Also, the earlier portion of those emissions may happen faster than MacDougall estimated. In all, this study’s modeled permafrost carbon emissions range from 35 to 70 ppm CO₂ by 2100 and from 54 to 441 ppm CO₂ by 2400, depending on the scenario. As stated earlier, this model simply assumes that other natural carbon reservoirs will add half as much carbon to the air as permafrost does, on the same time path. These sources include outgassing from soils now unfrozen year-round, the warming upper ocean, seabed CH₄ hydrates, and any net decrease in worldwide biomass.

Results

The Six Scenarios

1. “2035 Peak”. Fossil-fuel emissions are reduced 94% by 2100, from a peak about 2035, and phased out entirely by 2160. Phase-out accelerates to 2070, when CO₂ emissions are 25% of 2017 levels, then decelerates. Permafrost carbon emissions overtake human ones about 2080. Natural CO₂ removal (CDR) mostly further acidifies the oceans. But it includes 1 GT per year of CO₂ by rock weathering.
2. “2015 Peak”. Fossil-fuel emissions are reduced 95% by 2100, from a peak about 2015, and phased out entirely by 2140. Phase-out accelerates to 2060, when CO₂ emissions are 40% of 2017 levels, then decelerates. Compared to a 2035 peak, natural carbon emissions are 25% lower and natural CDR is similar.
3. “x Fossil Fuels by 2050”, or “x FF 2050”. Peak is about 2015, but emissions are cut in half by 2040 and end by 2050. Natural CDR is the same as for the 2015 Peak, but is lower to 2050, since human CO₂ emissions are less. This path has a higher GST from 2025 to 2084, while warming sooner from less SO₄ outweighs less warming from GHGs.
4. “Cold Turkey”. Emissions end at once after 2015. Natural CDR is only by rock weathering, since no new human CO₂ emissions push carbon into the ocean. After 2060, cooling from ending CO₂ emissions earlier outweighs warming from ending SO₂
5. “x FF 2050, CDR”. Emissions are the same as for “x FF 2050”, as is natural CDR. But human CDR ramps up in an S-curve, from less than 1% of emissions in 2015 to 25% of 2015 emissions over the 2055 to 2085 period. Then they ramp down in a reverse S-curve, to current levels in 2155 and 0 by 2200.
6. “x FF 2050, 2xCDR” is like “x FF 2050, CDR”, but CDR ramps up to 52% of 2015 emissions over 2070 to 2100. From 2090, it ramps down to current levels in 2155 and 0 by 2190. CDR = 71% of CO₂ emissions to 2017 or 229% of soil carbon lost since farming began [58], almost enough to cut CO₂ in the air to 313 ppm, for 2°C warming.

Projections to 2400

The results for the six scenarios shown in Figure 5 spread ocean warming over 1,000 years, more than half of it by 2400. They use the factors discussed above for sea level, water vapor, and albedo effects of reduced SO₄, snow, ice, and clouds. Permafrost emissions are based on MacDougall’s work, adjusted upward for a larger amount of permafrost, but also downward and to a greater degree, assuming much of the permafrost carbon stays as carbon-rich soil as in Iowa. As first stated in the introduction to Feedback Pathways, the model sets other natural carbon emissions to half of permafrost emissions. At 2100, net human CO₂ emissions range from -15 GT/year to +2 GT/year, depending on the scenario. By 2100, CO₂ concentrations range from 350 to 570 ppm, GLST warming from 2.9 to 4.5°C, and SLR from 1.6 to 2.5 meters. CO₂ levels after 2100 are determined mostly by natural carbon emissions, driven ultimately by GST changes, shown in the lower left panel of Figure 5. They come from permafrost, CH₄ hydrates, unfrozen soils, warming upper ocean, and biomass loss.

Figure 5: Scenarios for CO₂ Emissions and Levels, Temperatures and Sea Level.

Comparing temperatures to CO₂ levels allows estimates of long-run climate sensitivity to doubled CO₂. Sensitivity is estimated as ln(2)/ln(ppm/280) * ∆T. By scenario, this yields > 4.61° (probably ~5.13° many decades after 2400) for 2035 Peak, > 4.68° (probably ~5.15°) for 2015 Peak, > 5.22° (probably 5.26°) for “x FF by 2050”, and 8.07° for Cold Turkey. Sensitivities of 5.13, 5.15 and 5.26° are much less than the 8.18° derived from the Vostok ice core. This embodies the statement above, in the Partitioning Climate Sensitivity section, that in periods with little or no ice and snow [here, ∆T of 7°C or more – the 2035 and 2015 Peaks and x FF by 2050 scenarios], this albedo-related sensitivity shrinks to 3.3-3.4°. Meanwhile, the Cold Turkey scenario (with a good bit more snow and a little more ice) matches well the relationship from the ice core (and validated to 465 ppm CO₂, in the range for Cold Turkey: 4 and 14 Mya). Another perspective is the climate sensitivity starting from a base not of 280 ppm CO₂, but from a higher level: 415 ppm, the current level and the 2400 level in the Cold Turkey case. Doubling CO₂ from 415 to 830 ppm, according to the calculations underlying Figure 5, yields a temperature in 2400 between the x FF by 2050 and the 2015 Peak cases, about 7.6°C and rising, to perhaps 8.0°C after 1-2 centuries. This yields a climate sensitivity of 8.0 – 4.9 = 3.1°C in the 415-830 ppm range. The GHG portion of that remains near 1.8° (see Partitioning Climate Sensitivity above). But the albedo feedbacks portion shrinks further, from 6.4°, past 3.3° to 1.3°, as thin ice and most snow are gone, as noted above, plus all SO₄ from fossil fuels, leaving mostly thick ice and feedbacks from clouds and water vapor.

Table 5 summarizes estimated temperatures effects of 16 factors in the 6 scenarios to 2400. Peaking emissions now instead of in 2035 can keep eventual warming 1.1°C lower. Phasing out fossil fuels by 2050 gains another 1.2°C relatively cooler. Ending fossil fuel use immediately gains another 2.2°C. Also removing 2/3 of CO₂ emissions to date gains another 2.4°C relatively cooler. Eventual warming in the higher emissions scenarios is a good bit lower than what would be inferred by using the 8.2°C climate sensitivity based on an epoch rich in ice and snow. This is because the albedo portion of that climate sensitivity (currently 6.4°) is greatly reduced as ice and snow disappear. More human carbon emissions (the first three scenarios especially) warm GSTs further, especially from less snow and cloud cover, more water vapor, and more natural carbon emissions. These in turn accelerate ice loss. All further amplify warming.

Table 5: Factors in Projected Global Surface Warming, 2010-2400 (°C).

Carbon release from permafrost and other reservoirs is lower in scenarios where GSTs do not rise as much. GSTs grow to the end of the study period, 2400, except for the CDR cases. Over 99% of warming after 2100 is due to amplifying feedbacks from human emissions during 1750-2100. These feedbacks amount to 1.5 to 5°C after 2100, in the scenarios without CDR. Projected mean warming rates with continued human emissions are similar to current rates of 2.5°C per century over 2000-2020 [2]. Over the 21^st century, they range from 62 to 127% of the rate over the most recent 20 years. The mean across the 6 scenarios is 100%, higher in the 3 warmest scenarios. Warming slows in later centuries. The key to peak warming rates is disappearing northern sea ice and human SO₄, mostly by 2050. Peak warming rates per decade in all 6 scenarios occur this century. They are fastest not for the 2035 Peak scenario (0.38°C), but for Cold Turkey (.80°C when our SO₂ emissions stop suddenly) and xFF2050 (0.48°C, as SO₂ emissions phase out by 2050). Due to SO₄ changes, peak warming in the x FF 2050 scenario, from 2030 to 2060, is 80% faster than over the past 20 years, while for the 2035 Peak, it is only 40% faster. Projected SLR from ocean thermal expansion (OTE) by 2400 ranges from 3.9 meters in the 2035 Peak scenario to 1.5 meters in the xFF’50 2xCDR case. The maximum rate of projected SLR by 2400 is 15 meters from 2300 to 2400, in the 2035 Peak scenario. That is 5 times the peak 8-century rate 14 kya. However, the mean SLR rate over 2010-2400 is less than the historical 3 meters per century (from 14 kya) in the CDR scenarios and barely faster for Cold Turkey. The rate of SLR peaks from 2130 to 2360 for the 4 scenarios without CDR. In the two CDR scenarios, projected SLR comes mostly from the GIS, OTE, and the WAIS. But the EAIS is the biggest contributor in the three fastest warming scenarios.

