Feedbacks on zonal mean tropical precipitation shifts induced by land 12 surface change

Changes in land surface albedo and land surface evaporation modulate the atmospheric energy budget by changing temperatures, water vapor, clouds, snow and ice cover, and the partitioning of surface energy fluxes. Here idealized perturbations to land surface properties are imposed in a global model to understand how such forcings drive shifts in zonal mean atmospheric energy transport and zonal mean tropical precipitation. For a uniform decrease in global land albedo, the albedo forcing and a positive water vapour feedback contribute roughly equally to increased energy absorption at the top of the atmosphere (TOA), while radiative changes due to the temperature and cloud cover response provide a negative feedback and energy loss at TOA. Decreasing land albedo causes a northwards shift in the zonal mean intertropical convergence zone (ITCZ). The combined effects on ITCZ location of all atmospheric feedbacks roughly cancel for the albedo forcing; the total ITCZ shift is comparable to that predicted for the albedo forcing alone. For an imposed increase in evaporative resistance that reduces land evaporation, low cloud cover decreases in the northern mid-latitudes and more energy is absorbed at TOA there; longwave loss due to warming provides a negative feedback on the TOA energy balance and ITCZ shift. Imposed changes in land albedo and evaporative resistance modulate fundamentally different aspects of the surface energy budget. However, the pattern of TOA radiation changes due to the water vapour and air temperature responses are highly correlated for these two forcings because both forcings lead to near-surface warming. 28


Introduction
We meridionally integrate T OA net , under the assumption that atmospheric energy storage is negligible on annual time scales, to calculate cross-equatorial atmospheric energy transport AET eq , 188 and estimate the linear relationship between AET eq and the zonal-mean location of the ITCZ. 189 We measure the zonal-mean ITCZ location as the latitude φ p that is the center of mass of the 190 precipitation distribution between 20 • S-20 • N. Using the individual contribution to ∆T OA net from 191 each surface or atmospheric process resulting from the imposed change in land surface property 192 (e.g. the change in albedo from changes in snow/ice, or the change in water vapour), we determine 193 the ∆AET eq that would result from that individual component of the T OA net response alone. We 194 then leverage the derived relationship between AET eq and φ p to attribute portions of the total 195 modelled shift in the ITCZ to each individual atmospheric and surface process. The practice of 196 meridionally integrating anomalous TOA energy sources to obtain an AET eq change and then an 197 ITCZ shift follows Kang et al. (2008), and using this procedure to estimate radiative feedbacks 198 follows Peterson and Boos (2020). 199 We follow the methodology of Soden et al. (2008) and Pendergrass et al. (2018) to decompose 200 the response of TOA radiation into components associated with changes in imposed land surface evaporative resistance modifies the TOA energy budget mostly in the northern mid-to-high lati-210 tudes during boreal summer. Decreasing land albedo and increasing land evaporative resistance 211 both lead to overall more energy absorbed at the TOA over the Northern Hemisphere, though for 212 different reasons which are explored below. 213 The land albedo and evaporative resistance changes also produce changes in precipitation over The spatially uniform decrease in snow-free land albedo has a spatially non-uniform impact 226 on T OA net . Darkening land results in more SW being absorbed by Earth over most land areas, 227 while over oceans and parts of the northern high-latitudes, more energy is lost by the Earth system 228 (figure 1a). The peak anomalous energy gain resulting from the decreased land albedo is found in 229 the tropics in the annual mean, with smaller increases in the mid-latitudes. 230 To understand the mechanisms through which a spatially uniform change in land surface albedo 231 causes a spatially non-homogeneous and non-local change in TOA radiation, we decompose the 232 response into a forcing and several feedbacks, each of which impact the TOA flux of shortwave SW assuming temperatures, water vapour, snow and ice cover, and cloud cover do not change.