Perspectives

The results show that the GST is far from equilibrium; barely more than 20% of 5.12°C warming to equilibrium. However, the feedback processes that warm Earth’s climate to equilibrium will be mostly complete by 2400. Some snow melting will continue. So will melting more East Antarctic and (in some scenarios) Greenland ice, natural carbon emissions, cloud cover and water vapor feedbacks, plus warming the deep ocean. But all of these are tapering off by 2400 in all scenarios. Two benchmarks are useful to consider: 2°C and 5°C above 1880 levels. The 2015 Paris climate pact’s target is for GST warming not to exceed 2°C. However, projected GST warming exceeds 2°C by 2047 in all six scenarios. Focus on GLSTs recognizes that people live on land. Projected GLST warming exceeds 2°C by 2033 in all six scenarios. 5° is the greatest warming specifically considered in Britain’s Stern Review in 2006 [59]. For just 4°, Stern suggested a 15-35% drop in crop yields in Africa, while parts of Australia cease agriculture altogether [59]. Rind et al. projected that the major U.S. crop yields would fall 30% with 4.2°C warming and 50% with 4.5°C warming [60]. According to Stern, 5° warming would disrupt marine ecosystems, while more than 5° would lead to major disruption and large-scale population movements that could be catastrophic [59]. Projected GLST warming passes 5°C in 2117, 2131, and 2153 for the three warmest scenarios. But it never does in the other three. With 5° GLST warming, Kansas, until recently the “breadbasket of the world”, would become as hot in summer as Las Vegas is now. Most of the U.S. warms faster than Earth’s land surface in general [32]. Parts of the U.S. Southeast, including most of Georgia, become that hot, but much more humid. Effects would be similar elsewhere.

Discussion

Climate models need to account for all these factors and their interactions. They should also reproduce conditions for previous eras when Earth had this much CO₂ in the air, using current levels of CO₂ and other GHGs. This study may underestimate warming due to permafrost and other natural emissions. It may also overestimate how fast seas will rise in a much warmer world. Ice grounded below sea level (by area, ~2/3 of the WAIS, 2/5 of the EAIS, and 1/6 of the GIS) can melt quickly (decades to centuries). But other ice can take many centuries or millennia to melt. Continued research is needed, including separate treatment of ice grounded below sea level or not. This study’s simplifying assumptions, that lump other GHGs with CO₂ and other natural carbon emissions proportionately with permafrost, could be improved with modeling for the individual factors lumped here. More research is needed to better quantify the 12 factors modeled (Table 5) and the four modeled only as a multiplier (line 10 in Table 5). For example, producing a better estimate for snow cover, similar to Hudson’s for Arctic sea ice, would be useful. So would other projections, besides MacDougall’s, of permafrost emissions to 2400. More work on other natural emissions and the albedo effects of clouds with warming would be useful.

This analysis demonstrates that reducing CO₂ emissions rapidly to zero will be woefully insufficient to keep GST less than 2°C above 1750 or 1880 levels. Policies and decisions which assume that merely ending emissions will be enough will be too little, too late: catastrophic. Lag effects, mostly from albedo changes, will dominate future warming for centuries. Absent CDR, civilization degrades, as food supplies fall steeply and human population shrinks dramatically. More emissions, absent CDR, will lead to the collapse of civilization and shrink population still more, even to a small remnant.

Earth’s remaining carbon budget to hold warming to 2°C requires removing more than 70% of our CO₂ emissions to date, any future emissions, and all our CH₄ emissions. Removing tens of GT of CO₂ per year will be required to return GST warming to 2°C or less. CDR must be scaled up rapidly, while CO₂ emissions are rapidly reduced to almost zero, to achieve negative net emissions before 2050. CDR should continue strong thereafter.

The leading economists in the USA and the world say that the most efficient policy to cut CO₂ emissions is to enact a worldwide price on them [61]. It should start at a modest fraction of damages, but rise briskly for years thereafter, to the rising marginal damage rate. Carbon fee and dividend would gain political support and protect low-income people. Restoring GST to 0° to 0.5°C above 1880 levels calls for creativity and dedication to CDR. Restoring the healthy climate on which civilization was built is a worthwhile goal. We, our parents and our grandparents enjoyed it. A CO₂ removal price should be enacted, equal to the CO₂ emission price. CDR might be paid for at first by a carbon tax, then later by a climate defense budget, as CO₂ emissions wind down.

Over 1-4 decades of research and scaling up, CDR technology prices may drop far. Sale of products using waste CO₂, such as concrete, may make the transition easier. CDR techniques are at various stages of development and prices. Climate Advisers provides one 2018 summary for eight CDR approaches, including for each: potential GT CO₂ removed per year, mean US$/ton CO₂, readiness, and co-benefits [62]. The commonest biological CDR method now is organic farming, in particular no-till and cover cropping. Others include several methods of fertilizing or farming the ocean; planting trees; biochar; fast-rotation grazing; and bioenergy with CO₂ capture. Non-biological ones include direct air capture with CO₂ storage underground in carbonate-poor rocks such as basalts. Another increases surface area of such rocks, by grinding them to gravel, or dust to spread from airplanes. They react with weak carbonic acid in rain. Another adds small carbonate-poor gravel to agricultural soil.

CH₄ removal should be a priority, to quickly drive CH₄ levels down to 1880 levels. With a half-life of roughly 7 years in Earth’s atmosphere, CH₄ levels might be cut back that much in 30 years. It could happen by ending leaks from fossil fuel extraction and distribution, untapped landfills, cattle not fed Asparagopsis taxiformis, and flooding rice paddies. Solar radiation management (SRM) might play an important supporting role. Due to loss of Arctic sea ice and human SO₄, even removing all human GHGs (scenario not shown) will likely not bring GLST back below 2°C by 2400. SRM could offset these two soonest major albedo changes in coming decades. The best known SRM techniques are (1) putting SO₄ or calcites in the stratosphere and (2) refreezing the Arctic Ocean. Marine cloud brightening could play a role. SRM cannot substitute for ending our CO₂ emissions or for vast CDR, both of them soon. We may need all three approaches working together.

In summary, the paleoclimate record shows that today’s CO₂ level entails GST roughly 5.1°C warmer than 1880. Most of the increase from today’s GST will be due to amplification by albedo changes and other factors. Warming gets much worse with continued emissions. Amplifying feedbacks will add more GHGs to the air, even if we end our GHG emissions now. Further GHGs will warm Earth’s surface, oceans and air even more, in some cases much more. The impacts will be many, from steeply reduced crop yields (and widespread crop failures) and many places too hot to survive sometimes, to widespread civil wars, billions of refugees, and many meters of SLR. Decarbonization of civilization by 2050 is required, but far from enough. Massive CO₂ removal is required as soon as possible, perhaps supplemented by decades of SRM, all enabled by a rising price on CO₂.