242
The imposed decrease in land surface albedo causes an increase in net TOA SW radiation over all 243 non-glaciated land areas (that is, everywhere the albedo was directly changed; figure 3a). Within 244 snow-free land regions, the spatial pattern in the change in TOA SW radiation comes predomi-245 nantly from the spatial pattern of the radiative kernel itself, which reflects the pattern of insolation, 246 cloudiness, and clear-sky optical depth (figure S1). From the kernel, we see that the increase in 247 absorbed TOA SW for a spatially uniform decrease in land albedo is largest in low latitudes, where 248 incident solar radiation is highest and the annual mean atmospheric path length for downwelling 249 shortwave is smallest. The same albedo change imposed on regions with climatologically high 250 cloud cover (e.g. the Maritime Continent) has a smaller impact on TOA SW than regions at a 251 similar latitude with less cloud cover, as less SW reaches the surface in those regions. The direct 252 forcing of the imposed albedo change is calculated here specifically for snow-free albedo, i.e. how 253 the TOA SW would be affected in the absence of snow. However, land surface albedo in higher 254 latitudes is masked by snow for part of the year; the change in TOA radiation because of changes 255 in snow and ice is captured in the albedo feedback term discussed next.

256
2) ALBEDO FEEDBACK 257 We define albedo feedbacks as changes in TOA SW radiation due to changes in snow and ice 258 cover, which themselves result from changes to the climate system driven by our imposed change

267
Decreased land surface albedo can modify atmospheric water vapour both by modulating evapo-268 ration from the land surface and by modulating the winds that transport water vapour. Decreasing 269 land albedo leads to more water vapour over tropical land in our model, with atmospheric tem-270 peratures and specific humidities both generally increasing over land. There is also a meridional 271 dipole pattern in precipitable water over tropical oceans reflecting a northwards shift in the ITCZ 272 and a change in the humidity of the subtropical dry zones (figure S4). In idealized aquaplanet 273 models, the relative humidity of the subtropical dry zones increases in the hemisphere in which a 274 positive energy source is imposed and decreases in the subtropical dry zones on the other side of 275 the equator, amplifying the more traditional fixed-relative humidity water vapor feedback (Peter-276 son and Boos 2020); this also seems to occur in our model in response to land albedo changes. The 277 only statistically significant changes in SW at the TOA due to water vapour changes in response to 278 decrease land albedo occur over the Sahara and Arabian Peninsula, where the response is positive 279 (i.e. more SW absorbed by the enhanced water content; figure 3c). The LW effects of water vapour 280 changes are also positive, but are much more far reaching, spreading over most land and ocean 281 regions of the NH (figure 3d). Averaged globally, the LW effects of changes in atmospheric water 282 vapour are as large as the direct effect of both the albedo forcing and ice-albedo feedback on TOA 283 SW , with both contributing an extra 2 W/m 2 of energy to the Earth system at the TOA (table 1).

285
Temperature feedbacks are changes in TOA LW due to changes in surface temperature, T s , and 286 temperatures through the atmospheric column. These combine the Planck and lapse rate feed-287 backs, with the latter typically having a magnitude that is about one-third that of the former in 288 the global mean (Soden and Held 2006). Using the radiative kernel for temperature, we see that 289 temperature feedbacks produce an increase in outgoing LW that opposes the SW forcing, as ex-290 pected for negative feedbacks. Changes in T s drive an increase in outgoing LW mostly over NH  (table 1). This is expected for the negative Planck 296 and lapse rate feedbacks, which balance the sum of the forcing and the positive water vapor and 297 albedo feedbacks to achieve TOA energy balance in the new steady state. Unlike decreasing land albedo, which causes more SW energy to be absorbed by land, chang-317 ing the evaporative resistance of land does not directly modify the total energy absorbed by land.

318
Increasing evaporative resistance drives a repartitioning of surface energy fluxes, where energy 319 previously used to evaporate water is instead partitioned into sensible heat flux or emitted long-320 wave radiation, both of which result from the increase in surface temperature that is driven by the 321 reduced evaporative cooling. Changes in evaporative resistance can only modify latent heat flux 322 from the surface to the atmosphere in regions where there is water stored on the land surface; there 323 is little to no effect of changing this surface property over desert regions.

324
Here we discuss the net response to the evaporative resistance forcing, and briefly summarize all 325 of the individual components of that response. In contrast to the response of T OA net to decreasing 326 land albedo, increasing the evaporative resistance of land results in an increase in T OA net that 327 is strongest in the northern mid-latitudes during June-August (figure 1b, d). As stated above, 328 changing the evaporative resistance of land has no direct impact on the total energy absorbed by 329 land, so there is no "forcing" in the context used for the albedo simulations. However, we can still 330 decompose changes in the TOA energy budget into components due to snow/ice changes, water 331 vapour, temperatures, and clouds.

332
Increasing the evaporative resistance of land leads to warming by suppressing latent cooling 333 of the land surface, which causes a reduction of snow and sea-ice ( figure S3). This reduces the 334 surface albedo and leads to an increase in absorbed SW at the TOA, mostly in the northern high 335 latitudes during boreal summer (figure 1d, 4b; note the change in color scale in figure 4). There are 336 no statistically significant changes in TOA SW due to changes in atmospheric water vapour, while 337 the LW effects of water vapour changes lead to a slight increase in energy absorbed by Earth at 338 the TOA over parts of the low latitude ocean (figure 4c, d). We note that total column water vapor 339 actually increases over most of the Northern Hemisphere, which has the largest land area (figure 340 4b). That is, increased land resistance leads to decreased land evaporation and less low cloud 341 cover, which drives warming which itself results in more atmospheric water vapor, particularly 342 over the oceans, resulting from suppressed terrestrial evaporation. Increased surface temperatures 343 in the Arctic lead to more TOA LW loss, while atmospheric warming in the northern mid-to 344 high-latitudes also increases TOA LW loss (figure 4e).

345
The largest change to TOA radiation as a result of increasing the evaporative resistance of land 346 comes from the SW effects of changes in cloud cover (figure 4f,g). Loss of cloud cover over 347 southeastern North America and western Eurasia results in an increase in SW absorption by Earth.

348
This signal is strongest during NH summer, but persists with weaker magnitude over southeastern cover. The global mean SW cloud forcing is more than twice as large as the global mean cloud 362 feedback for the albedo forcing (-1.2 W/m2 vs -0.5 W/m2). However, the SW cloud forcing and 363 SW cloud feedback are very similar for an increase in land evaporative resistance, because in that 364 case nearly all of the change in SW cloud forcing comes directly from a change in cloud cover.

365
The same physical process is thus captured by the SW cloud feedback and SW cloud forcing for 366 changes in evaporative resistance (figure S7). 2). This is because both the water vapour and temperature components of the TOA energy budget 379 decomposition are directly related to warming, and both decreasing the land surface albedo and 380 increasing land surface evaporative resistance lead to large-scale warming of the Earth system.

381
The mechanisms responsible for the surface warming are different; in the case of albedo, warming 382 is the direct result of increased SW absorption at the surface, while in the case of evaporative re-383 sistance warming is the result of suppressed evaporative cooling and increased SW absorption due 384 to regional loss of cloud cover. However, in both cases, warming at the surface is accompanied by 385 warming aloft and an increase in atmospheric water vapour over large parts of the northern hemi-386 sphere remote from the forcings (figure S8), presumably due to homogenization of atmospheric  2).