List of Acronyms

References

1. https://data.giss.nasa.gov/gistemp/tabledata_v3/
2. https://data.giss.nasa.gov/gistemp/tabledata_v3/GLB.Ts+dSST.txt
3. Hansen J, Sato M (2011) Paleoclimate Implications for Human-Made Climate Change in Berger A, Mesinger F, Šijački D (eds.) Climate Change: Inferences from Paleoclimate and Regional Aspects. Springer, pp: 21-48.
4. Levitus S, Antonov J, Boyer T (2005) Warming of the world ocean, 1955-2003. Geophysical Research Letters
5. https://www.nodc.noaa.gov/OC5/3M_HEAT_CONTENT/
6. https://www.eia.gov/totalenergy/data/monthly/pdf/sec1_3.pdf
7. Hansen J, Sato M, Kharecha P, von Schuckmann K (2011) “Earth’s Energy imbalance and implications. Atmos Chem Phys 11: 13421-13449.
8. https://www.eia.gov/energyexplained/index.php?page=environment_how_ghg_affect_climate
9. Tripati AK, Roberts CD, Eagle RA (2009) Coupling of CO₂ and ice sheet stability over major climate transitions of the last 20 million years. Science 326: 1394-1397. [crossref]
10. Shevenell AE, Kennett JP, Lea DW (2008) Middle Miocene ice sheet dynamics, deep-sea temperatures, and carbon cycling: a Southern Ocean perspective. Geochemistry Geophysics Geosystems 9:2.
11. Csank AZ, Tripati AK, Patterson WP, Robert AE, Natalia R, et .al. (2011) Estimates of Arctic land surface temperatures during the early Pliocene from two novel proxies. Earth and Planetary Science Letters 344: 291-299.
12. Pagani M, Liu Z, LaRiviere J, Ravelo AC (2009) High Earth-system climate sensitivity determined from Pliocene carbon dioxide concentrations, Nature Geoscience 3: 27-30.
13. Wikipedia – https://en.wikipedia.org/wiki/Greenland_ice_sheet
14. Bamber JL, Riva REM, Vermeersen BLA, Le Brocq AM (2009) Reassessment of the potential sea-level rise from a collapse of the West Antarctic Ice Sheet. Science 324: 901-903.
15. https://nsidc.org/cryosphere/glaciers/questions/located.html
16. https://commons.wikimedia.org/wiki/File:AntarcticBedrock.jpg
17. DeConto RM, Pollard D (2016) Contribution of Antarctica to past and future se-level rise. Nature 531: 591-597.
18. Cook C, van de TF, Williams T, Sidney RH, Masao I, al. (2013) Dynamic behaviour of the East Antarctic ice sheet during Pliocene warmth, Nature Geoscience 6: 765-769.
19. Vimeux F, Cuffey KM, Jouzel J (2002) New insights into Southern Hemisphere temperature changes from Vostok ice cores using deuterium excess correction. Earth and Planetary Science Letters 203: 829-843.
20. Snyder WC (2016) Evolution of global temperature over the past two million years, Nature 538: 226-
21. https://www.wri.org/blog/2013/11/carbon-dioxide-emissions-fossil-fuels-and-cement-reach-highest-point-human-history
22. https://phys.org/news/2012-03-weathering-impacts-climate.html
23. Smith SJ, Aardenne JV, Klimont Z, Andres RJ, Volke A, al. (2011). Anthropogenic Sulfur Dioxide Emissions: 1850-2005. Atmospheric Chemistry and Physics 11: 1101-1116.
24. Figure SPM-2 in S Solomon, D Qin, M Manning, Z Chen, M. Marquis, et al. (eds.) IPCC, 2007: Summary for Policymakers. in Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the 4th Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK and New York, USA.
25. ncdc.noaa.gov/snow-and-ice/extent/snow-cover/nhland/0
26. https://nsidc.org/cryosphere/sotc/snow_extent.html
27. ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/
28. Chen X, Liang S, Cao Y (2016) Satellite observed changes in the Northern Hemisphere snow cover phenology and the associated radiative forcing and feedback between 1982 and 2013. Environmental Research Letters 11:8.
29. https://earthobservatory.nasa.gov/global-maps/MOD10C1_M_SNOW
30. Hudson SR (2011) Estimating the global radiative impact of the sea ice-albedo feedback in the Arctic. Journal of Geophysical Research: Atmospheres 116:D16102.
31. https://www.currentresults.com/Weather/Canada/Manitoba/Places/winnipeg-snowfall-totals-snow-accumulation-averages.php
32. https://data.giss.nasa.gov/gistemp/tabledata_v3/ZonAnn.Ts+dSST.txt
33. https://neven1.typepad.com/blog/2011/09/historical-minimum-in-sea-ice-extent.html
34. https://14adebb0-a-62cb3a1a-s-sites.googlegroups.com/site/arctischepinguin/home/piomas/grf/piomas-trnd2.png?attachauth=ANoY7coh-6T1tmNEErTEfdcJqgESrR5tmNE9sRxBhXGTZ1icpSlI0vmsV8M5o-4p4r3dJ95oJYNtCrFXVyKPZLGbt6q0T2G4hXF7gs0ddRH88Pk7ljME4083tA6MVjT0Dg9qwt9WG6lxEXv6T7YAh3WkWPYKHSgyDAF-vkeDLrhFdAdXNjcFBedh3Qt69dw5TnN9uIKGQtivcKshBaL6sLfFaSMpt-2b5x0m2wxvAtEvlP5ar6Vnhj3dhlQc65ABhLsozxSVMM12&attredirects=1
35. https://www.earthobservatory.nasa.gov/features/SeaIce/page4.php
36. Shepherd A, Ivins ER, Geruo A, Valentina RB, Mike JB, et al. (2012) A reconciled estimate of ice-sheet mass balance. Science 338: 1183-1189.
37. Shepherd A, Ivins E, Rignot E, Ben Smith (2018) Mass balance of the Antarctic Ice Sheet from 1992 to 2017. Nature 558: 219-222.
38. https://en.wikipedia.org/wiki/Greenland_ice_sheet
39. Robinson A, Calov R, Ganopolski A (2012) Multistability and critical thresholds of the Greenland ice sheet. Nature Climate Change 2: 429-431.
40. https://earthobservatory.nasa.gov/features/CloudsInBalance
41. Figures TS-6 and TS-7 in TF Stocker, D Qin, GK Plattner, M Tignor, SK Allen, J Boschung, et al. (eds.). IPCC, 2013: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK and New York, NY, USA.
42. Zelinka MD, Zhou C, Klein SA (2016) Insights from a refined decomposition of cloud feedbacks. Geophysical Research Letters 43: 9259-9269.
43. Wadhams P (2016) A Farewell to Ice, Penguin / Random House, UK.
44. https://scripps.ucsd.edu/programs/keelingcurve/wp-content/plugins/sio-bluemoon/graphs/mlo_full_record.png
45. Fretwell P, Pritchard HD, Vaughan DG, Bamber JL, Barrand NE et al. (2013) Bedmap2: improved ice bed, surface and thickness datasets for Antarctica .The Cryosphere 7: 375-393.
46. Fairbanks RG (1989) A 17,000 year glacio-eustatic sea- level record: Influence of glacial melting rates on the Younger Dryas event and deep-ocean circulation. Nature 342: 637-642.
47. https://en.wikipedia.org/wiki/Sea_level_rise#/media/File:Post-Glacial_Sea_Level.png
48. Zelinka MD, Myers TA, McCoy DT, Stephen PC, Peter MC, al. (2020) Causes of Higher Climate Sensitivity in CMIP6 Models. Geophysical Research Letters 47.
49. https://earthobservatory.nasa.gov/images/4073/panama-isthmus-that-changed-the-world
50. https://sunearthday.nasa.gov/2007/locations/ttt_cradlegrave.php
51. Hugelius G, Strauss J, Zubrzycki S, Harden JW, Schuur EAG, et al. (2014) Improved estimates show large circumpolar stocks of permafrost carbon while quantifying substantial uncertainty ranges and identifying remaining data gaps. Biogeosciences Discuss 11: 4771-4822.
52. DeConto RM, Galeotti S, Pagani M, Tracy D, Schaefer K, al. (2012) Past extreme warming events linked to massive carbon release from thawing permafro.st Nature 484: 87-92.
53. Figure SPM-2 in IPCC 2007: Summary for Policymakers. In: Climate Change 2007: The Physical Science Basis.
54. Figure 22.5 in Chapter 22 (F.S. Chapin III and S. F. Trainor, lead convening authors) of draft 3rd National Climate Assessment: Global Climate Change Impacts in the United States. Jan 12, 2013.
55. Dorrepaal E, Toet S, van Logtestijn RSP, Swart E, van der Weg, MJ, et al. (2009) Carbon respiration from subsurface peat accelerated by climate warming in the subarctic. Nature 460: 616-619.
56. MacDougall AH, Avis CA, Weaver AJ (2012) Significant contribution to climate warming from the permafrost carbon feedback. Nature Geoscience 5:719-721.
57. Shakhova N, Semiletov I, Leifer I, Valentin S, Anatoly S, et al. (2014) Ebullition and storm-induced methane release from the East Siberian Arctic Shelf. Nature Geoscience 7: 64-70.
58. Sandeman J, Hengl T, Fiske GJ (2018) Soil carbon debt of 12,000 years of human land use PNAS 114:36, 9575-9580, with correction in 115:7.
59. Stern N (2007) The Economics of Climate Change: The Stern Review. Cambridge University Press, Cambridge UK.
60. Rind D, Goldberg R, Hansen J, Rosenzweig C, Ruedy R (1990) Potential evapotranspiration and the likelihood of future droughts. Journal of Geophysical Research. 95: 9983-10004.
61. https://www.wsj.com/articles/economists-statement-on-carbon-dividends-11547682910
62. www.climateadvisers.com/creating-negative-emissions-the-role-of-natural-and-technological-carbon-dioxide-removal-strategies/

Introduction

Climate change is having great impact on agricultural productivity worldwide. Agriculture is strongly influenced by weather and climate [1,2]. Climate change and variability adversely affect environmental resources such as soil and water upon which agricultural production depends, which poses a serious threat to sustainable agricultural production [2]. In Ghana climate variability and change is expected to have an adversely effect on the agriculture sector. According to the NIC, (2009) by 2030 temperature are projected to rise by 0.5 °C. This situation would result in fewer rainy days and more extreme weather conditions like prolonged droughts. The impacts of a changing climate will have direct and indirect effects on global and domestic food systems [3,4]. Rioux [5] reported that climate change has affected yields in food crop production in many Africa countries. If the issues of climate change and variability are not addressed incomes and food security of rural households in Ghana would be undermined because there would be increased incidence of diseases and pest as well as prolonged variable rainfall patterns.

Cocoa production employs over 15 million people worldwide with over 10.5 million workers in West Africa [6]. Cocoa, in addition to cereals and other root and tuber  crops  contribute  largely  to  food security in Ghana. In Ghana cocoa production is an essential component of  rural  livelihoods  and  its  cultivation  is  considered a ‘way of life’ in many production communities [7]. The cocoa sub sector cocoa employs about 800,000 farm families spread across the cocoa growing regions of Ghana and generating about $2 billion in foreign exchange annually [8,9]. The expansion of cocoa production is replacing substantial areas of primary forest. It’s of no surprise that the total area under cocoa cultivation increased by 50,000 hectares between 2012 and 2013 and there is no indication that the rate is slowing down. According to Anim Kwapong et al. [10] the government of Ghana recognizes that climate change is already negatively affecting Ghana’s cocoa sector in myriad ways and that, it is likely to continue hampering Ghana’s environmental and socio-economic prospects in the coming decades. Cocoa agroforestry system has been identified as is an important strategy that can ameliorate climate change [11].

This system can play a dual role of mitigation and adaptation, which makes it one of the best responses to climate change. It is noted that agroforestry has multi-functional purposes which makes it one of  the most promising strategies for climate change adaptation [11,12]. The use of trees and shrubs in agricultural systems help to tackle the triple challenge of securing food security, mitigation and reducing the vulnerability and increasing the adaptability of agricultural systems to climate change [13,14]. With this view, serious attention must be given to cocoa agroforestry which is capable of reducing temperatures and enhancing the growing of cocoa thus sustaining livelihood of many households in this climate changing pattern. According to previous studies [11,13,15] agroforestry as an adaptation strategy could sustain agricultural production and enhance farmers’ ability to improve livelihoods and will minimize the impacts of climate change which include drought, variable rainfall and extreme temperatures. Agroforestry as a forest-based system plays a significant role in conserving existing carbons, thereby limiting carbon emissions and also absorbing carbons that are released into the atmosphere [16]. Nair [17] also indicated that agroforestry has received international attention as an effective strategy for carbon sequestration and greenhouse mitigation. Cocoa agroforestry can increase farmers’ resilience and position them strategically to adapt to the impacts of a changing climate. This system of cocoa production can be very useful because it generates quite substantial benefits on arable lands in diverse ways; trees in agricultural fields improve soil fertility through control of erosion, improve nitrogen content of the soil and increase organic matter of the soil [18,19]. Agroforestry can also transform degraded lands into productive agricultural lands and improves productive capacities of soils [18]. Although agroforestry is not new in Ghana, it is quite optimistic that effective adoption to climate change will contribute towards the achievement of sustainable development and to a large extent, the attainment of the Sustainable Development Goals (SDGs). Despite the immeasurable benefits of cocoa agroforestry system, adoption is not widespread and for that matter success stories are found in isolated cocoa farming areas among few adapters of cocoa agroforestry system initiatives. Aidoo and Fromm [20] report that although cocoa farmers are aware about sustainability issues, they hardly adopt sustainable production practices. It is quite not always the case that policies are implemented as they were intended and so the need to assess farmers’ perspectives on cocoa agroforestry adoption and implementation especially when climate change has become a serious constraint to cocoa production in Ghana. Traditional coping mechanisms to the impact of climate change in the Western Region of Ghana include mixed cropping, non-farm activities and traditional agroforestry practices by some individual cocoa farmers. However, non-shade cocoa production systems, bush burning, slash and burn farming methods expose the cocoa communities to further impacts of climate change. This calls for swift attention from all, especially cocoa farmers in the study communities to tackle the problem. Despite the economic, environmental and sustainable cocoa production potential via agroforestry systems, farmers have not adopted cocoa agroforestry practices entirely especially in Sefwi Wiawso District. Understanding cocoa farmers decision making processes in ensuring sustainable food supply and cocoa yield in cocoa agroforestry system is critical. Research frontiers in cocoa agroforestry systems need to be identified and better understand barriers to adoption and the development of strategies to support cocoa agroforestry that enhance food security in climate changing conditions. The objectives of this study are therefore to empirically assess the factors that affect farmers’ decision to adopt cocoa agroforestry systems and determine cocoa farmers’ perception on cocoa agroforestry as an adaptation strategy to climate change.

Methodology

The study was conducted at Sefwi Wiawso in the Western and region of Ghana. The district lies within latitudes 6º 00“and 6º 30 North and Longitudes 2º 15‟ and 2º 45 West. The District covers an area of about 2,634 square kilometers. The detailed hydrometeorological characteristics of the study area are provided in Table 1.

A stratified random sampling technique was employed in the selection of the 300 cocoa farmers interviewed for the study. In the first stage, Western Region was purposively selected due to the fact that apart from being one of the highest cocoa producing regions    in Ghana, it is one of the regions which has experienced significant impact as a result of climate change. In the second stage, Sefwi Wiawso was randomly selected. In the third stage, five communities were randomly selected. In the final stage 60 cocoa farmers were randomly selected from each village. Primary data were employed in the study. The primary data consisted of qualitative data and household survey interviews. Specifically, the primary data were collected through focus group discussions (FGD), stakeholder interviews, and field observations. The household survey interviews employed both open- ended and close ended survey instruments.

To examine the factors that influence a household’s decision to participate in agroforestry a logistic regression model was employed.

The model was specified as:

Where: i = 1, 2, 3………., k are the observations, α= constant. β = the regression parameter to be estimated. βX= linear combination of independent variables.  Z_i= the log odds of choice for the  ithobservation. P_i= the probability of observing
a specific outcome of the dependent variable (adoption). X_n = nth explanatory observation. u = the error term.

Results and Discussion

The gender composition of the cocoa farmers among revealed that 81.5 percent of the respondent are males with 19.5 percent been females. This indicates that cocoa production is a male dominated occupation in the study area. In Ghana cocoa production is considered a male job but this is not the situation at the study sites because both women and men play a critical role in the production cycle. Within the last 30 years, cocoa farmers observed some impacts of climate change in the study communities, information gather from the cocoa farmers showed that there has been varying pattern in rainfall and sunshine. With regards to drought, overwhelming 98 percent of cocoa farmers reported the occurrence of drought in the study area and linked it to climate change. The pattern of rainfall distribution has changed as reported from the study. The study reported high level of windstorm, high incidence of flooding and frequent occurrences of pests and disease on their cocoa farms in recent time. These are attributed to climate change. Frequent felling of trees, non-shade cocoa production systems, wood harvesting for charcoal and firewood and bush burning among others were mention as some course of changing climate in the farming communities. About two thirds of the farmers reported unplanned trees harvesting as a major cause for variable rainfall thus climate change. This suggest that majority of farmers are aware of some of the causes of climate change in the study area. About 58 percent of cocoa farmers are using doing the non-shade cocoa production system. This result confirms a report [21], indicating that high proportion of Ghana’s cocoa is grown in full sun at the expense of primary or secondary forest conversion. A study [22] reported that shaded tree densities, and average number of tree species per hectare vary according to cultural tradition and ethnic group, age of farms, proximity to markets, and intensity of farming, this situation is similar to that of the study area after personal interaction with the cocoa farmers. This current trend of no shade is not only common in Ghana but other cocoa growing countries like Cote d’Ivoire, Malaysia, Indonesia and Ecuador. A study [23] in Ecuador reported that half of the new cocoa plantations are now full-sun and are from high-yielding variety. A study [24] also revealed that in Sulawesi cocoa farmers are switching from long-fallow shifting cultivation of food crops to intensive full-sun cocoa. This current trend of cocoa production put the food security of these cocoa farmers in doubt with the impact of climate change.

Cocoa farmers acknowledge the benefits of adopting cocoa agroforestry system in cocoa production. Farmers indicated that cocoa agroforestry has the potential of maintaining soil moisture, improving soil fertility as well as suppressing weeds within the cocoa farm. A study by Bentley [23] on cocoa farmers in Ecuador also indicated similar characteristics. Cocoa farmers acknowledged that no shade cocoa system is agriculturally unsustainable and is becoming common in the study area. The study reported that cocoa agroforestry mimics the natural sub canopy cover of traditional cocoa tree in the forest thus good practice to mitigate climate change. The shade trees selected by the cocoa farmers need to provide products and additional income when sold. Terminalia superb, Milicia excels, Terminalia ivorensis, Cedrella odorata,Ceiba pentandra and Ceiba pentandraas are the most dominant shade tree on cocoa farms  and are retained because of their economic importance. Eighty-five percent have little knowledge about the tree rights in the community although there are existing policies and legislations in Ghana. The average knowledge of useful species in this cocoa farming communities are fading out. For example, some of the younger farmers interviewed retain shade trees on an interest in the knowledge of their parents and grandparents.

Cocoa farmers have various levels of perception on certain characteristics of cocoa agroforestry. About 54 percent of cocoa farmers strongly perceive that cocoa agroforestry improves yield of cocoa. These trees ensure a microclimate condition which enhance the yield of the cocoa and thus mitigate climate change. Other perception held by cocoa farmers for cocoa agroforestry are enhancing soil moisture, improve farm humidity and environment, protecting young cocoa trees from pest and diseases and direct sun rays (Table 2).

Factors Affecting Adoption of Climate-Smart Agriculture Innovations in Isolation and in Combination

Farmers’ adaption decisions were found to be influenced by several varying factors. The factors include farming experience, agricultural land size, belonging to farmer association, access to extension services, awareness of climate change, and experience in farming.

Results from the regression are reported here to tell the factors determining of adoption of individual farmer. The base category used in the analysis was non-adoption. Table 3 report coefficients and marginal effects from MNL regression respectively. Marginal effects (Table 3) are reported and discussed here. In this instance,  the marginal effects measure the expected change in probability of   a certain choice (of a cocoa agroforestry system) being made with respect to a unit change in an explanatory variable, all in comparison to the no adoption category.

***Significant at 1%, **Significant at 5%, *Significant at 10%.

Results are compared to the base category of no-adoption. The results indicated that adoption of cocoa agroforestry is negatively associated with age of farmer and positively associated with agriculture land size, experience in farming, member of farmer association, gender, awareness of climate change and access to extension service. Results imply that probability of adopting cocoa agroforestry decreases with ageing of cocoa farmer possibly due to risk aversion of innovative practices like cocoa agroforestry by older cocoa farmers. The positive association of cocoa agroforestry adoption with agriculture land size imply that larger plot sizes could be more flexible to experiment with cocoa agroforestry. Also, the positive association of extension could be due to availability of information for cocoa farmers with access to it. The factors of cocoa agroforestry adoption is in agreement with studies [25,26]. Extension services are very critical for availing necessary information on cocoa agroforestry. Overall, results show the importance of cocoa agroforestry system at the farmer level in building resilience to climate variability and change as well as other productivity related challenges in cocoa farming in Ghana. Adoption of cocoa agroforestry system reduces the impacts of climate change on cocoa productivity and hence farmer incomes. The enhanced impact of adopting cocoa agroforestry systems possibly arise as a result of the micro climatic conditions that is favorable for cocoa production. Findings of the study conform to other related literature that indicates that, adoption  of new agricultural technologies needs to positively impact on productivity, income and other welfare related variables of the adaptors.