404
The relationship between annual mean cross-equatorial atmospheric energy transport and the 405 zonal mean ITCZ latitude φ p is strongly linear in our simulations (figure 5). We find a -4.4 • shift 406 in the ITCZ per 1 PW increase in annual mean northwards cross-equatorial atmospheric energy 407 transport (figure 5). This slope is slightly larger in magnitude than that found by Donohoe et al.  The relationship between the zonal mean ITCZ location, φ p , and cross-equatorial atmospheric energy transport, AET eq , in response to perturbed land surface properties is also tightly correlated 412 during Northern Hemisphere summer (figure 5a, c). However, we wish to decompose the ITCZ 413 shift into components associated with individual feedbacks (e.g. water vapor and Planck feed-414 backs), which requires meridionally integrating the anomalous TOA energy flux due to each feed-415 back to obtain its contribution to the net cross-equatorial energy transport (e.g. Kang et al. 2008;416 Peterson and Boos 2020); this can only be done exactly in the annual mean, when the transient 417 atmospheric storage term is zero in a steady state climate. In order to leverage our decomposition 418 of the TOA energy budget, we thus focus our analysis of shifts in the ITCZ on the annual mean.

419
For each component of the TOA energy budget response to changes in land surface albedo and 420 evaporative resistance, we calculate the anomalous cross-equatorial energy flux needed to bal-421 ance the specific pattern and magnitude of TOA SW and LW change comprising that component.  Decreasing land albedo drives a northwards shift in the ITCZ as a result of the direct effect of 431 the imposed change in albedo, with positive (northward) contributions from the albedo feedback 432 due to changes in snow and ice, the SW and LW water vapour feedbacks, and the LW cloud 433 feedback (figure 6). It is notable that the LW cloud effects provide a negative feedback on the 434 global mean TOA energy balance response to the albedo forcing (Table 1)  is a contribution to a northward ITCZ shift from loss of high-latitude snow and ice resulting from 468 warming, this contribution is smaller than the contributions from temperatures, water vapour, and 469 SW cloud feedbacks, and is not statistically significant.

470
The ITCZ shift predicted by the sum of the feedbacks is larger than the modelled ITCZ shift, 471 more so for evaporative resistance than for albedo (light gray bars in figure 6). This disagreement 472 is the result of the linear fit used to predict the ITCZ shift associated with a given change in cross-   The idealized nature of these simulations necessarily presents some limitations. The perturba-518 tions made to land surface albedo and evaporative resistance were applied to all non-glaciated land 519 surfaces, and as such the hemispheric imbalance in response to these land surface perturbations is 520 largely a result of the hemispherically asymmetric distribution of the continents in their present-521 day configuration; other patterns of land surface change would yield their own specific patterns 522 of TOA energy flux changes and individual forcing/feedback terms. The radiative kernel we use 523 to decompose the TOA energy budget response into its components was generated with the same 524 atmospheric model as we use in this study (CAM5). However, any differences in the mean state of 525 atmospheric temperatures, humidity, and cloud cover between the CLM-CAM5 simulation used 526 for the kernels and the baseline SLIM-CAM5 simulation used in this study could introduce errors 527 in the kernel-predicted change in TOA radiation. Furthermore, because we do not have an explicit 528 radiative kernel for cloud fraction, any residuals that may exist in our calculations are lumped in 529 with the impact of clouds on TOA SW and LW , by virtue of the methods we use to decompose the TOA energy balance (see Appendix). However, we expect these residuals to be small for two  Despite these caveats, the method we present here allows us to understand the mechanisms 546 through which changes in the land surface drive changes in zonal mean atmospheric circulation and 547 tropical precipitation. Understanding these mechanisms is critical to understanding how changes 548 in the land surface-both historical and in the future-impact climate locally and globally. The data presented in this paper will be archived on Dryad and the link added here upon accep-551 tance of this manuscript. The source code for the models used in this study are publicly available 552 on github at https://escomp.github.io/CESM/release-cesm2/downloading_cesm.html 553 for CESM, and https://github.com/marysa/SimpleLand for SLIM.