Conclusion and Recommendation

Cocoa researchers and development partners are becoming more concern with welfare of cocoa farm in Ghana by promoting cocoa agroforestry systems which is essential in a bid to improve climate resilience. Cocoa agroforestry has the potential to improve soil fertility, regulate soil temperature, control soil moisture among other benefits. The study outcomes have shown that climatic changes have occurred over the years and these have had effect on the annual cocoa yield. The study revealed that some cocoa farmers are presently ignorant about their tree ownership on their farms. It therefore recommended that agricultural extension officers should educate these farmers on tree rights. Cocoa farmers in the study areas have noticed changes   in climate conditions through their own experiences and careful observations over the year of farmers. Also, respondents reported that cocoa agroforestry systems can offer numerous environmental, social and financial benefits, and can lead to an alternative way to mitigate climate change and variability. Land size, member of farmer association, experience in farming, awareness of climate change and access to extension service are the main factors that influence cocoa farmers’ decision to adopt cocoa agroforestry system. There is the need for effective provision of extension services through farmer field school programs. Programs of this nature have the potential to change farmers’ attitudes towards adopting a technology. Access to information and credit needs to be enhanced so as to get the needed logistics for managing cocoa agroforestry systems. This would facilitate farmers’ access to information about technical issues of the systems and how it can be managed in mitigating climate change. Finally, government should support cocoa famers through subsidies and long-term loans. There is also the need for more concerted and strong collaborative effort among Ghana COCOBOD, the Ministry of Food and Agriculture and Forestry Commission so as to reach greater a policy impacts on cocoa agroforestry system.

References

1. Parry L (2019) Climate Change and World Agriculture. Routledge Library Editions: Pollution, Climate and Change, London, 172.
2. Gornall J, Betts R, Burke E, Clark R, Camp J, et al. (2010) Implications of climate change for agricultural productivity in the early twenty-first century. Philos. Trans R Soc B Biol Sci 5: 2973-2989. [crossref]
3. Lake IR, Hooper L, Abdelhamid A, Bentham G, Boxall ABA, et al. (2012) Climate change and food security: Health impacts in developed countries. Environ Health Perspect 120: 1520-1526. [crossref]
4. Edwards F, Dixon J, Friel S, Hall G, Larsen K, et al. (2011) Climate change adaptation at the intersection of food and health. Asia Pac J Public Health 23: 91-104. [crossref]
5. Rioux J (2012) Nature & Faune 26: 63-68.
6. De Lattre-Gasquet M, Despéraux D, Barel M (1998) ‘Prospective de la Filière du Cacao Plantation’. Recherche Développment 5: 423-434
7. Nunoo I and Owusu V (2015) Comparative analysis on financial viability of cocoa agroforestry systems in Ghana. Environment Development and Sustainability 19.
8. COCOBOD (2018) Ghana Cocoa Board Handbook16th ed. Jamieson’s Cambridge Faxbooks Ltd, Accra.62pp.
9. Ministry of food and Agriculture (2017) Directorate of Agricultural Extension Services: Agricultural Extension Approaches Being Implemented in Ghana.
10. Anim Kwapong, et al. (2005) Vulnerability and Adaptation Assessment under the Netherlands Climate Change Studies Assistance Programme Phase 2.
11. Kuyah S, Whitney CW, Jonsson M, et al. (2019) Agroforestry delivers a win-win solution for ecosystem services in sub-Saharan Africa. A meta-analysis. Agron Sustainm Dev 39: 47.
12. Campbell ID, Durant DG, Hunter KL, Hyatt KD (2014) Food production. In Canada in a Changing Climate: Sector Perspectives on Impacts and Adaptation 99–134
13. Carsan S, Stroebel A, Dawson I (2014) Can agroforestry option values improve the functioning of drivers of agricultural intensification in Africa? Curr Opin Environ Sustain 6: 35-40.
14. McCabe Colin (2013)”Agroforestry and Smallholder Farmers: Climate Change Adaptation through Sustainable Land Use” Capstone Collection.
15. Syampungani SC (2010) The Potential of Using Agroforestry as a Win-Win Solution to Climate Change Mitigation and Adaptation and Meeting Food Security Challenges in Southern Afri. Agricultural Journal 5: 80-88.
16. Mbow C, Smith P, Skole D, et al. (2014) Achieving mitigation and adaptation to climate change through sustainable agroforestry practices in Africa. Curr Opin Environ Sustain 6: 8-14.
17. Nair PK (2009). J Plant Nutr Soil Sci 172: 10-23.
18. Pinho CR, Miller PR, Alfaia SS (2012) Agroforestry and the Improvement of Soil Fertility: A View from Amazonia. Applied and Environmental Soil Science 2012: 11.
19. Thangataa PH, Hildebrand PE (2012) Carbon stock and sequestration potential      of agroforestry systems in smallholder agroecosystems of sub-Saharan Africa: mechanisms for reducing emissions from deforestation and forest degredation (REDD+). Agric Ecosyst Environ 158: 172-183.
20. Aidoo R, Fromm I (2015) Willingness to Adopt Certifications and Sustainable Production Methods among Small-Scale Cocoa Farmers in the Ashanti Region of Ghana. Journal of Sustainable Development 8: 33-43.
21. UNDP (2011) Greening the sustainable cocoa supply chain in Ghana.
22. Sonwa DJ (2004) Biomass management and diversification within cocoa agroforests in the humid forest zone of southern Cameroon. PhD thesis. Institute fur Gartenbauwissenshaft der Rheinischen FriedrichWilhelms-Universitat Bonn.
23. Bentley JW, Boa E, Stonehouse J (2004) Neighbor trees: Shade, intercropping, and cacao in Ecuador. Human Ecology 32: 241-270.
24. Belsky JM, Seibert S (2003) Cultivating cacao: implications of sun-grown cacao on local food security and environmental sustainability. Agric Human Values 20: 277- 285.
25. Mazvimavi K, Twomlow S (2009) Socioeconomic and institutional factors influencing adoption of conservation farming by vulnerable households in Zimbabwe. Agric Syst 101: 20-29.
26. Makatea C, Makateb M, Mangoc N, Sizibad S (2019) Increasing resilience of smallholder farmers to climate change through multiple adoption of proven climate-smart agriculture innovations. Lessons from Southern Africa. Journal of Environmental Management 231: 858-868.

Abstract

Management of climate via vegetation mainly focuses on the CO₂ sequestration activity of plants. Ecologists and Meteorologists so far agree that vegetation has an impact on micro and meso climatic level. Settlement of new vegetation on bare steppe ground over thousands of square km created within short time as seen in a “Great Green Wall” (GGW) – this still is a “new” engineering event, climatic evaluation of greening of entire regions is only starting. Large scale vegetation in semi-arid areas may have a role as direct meso – and macro climatic factor, developing over decades. Discrepant results are found in simulation models (afforestation related risk of heat in same or neighboring region) versus biophysical analysis of satellite data (warming effect of deforestation in dry climate). In trying to explain this discrepancy the reported effects of large scale afforestation in the Chinese GGW on regional and continental climate are reviewed, as reported for model regions sized a few thousand square km. Long term data showing a mitigating effect on wind, temperature and dryness, an important function of trees in breaking hot dry desert wind, a change to moderately humid climate and the critical minimum density of tree cover are reported. Potential errors underlying the simulation models are being discussed. We derive that the first signs of a potential direct meso and macroclimatic effect of additional vegetation in dry semi-arid and arid areas may become visible in the Chinese GGW, which would mean a duplicate climate mitigating effect here. As more and more afforestation areas of this GGW are established this effect is expected to develop in even larger regions during the next two decades.

Keywords

semi-arid, macro climatic, afforestation, CO₂ independent, climate protection, Great Green Wall

Introduction

There is an increasing interest in managing climate, globally. The topic of managing global climate by means of additional vegetation so far is focused mainly on the CO₂ sequestration activity of plants. Natural climate solution projects are aiming to gain a maximum amount of CO₂ fixation, therefore plantation projects were started preferably in regions where high amount of CO₂ can be sequestered within short time, ie where fast growth of trees is supported by humidity of the local climate. The direct climatic impact of vegetation observed in hot dry regions is: breaking of hot desert winds, cooling effect via evapotranspiration and shadow, increasing water storage capacity of the ground, introduction of a hydrate cycle, etc. Ecologists and Meteorologists agree in that direct effects of vegetation can be demonstrated on micro and meso climatic level. The reason for limiting it to a local and regional scale possibly is because the settlement of new tree vegetation over ten thousands of square km, developing, e.g. on bare steppe ground within short time as seen in the “Great Green Walls” (GGW)– this rarely happened before in human history, so there was no opportunity to observe a macro climatic impact. If afforestation of a huge area with desert-like climate and formerly very sparse vegetation has an impact on climate – how long would it take for such new „savannah plantation“ to really show an impact on the dominating semi-arid (almost desert like) climate?

Evidence of such effects of vegetation so far can only be shown indirectly from analysing the meteorological outcome of large scale deforestation, which is leading to weather extremes. If taking place on several continents at the same time (as is the case with tropical forests) such deforestations are expected to bear risk of global temperature rise [1]. Climatic evaluation of the newly existing large scale GGW in Northern China is only starting [2, 3]. This “North Shelterbelt Development Program” was built on the Southern edges of the Gobi and Taklamakan deserts, predominantly in semi-arid climate, an area measuring up to 4,800 km from West to East.

The oldest GGW was started in the early 1970s in Algeria (1,500 km), and the so far largest GGW is planned since 2005 in the South (Sahel) and North of the Sahara Desert. Today the idea is to create a network of vegetation areas over a territory of more than 7,500 km from coast to coast across the North African continent. It is coordinated by the African Union Commission [4, 5]. New large scale afforestation in hot dry climate needs several decades to get established, vegetation here may directly reduce ground temperature, cause regularity of precipitation and humidity of soil and air. Additional evidence of trees and shrubs in semi-arid areas to have a possible cooling effect on surface temperature is coming from the biophysical investigation of vegetation changes on the energy balance in a global context [6]. On the other hand, two studies simulating the impact of large scale afforestation in semi-arid climate are reviewed here which are finding a risk of surface temperature warming in the planted area or the neighboring regions. The warming trend typically found in simulations is connected to the expected changes in albedo. To better understand the discrepant finding I will review the first long-term real life climate data published for the Chinese GGW.

Review

Afforestation in dry climate – Biophysical analysis and Simulation models

Only recently it has become possible to evaluate the effects of vegetation changes on the energy balance in a global context by means of satellite data analysis. Duveiller et al. [6] have shown that the conversion of forests into grassland or agricultural land in dry climate will lead to rise in mean land surface temperatures. The local effects of vegetation loss or land degradation are an increase in the reflected portion of the short wave radiation, and this was most significant in dry climate regions. The resulting emitted long wave radiation is higher in dry regions and lower in northern latitudes. In addition, vegetation loss will lead to strongly reduced latent heat stream, particularly in tropical climate.

„The type of vegetation covering the landscape has a direct influence on local climate through its control of water and energy fluxes. The albedo (brightness) of the vegetation cover will determine how much energy is reflected back into space as shortwave radiation. Its roughness determines how much mixing of air occurs between the atmosphere and the vegetation canopy. The depth and structure of its rooting system can determine how much soil moisture and groundwater might be tapped and thus how much heat can be dissipated through evapotranspiration or latent heat flux. The balance of all these surface properties determines the direct influence of vegetation on the surface energy budget and ultimately on the local temperature” [7]. It therefore seems that changes in surface properties resulting in reduced number of trees in regions with dry and warm climate can lead to a local warming effect. Can we derive from this finding that, vice versa plantation of trees on bare ground in dry climate will have a cooling effect on surface temperatures? The albedo changing effect of vegetation cover is a strong factor in current climate simulation models. Typically, these models are finding that albedo will be reduced by vegetation, i.e. in comparison with the highly reflective bare ground of steppe or desert, the reduced reflection of sun heat radiation will lead to warming of surface temperature via the non-reflected portion absorbed by vegetation.

Simulation of a Western African GGW [8] has investigated how a seamless vegetation cover with evergreen broad leafed plants in a several hundred km wide area (here called „Savannah“) in the South of the Sahel would impact the number of days with extreme hot temperature, since heat waves are becoming more frequent in some areas of the world and they could as well become a risk for man and agriculture here in the South of the Sahel. The simulation is finding that indeed the number of days with extremely hot temperature will increase over the Savannah region, whereas temperature reduction will be found in the „Guinea“ region in the South along the Ivory Coast and in the Sahel region in the North. The increase in days with hot temperature in the afforested „Savannah“ region would predominantly occur during the dry season. At the end of this article it is stated that further analysis work is needed due to some uncertainty factors in order to come to more robust conclusions but that afforestation would lead to increased risk of heat waves in the „Savannah“ and a reduced risk for regions of comparable size in the North and South of it. Another investigation has been conducted to simulate afforestation in the East of South Africa [9], and the results are similar: In the simulated scenario afforestation would lead to reduced albedo and an increase in surface temperature over the plantation area, as well as a certain cooling effect on the neighbouring regions. Some areas would become more dry, other areas of South Africa may get more precipitation than before. Therefore, afforestation could lead to unfavourable changes in local climate in unpredictable areas, which is why besides the positive biogeochemical impact of large scale afforestation also its possible biophysical effects needed to be considered. In trying to overcome the dilemma of the conflicting results, I will review the first real life climate data published for the world´s largest GGW in China, separated by their potential regional (meso climatic) and continental (macro climatic) effects.

Examples of direct meso climatic effects of vegetation in semi-arid climate

The Chinese state and part of its population have been tackling the “Northern Shelter Belt” project since the 1970’s by planting a reported 66 billion trees along the roads, ditches, ponds, and cultivated land ridges, with the aim of a total of 100 billion trees and shrubs planted by 2050.Today these activities are supported by increasingly sophisticated technology resulting in the (re-) greening of steppes and even sand deserts in a gigantic large scale, on an area of 4,800 x 1,500 km. By 2050 these measures are hoped to improve soil degradation on 40% of China´s total area [2, 3].  A detailed case study by Zhuang et al. [10] published in 2017 is based on long-term data from a ”show case“ region of 152,000 hectars (an area of about 50 x 30 km) in Northern Jiangsu. Some 132,600 hectares of the total ground had been desertified until the early 1950´s when afforestation with millions of trees was started here as one of the first regions in the fight against desertification. It is not reported but the proximity to the Yellow River may have made afforestation easier. A marked improvement of the regional climate data is reported: air humidity has increased, the number of days with dust winds is decreased and the formerly steppe landscape was transformed into a green patchwork of forests and agriculture. Reliable publications of detailed afforestation related climate parameters are still rare, therefore the findings are presented in more detail.

The authors claim that today, “the formerly extremely severe climate along the old course of the Yellow River has been fundamentally changed. The improved quality of the regional environment is verified by the greatly increased productivity and welfare of the people. The saline alkali soil has been treated, along with poverty, transforming a beggar’s hometown into a modern region, famous as a producer of food, fruits, vegetables and wood.“ [10]. Regional climate data for the last 66 years were documented by the Fengxian Meteorological Bureau. Data are showing a reduction in strong wind days per year by 80%, reduction in maximum wind speed from 26 to 11 m/sec and reduction of the average wind speed over the ground by 90%. The forested area has been expanded within 60 years, starting from 3% in the 1950´s to 36.9% in the 2010´s. This would have transformed the long term trend of sand storms and desertification into more humid climate in which catastrophic droughts have become rare, despite the underlying global warming mega trend. The local climate has benefitted from reduced temperature extremes, reduced strength and frequency of sand storms and more days with fog.

Precipitation data before / after afforestation are not presented in this paper. However, from 1958 to 1980 the average relative humidity in June had been between 55 and 80%, then during the last 30 years it has varied from 78 to 90%. The increase in relative humidity possibly would result from an increase in evapotranspiration of trees and shrubs and on the other hand from the markedly reduced frequency of strong winds and decreased average yearly wind speed. The number of foggy days per year in this region is reported to have increased from 10 to 20 days (1958 -1971), to 18 to 35 days (1972 – 2000), and 35 to 45 days (2001 – 2013). Before 1960 there were 1 to 22 days with hot dry wind per year, this value has gone down to 0 to 6 (1981 – 2005) and 0 to 3 per year (2005 – 2013). This finding is interesting as it is showing an important impact of vegetation on hot dry desert wind that had caused extreme temperatures in the past – and it contradicts the expected role of a reduced albedo that is to be assumed for an increase in vegetation coverage by 33%. Today this feature of hot dry desert winds seems to have mostly gone and a reduced average wind speed is caused by the additional vegetation. In China, as in many other countries a recent trend of warming and increase in droughts is found, as shown for the period from 1982 to 2011 in [7]. Despite this fact, Zhuang et al report that the June average temperatures have remained constant over the last 60 years in this Northern Jiangsu region. Given the global warming trend (with a reported increase of about 1.5 degrees for China during the last three decades) this arguably may be considered a net decrease of surface temperature.

The authors conclude that, “with constant application of reforestation for 50 years, the regional climate in the old course of Yellow River has improved greatly, from its former long term status as a region of sandstorms and desertification, into a region that can be considered as being intermediate between mesic and humid in weather, and with few natural disasters. Sandstorms, dry-hot wind and saline alkali soil have been eliminated at the root source, along with poverty of the local population.”[10]. The authors propose that, “even though a single or several plots of trees might be net consumers of water in arid and half arid region, millions of trees may have a ‘‘mass effect function on improving regional climate.’’