554
At the TOA, the energy balance is between incoming shortwave (SW ) radiation, reflected SW 571 radiation, and outgoing longwave radiation (LW ). At the surface, the balance is between the net 572 flux of SW and LW radiation, and the turblent fluxes of sensible heat (SH) and latent heat (LH).

573
The sign convention in equations A1-A2 is such that SFC net > 0 means more energy absorbed 574 by the surface (land or ocean). More energy is absorbed by the Earth system in regions with 575 T OA net > 0, while more energy is lost by the Earth system in regions with T OA net < 0. On land 576 over sufficiently long timescales (e.g. the annual mean, which we examine here), the surface where v is the meridional wind and h is the moist static energy. vh is calculated from the heat  We use the following notation when referring to calculations using the radiative kernel. The 632 change in net TOA SW as a result of a 1% change in surface albedo is given by K α . The change 633 in net TOA LW resulting from a 1K increase in surface temperature is given by K T s . The change 634 in TOA LW resulting from a 1K increase in air temperature vertically through the atmosphere is 635 given by K T . The change in TOA SW and LW resulting from the imposed change in water vapour 636 are given by K q,SW and K q,LW , respectively.

637
We impose a change in snow-free albedo ∆α i on the land surface. Using ∆α i , we can quantify 638 the change in top of atmosphere SW radiation directly attributable to the imposed change in surface 639 albedo ∆SW α i (equation A6), where ∆α i is multiplied by 100 to convert it to a percent value.
The total modeled change in albedo includes both our imposed snow-free change in albedo as 641 well as albedo changes due to snow and ice responses. We can calculate the change in albedo due to where δ T is the modelled change in air temperature and δ q is 661 the change in specific humidity that would have resulted from ∆T given constant relative humidity.

662
Then, we can use K * q to determine the change in TOA SW and LW attributable to the modelled 663 change in specific humidity ∆q (equations A10-A11).
To determine the effect of changes in cloud cover on T OA net , we do not use a radiative kernel 666 for cloud cover. Rather, we determine how much the modelled change in cloud fraction impacts 667 SW and LW at the TOA, by calculating the total modelled response of T OA net then subtract the 668 change in T OA net due to the combined effects of albedo, temperature, and water vapour (equations 669 A12-A13).
Because we do not diagnose ∆LW cloud or ∆SW cloud directly from a cloud kernel, the ∆LW cloud or 672 ∆SW cloud terms necessarily also include any potential residual terms associated with the kernel.

673
That is, if the actual direct response of TOA SW to ∆α i in our simulations differs from the ∆SW α i 674 predicted by K α because, for example, the mean state of cloud cover in our SLIM-CAM5 sim-675 ulations differs substantially from the mean state of cloud cover in the CLM-CAM5 model, that 676 difference would necessarily be included in the ∆SW cloud and ∆LW cloud terms here.

677
We also consider changes in the shortwave cloud forcing (SWCF) and longwave cloud forcing 678 (LWCF). This is a different quantity than ∆SW cloud and ∆LW cloud (see, for example, figure 6. The breakdown of the change in the zonally averaged annual mean location the ITCZ (measured by φ p ) resulting from each component, re-scaled to a 1 • total northwards shift. Solid (hatched) bars show the change in the zonal mean ITCZ location for a uniform decrease of land surface albedo (increase of evaporative resistance). From left to right, bars show: the total modelled change (dark grey); the change due to the sum of all of the individual components (light gray); the change attributable to the imposed change in albedo (oragne), the change in albedo due to changes in snow and ice (yellow), LW effects due to changes in surface temperature (dark purple), LW effects to due vertical changes in the atmospheric temperature profile (lilac), SW changes due to changes in water vapour (light green), LW changes due to changes in water vapour (dark green), SW changes due to changes in cloud cover (light blue), and LW changes due to changes in cloud cover (dark blue). The magnitude of the ITCZ shift is noted above each bar, as well as the p value taken from a students' t-test, where p < 0.05 indicates a significant shift from the baseline simulation.