Furthermore, based on long term climate data a „critical mass“, i.e. a minimum number of trees per area required to find measurable climatic effects of vegetation was discovered. In the hostile semi-arid basic climate a reported 16% and higher tree coverage had led to the moderately sub humid conditions observed today. Less marked results so far are reported in another example. The arid Kubuqi desert is located in the Ordos prefecture of Inner Mongolia, an Autonomous Region in the Northwest of China. Here, a total area of almost 6,000 square km of sand desert has been greened. This achievement was sponsored by a private ecology and investment company, Elion Research Ltd. since 1988 [11, 12]. “Emerging private enterprises such as Elion have played an important role in desertification control and governance in the Kubuqi Desert with the support of local government in terms of policies, planning, and infrastructure construction” [12]. Along the south bank of the Yellow River, Elion has established shelter forest in a belt of 242 km length and 5 to 20 km wide, consisting of trees, bushes and grass. Kubuqi has a temperate continental arid monsoon climate (Köppen class BWk, desert climate), with a long cold winter and warm short summer. January is the coldest month with average of –11.7ºC, July is the hottest month with average of 22.1ºC [12].

In a newspaper article precipitation is reported to have increased during the last 30 years from 100 mm to more than 400 mm in 2018 in this part of the desert [11]. However, there is a constant risk for such reports to originate from biased source. A reliable report including meteorological long term data published by the United Nations (UNEP) in 2015 [12] has found only around 10 % increase in precipitation. The shape of the main tree plantation area is a stretched, rather narrow belt. Evapotranspiration is reported to be generally low due to the low temperatures in autumn and winter. Precipitation results published in [12] are between 260 and 280 mm from the 1960´s to the 2000´s, and 310 mm for the 2010´s decade.

As to sand storms, the UNEP report concludes: “The Kubuqi Project area displays a consistent greening trend that could have caused a decrease in dust storms. This is supported by evidence from the meteorological records at Hangjin Qi which indicated that the number of sandstorm days per year decreased dramatically after the 1970s. Although the decreasing trend was evident before the Kubuqi Project started it has continued until now.”

Days of sandstorm per year until 1985 was between 10 and 50, and since then has decreased to 0 to 8 days.

Annual air temperatures recorded at Hangjin Qi station seems to follow the continental trend as given in [7].

The UNEP report identifies a risk typical for large scale afforestation in semi-arid areas: “While there is currently some risk of overuse of the water table, that is… mitigated by the fact that high water use species, such as non-native vegetables and trees, are only a portion of the developed area, the remainder being mostly plants native to desert areas.” The report recommends “a thorough assessment of water resources before extending to new areas so that the risk of water table depletion can be managed in terms of planting the appropriate species at suitable densities for the local hydrological conditions” [12]. In summary, in both example regions which may belong to the most advanced areas of the Chinese GGW, a reduction in sand storms and events of strong wind can be found following afforestation. In addition, for Jiangsu region an increase in air humidity and constant temperatures over the last 6 decades are found which on basis of global warming trend could indicate a slight cooling effect resulting from afforestation. This may be evidence of direct effects of afforestation on meso climatic level, leading to mitigation of dryness, heat, wind and sand storms in semi-arid and arid climate. Regional transformation from semi-arid to now moderately humid climate was reported.

Direct macro climatic effect of vegetation

During the last three decades, increased drought severity has led to loss of biomass in China, particularly around the year 2000 [7]. This trend clearly will have impacted the Chinese GGW afforestation efforts but plantations may have recovered since then. However, significant increase of forested area in Northern China also has been confirmed for other regions of the GGW. In a study published in 2013 the forested area in the district of Yulin (Shaanxi province) was analysed by mapping afforestation and deforestation from 1974 to 2012. Here, the forested area grew from 14.8% (380,394 hectares) in 1974 to 43.9% (1,128,380 hectares) in 2010. This was determined in a validated evaluation of time series stacks taken by Landsat satellite [13]. The semi-arid continental climate here has an average annual precipitation as low as 400 mm, falling mostly in the hot months of July and August.

In the last century sand and dust from the Gobi and Taklamakan deserts have been reported to be blown over thousands of km, leading to regular heavy air pollution in the capital of Beijing, and even causing coloration of rain and surfaces in Korea and Japan. These dust storm events, so called “Yellow dragon”, probably have been worsened in the last century by deforestations and over use of vegetation and ground water in the climatically sensitive semi-arid Northern territories of China, thereby leading to desertification of wide areas. A publication of Feng Wan et al. (2013) is showing that the frequency of sand storms of different strength in China indeed has gone down since 1954. According to this study, until 2010 the last strong sand storm in Beijing has been registered in 1995 [2]. The reduction of these events has been connected by local meteorologists to the large scale fixation of sand dunes and steppes of Northern and Northwest China. Evidence is given in a study (2015) of time trends in vegetation index in the GGW region, showing that, when compared with adjacent regions the GGW has improved the vegetation index and effectively reduced dust storm intensity (frequency, visibility, duration) in Northern China [14]. The Normalized Vegetation Difference Index (NDVI) is a measure of green vegetation cover from satellite imagery. For this parameter time trends were analyzed together with rainfall and dust storm data from weather stations. An index of dust storm intensity was deployed that takes frequency, visibility, and duration of dust storm events into account. The study found that NDVI was not related to rainfall trends, whereas dust storm intensity was decreased, resulting from increase in NDVI.

An investigation published by the same author in 2016 [15] has analysed air pollution data as generated by 186 observational stations across China. Average NDVI values in the 20-km radius of the 186 stations for six selected years in the period from 1983 to 2003 were analysed. Tan concludes that sand storms and dust storm intensity had decreased markedly during this time period in the area of the GGW and that in parallel the vegetation had recovered here. Thus, reduction in sand and dust storm events seems to be the first and most obvious climatic change introduced by afforestation. It is mentioned in all studies on the subject and noticed on regional as well as on continental level.

Discussion

The biophysical analysis of satellite derived data by Duveiller et al. [6] is showing that in dry climate, when compared to any other form of vegetation, forests have a cooling effect on surface temperature. This finding is surprising and it may necessitate a correction of the existing albedo simulation models for this climate zone. The review of simulation results in contrast suggests that afforestation of steppes and desert bordering areas may lead to a rise in surface temperature and risk of extreme hot temperatures in the same or neighbouring regions.

The West African simulation study [8] is mentioning uncertainty factors to be considered, e.g., choice of simulation model, and definition of extreme heat. At least for the vegetation found naturally in this climate it is fair to say that it would not show dark green colour throughout the year: Leaves and bark of sclerophylls often show bright wax cover, white „hair“, spines, prickles or thorns reflecting the sun light, thereby protecting them from UV radiation. In the typical savannah landscape trees are not standing very close and the grass in between will show light yellow colour during the hot dry season, i.e. for 7 to 9 months of the year. During this time fresh green leaves will dry out and fall off, they typically do not exist on the local species during most of the year. Therefore it seems doubtful whether the standard used here (“evergreen broad leafed plants“) can be applied to simulations of semi-arid conditions. Any existing or newly developed savannah vegetation cover that is adapted to this climate would probably not present dark green colour during dry season thus have a lower effect on albedo most time of the year.

With more and more simulations and investigations being undertaken to analyse the climatic impact of trees and forests scientists are now heavily discussing the overall net contribution of afforestation on global warming. In this situation it may help to search for afforestation related real life (semi-arid) climate data. The dilemma being that for such experiment we would need 1. about 50 years of time to establish tree vegetation cover in semi-arid area, 2. perfect baseline regional climate data and 3. to come to a reliable conclusion the temperature rise of global warming trend over these decades is to be taken into account. Afforestation in the Chinese example areas more or less is those 50 years ahead. By using long term climate data collected in one and the same meteorological station an interesting regulation mechanism was found that may not be as predominant in simulation models [10]. In the Jiangsu region, an actual tree coverage as high as 36% (starting from 3%) led to significantly reduced wind speed so that days of hot dry desert wind and extreme temperature in this region are almost gone. Where did the heat go, has it led to warming of the neighbouring regions?

We do not know, but certainly part of the energy will have gone into evapotranspiration of the newly developed forests. In the context of a GGW we need to ”think big“, we should find similar, maybe weaker cooling trends in other parts of the gigantic large Chinese GGW. The example described in above is a simple cause-and-effect relationship but it seems to have large meso climatic effect. It is questionable whether simulation would have shown that the vegetational impact on wind speed would by far overweight the impact of reduced albedo in the investigated region – and probably to some degree in all surrounding regions that have been enriched with vegetation. Albedo change and reduction in wind speed – these are only two of the factors of the complex regional interaction between vegetation, ground and climate. Other single factors that speak against warming effect of afforestation in dry hot climate which seem difficult to analyse or simulate are:

1. Shade: Shade reduces temperature, increases humidity, leading to constant soil moisture and thereby an increased uptake of rare precipitation. In this climate partial shadow may enable life, whereas absence of shadow does not. On a ground that is entirely dried out water can stay for long time without being taken up. During this time the majority of precipitation may already have run away in a wadi.
2. Root system: Deep root penetration of the ground will give it structure that allows for infiltration and long term storage of precipitation in deeper zones of the ground, leading to rise of the ground water level. Strong main roots will break up soil compaction in deep ground, fibre roots will increase the water holding capacity of desert sand… All of which will contribute to the water cycle, thereby increasing the cooling effect via evapotranspiration. In this climate zone water is of ultimate importance as it dominates life here in a “all or nothing rule”.

Can these and other biophysical single factors related to afforestation be simulated appropriately, with their specific intensity and meaning in semi-arid regions? Today, where would we get reference data to underlie a realistic simulation, even if these were given for an area of ”only“ 1,500 square km, like the Jiangsu region in China? When in this climate it takes at least 4, 5 or more decades until such „test area“ is established? Probably it is not a simple task to simulate an overall cooling effect of new vegetation that may evolve to its full extent in nature only after 5 to 7 decades, as can be derived from review of dry climate ecological data [16].

Often questions around publication of data of the Chinese GGW are raised: Where do data come from, what is the source? Analysts may be over motivated to sell a positive outcome of such huge project leading to biased reporting of results. As shown in the Kubuqi example, true facts (increase in precipitation) can be mixed with “half truth” (100 mm instead 260 mm initial value, 400 mm instead of 310 mm averaged value today) which is disappointing as it may mask any helpful interpretation based on true data. The UNEP (2015) report in this respect seems trustworthy, doing only interpretation of long term data, gathered from a local weather station.

Likewise, climate data from the Jiangsu region do meet these criteria. They are also underlined by positive agro economical facts on the development from steppe land in the 1950´s to farming land and fruit garden today. The authors even complain that, because harvests here are so rich farmers would now start to cut trees on the field borders in order to gain more arable land, they see a risk that very soon such behaviour could bring back the old days of hot sand storms.

The Northern Jiangsu region seems to be a “show case” or “model” region: Its´ proximity to the Yellow River is likely to have enabled irrigation of young plants, maybe there is a high water table as in the example of Kubuqi where the stretched narrow green belt was built along the banks of Yellow River. Here a high water table in a few meters depth is reported which clearly has made afforestation easier. Other regions of the Chinese GGW may not have this luxury. Their development into a state of 16, 20 or more percent tree cover when transition to sub-humid climate can be expected (as in Jiangsu) – this is likely to take more years here.

In this review we do not cover the question of which kind of species or vegetation to be chosen for which climate situation. Reports of the Chinese GGW make clear that also plantation of grassland and shrubs is part of the afforestation campaign. The high mortality rate of trees in this climate and a preferably lower water consumption rate of grassland are typical points of criticism of scientists [17]. Today China looks back on three generations of large scale planting efforts, and correction of some of the ecological parameters has been and will need to be done.

Any improvement of climate parameters may only develop over years and decades, hand in hand with the establishment of the new vegetation. Such improvement likely will be counteracted by the effects of global warming. It is therefore difficult to measure any balancing climate effect of new vegetation since these two activities minimize each other. In addition, the larger the region being analysed the more difficult it seems to relate any change in climate parameters to afforestation activity. Long term investigations over the next decades will show whether a still growing new vegetation cover in semi-arid Northern China, besides sand storm events also will modify temperature and humidity on a continental level, similar to the reports on regional level.

Conclusion

It is surprising that we may be able to already find meso and macro climatic effects of vegetation only 50 years after the first plantations were started. In comparison to afforestation in humid climate, newly planted vegetation in semi-arid climate is expected to need significantly more time to take root and get established. Based on semi-arid ecological observations it may take up to 70 years until new vegetation in steppes is well established [16] and only then would also have developed its´ full climatic potential for increased deposition of precipitation in the ground, for the activation of hydrate cycle, formation of clouds, reduction of wind speed, and stabilized surface temperatures. Similarly, a recent report from a Chinese science journalist indicates that it is expected to take another 20 years until we will see the full spectrum of positive results from the Chinese GGW that was started in 1978 (“France 24”, online news 2018). We need to „think big“ in terms of geography – and time. The ability of such plantation to develop, spread and expand further, this certainly can be used as a measure of persistence and success of semi-arid afforestation.

Striking discrepancy was found between the theoretical impact of albedo changes outweighing afforestation results in simulation studies when compared to the importance of wind breaking activity of vegetation on the border of deserts, as seen in real life. The difficult afforestation of steppes and desert border regions may be of high value, functioning as a „vegetational climate barrier“, in addition to the climate protective effect via CO₂ fixation. Here the desert climate parameters are being controlled, vegetation here is buffering the climatic impact of deserts on their adjacent regions.

The first results from Jiangsu region with a size of about 30 x 50 km are showing a threshold with 16 to 20 % of minimum tree cover that is leading to beneficial regional climate changes, i.e. an increase in humidity that is enabling agricultural production in an area dominated by hostile semi-arid climate before. Especially in semi-arid and arid climate it seems obvious that vegetation should have a minimum density over a larger area, a certain minimum percentage in order to show climatic impact and to support or enable agriculture via humidity and precipitation induced by the additional vegetation.

Current natural climate solution projects are focused on maximum fixation of CO₂ amounts, consequently plantation projects were supported preferably in regions where high amount of CO₂ can be sequestered within short time, i.e. where fast growth of trees is supported by humidity of the local climate. In the semi-arid climate of desert border regions however, viability of vegetation depends on a certain regularity of precipitation, additional vegetation may create a duplicate climate mitigating effect, leading to additional humidity and reduced surface temperatures on formerly bare ground.

What if many or most of the desert bordering regions and semi-arid areas with signs of desertification, globally are considered as GGW candidates, getting re-greened in order to maintain soil fertility and a balanced regional and continental climate? Regions in question for respective activities are the Sahel, South Africa, the entire region from Syria to Pakistan, parts of India, and Australia. GGWs and networks of existing and new vegetation in desert bordering areas may be stabilizing in many regards, leading to a more balanced climate regionally and perhaps, globally – in addition to the benefit for agriculture and economy.

Acknowledgement

I would like to thank Professor Dr. Klaus Becker, University of Hohenheim, Germany for all helpful feedback and discussion of the topic.

References

1. Sven Ploeger, Frank Boettcher: Klimafakten“, Westend Verlag GmbH, Frankfurt (2013). ISBN 978-3-86489-048-2. Pg No: 124–125.
2. China’s ‘Great Green Wall’ Fights Expanding Desert by Alexandra Petri E (2017) National Geographic, Apr 21.
3. Feng Wan, Xubin Pan, Dongfang Wang, ChongyangShen, Qi Lu (2013) Combating desertification in China: Past, present and future“. Land Use Policy 31: 311–313.
4. World leaders renew commitment to strengthen climate resilience through Africa’s Great Green Wall“. Article based on The African Union Commission Press Release, Paris, France, 02 December 2015. In: The African Union, Link: http://www.au.int/en/
5. Eduardo Mansur in Great Green Wall’ initiative offers unique opportunity to combat climate change in Africa. UN agency. 17 November 2016. Link: http://www.un.org/sustainabledevelopment/blog/2016/11/great-green-wall-Ca-un-agency/
6. Gregory Duveiller, Giovanni Forzieri, Eddy Robertson, Wei Li, Goran Georgievski, et al. (2018) Biophysics and vegetation cover change: a process-based evaluation framework for confronting land surface models with satellite observations. Earth Syst Sci Data 10: 1265-1279, https://doi.org/10.5194/essd-10-1265-2018
7. WAD – World Atlas of Desertification, European Commission, Joint Research Center Update 21.11.2018, Link: https://wad.jrc.ec.europa.eu/
8. Odoulami RC, Abiodun BJ, Ajayi AE, Diasso UJ, Saley MM (2017) Potential impacts of forestation on heatwaves over West Africa in the future. Ecological Engineering 102: 546–556.
9. Myra Naik, Babatunde J (2016) Potential impacts of forestation on future climate change in Southern Africa”. Int. Journal of Climatology 36: 4560–4576.
10. Jia-Yao Zhuang, Jin-Chi Zhang, Yangrong Yang, Bo Zhang, Juanjuan Li (2017) Effect of forest shelter-belt as a regional climate improver along the old course of the Yellow River, China. AgroforestSyst 91: 393–401.
11. Li Yang: Kubuqi a successful example of desert greening.China Daily, Updated: 2018-08-06, 07:39h.
12. UNEP: Review of the Kubuqi Ecological Restoration Project: A Desert Green Economy Pilot Initiative.” 2015. United Nations Environment Programme, Nairobi.
13. Liangyun Liu,Huan Tang, Peter Caccetta, Eric A. Lehmann, Yong Hu,Xiaoliang Wu (2013) Mapping afforestation and deforestation from 1974 to 2012 using Landsat time-series stacks in Yulin District, a key region of the Three-North Shelter region, China Environ Monit Assess. 185: 9949–9965.
14. Minghong Tan, Xiubin Li (2015) Does the Green Great Wall effectively decrease dust storm intensity in China? A study based on NOAA NDVI and weather station data. Land Use Policy 43: 42–47.
15. Minghong Tan (2016) Exploring the relationship between vegetation and dust-storm intensity (DSI) in China. Journal of Geographical Science 26: 387–396.
16. Lorenz Huebner (2019) Der Gruene Rettungsring. MitvernetzterSteppenbegruenung global der Klimakrisebegegnen”.OekomVerlag, Munich, Germany.
17. Alexandra E. Petri (2017) China’s ‘Great Green Wall’ Fights Expanding Desert”. National Geographic Link: https://news.nationalgeographic.com/ 2017/04/china-great-green-wall-gobi-tengger desertification/