The contribution of transpiration , ground evaporation , and canopy 1 evaporation to local and remote precipitation across North America 2 3

Abstract. Land surface evapotranspiration (ET) is a major source of moisture for the global hydrologic cycle. Though the influence of the land surface is well documented, moisture tracking analysis has often relied on offline tracking approaches that require simplifying assumptions and can bias results. Additionally, the contribution of the ET components (transpiration (T), canopy evaporation (C), and ground evaporation (E)) individually to precipitation is not well understood, inhibiting our understanding of moisture teleconnections in both the current and future climate. Here we use the Community Earth System Model version 1.2 with online numerical water tracers to examine the contribution of local and remote land surface ET, including the contribution from each individual ET component, to precipitation across North America. We find the role of the land surface and the individual ET components varies considerably across the continent and across seasons. Much of northern and northeastern North America receives up to 80% of summertime precipitation from land surface ET, and over 50 % of that moisture originates from transpiration alone. Local moisture recycling constitutes an essential source of precipitation across much of the southern and western regions of North America, while remote land surface moisture supplies most of the land-based precipitation across northern and eastern North America. Though the greatest contribution of remotely sourced land ET occurs in the north and east, we find the primary sources of North American land surface moisture shifts seasonally. The results highlight regions that are especially sensitive to land cover and hydrologic changes in local and upwind areas, providing key insights for drought prediction and water resource management.



12
Abstract. Land surface evapotranspiration (ET) is a major source of moisture for the global 13 hydrologic cycle. Though the influence of the land surface is well documented, moisture 14 tracking analysis has often relied on offline tracking approaches that require simplifying 15 assumptions and can bias results. Additionally, the contribution of the ET components 16 (transpiration (T), canopy evaporation (C), and ground evaporation (E)) individually to 17 precipitation is not well understood, inhibiting our understanding of moisture teleconnections 18 in both the current and future climate. Here we use the Community Earth System Model 19 version 1.2 with online numerical water tracers to examine the contribution of local and 20 remote land surface ET, including the contribution from each individual ET component, to 21 precipitation across North America. We find the role of the land surface and the individual 22 ET components varies considerably across the continent and across seasons. Much of 23 northern and northeastern North America receives up to 80% of summertime precipitation 24 from land surface ET, and over 50% of that moisture originates from transpiration alone. 25 Local moisture recycling constitutes an essential source of precipitation across much of the 26 southern and western regions of North America, while remote land surface moisture supplies 27 most of the land-based precipitation across northern and eastern North America. Though the 28 greatest contribution of remotely sourced land ET occurs in the north and east, we find the 29 primary sources of North American land surface moisture shifts seasonally. The results 30 highlight regions that are especially sensitive to land cover and hydrologic changes in local 31 and upwind areas, providing key insights for drought prediction and water resource 32 management. 33 34 35 1. Introduction: Evapotranspiration (ET) from the land surface is a major atmospheric moisture source 38 responsible for approximately 35% of precipitation over land (van der Ent et al., 2014). This 39 terrestrial-sourced moisture is supplied by advection of ET from upwind land surfaces 40 (moisture transport) and by ET from within the land region of interest (moisture recycling). 41 Regions that rely heavily on terrestrial ET moisture for precipitation are susceptible to 42 mechanisms that alter land surface ET such as land use and land cover change (Weng et al., that can mitigate or exacerbate dry and wet periods, including high-impact drought and flood 46 events (Dirmeyer and Brubaker, 1999;Seneviratne et al. 2010; Kelemen et al., 2016). It is 47 therefore critical to understand the breakdown of a region's terrestrial moisture sources, 48 including the contributions from moisture transport and recycling, and the contributions from 49 different land cover types, such as vegetation and soil, for water resource management. 50 Previous studies have identified terrestrial ET as an important source of moisture for 51 North American precipitation. For example, van der Ent et al. (2010) used reanalysis data 52 and an atmospheric moisture budget-based accounting model to quantify a continental 53 precipitation recycling ratio and found that terrestrial ET is responsible for approximately 54 40% of annual precipitation across most of the North American continent (van der Ent et al., 55 2010). The authors also quantified the continental evaporation recycling ratio and found that 56 nearly 60% of evaporated moisture from the western half of North America falls as 57 precipitation over land (van der Ent et al., 2010). Both the continental precipitation recycling 58 ratio and the continental evaporation recycling ratio are projected to decrease across North 59 America with future warming, highlighting the sensitivity of moisture source-sink 60 relationships to varying environmental conditions, and the need to understand the underlying 61 processes that influence these relationships to confidently predict future water availability 62 ). 63 While the continental recycling ratios point to a key role for the land surface as a 64 whole in shaping North American precipitation, within this continental-scale framework, all 65 precipitation sourced from the land surface is considered "recycled" even though it might 66 have evaporated thousands of kilometers upwind. A regional-scale recycling framework is 67 necessary to identify the contribution of proximate and remote terrestrial sources to 68 precipitation. Though the definition of "regional scale" within the context of moisture 69 recycling is subjective and influences the quantity of recycled precipitation (Dirmeyer and 70 Brubaker previous work has found that changes to crop management practices, such as a reduction in 76 irrigation in California's Central Valley, would reduce precipitation in California and the 77 surrounding Southwest U.S. (Lo and Famiglietti, 2013). Changes to land use and agricultural 78 practices in the Central Valley should therefore be considered carefully to limit potential 79 adverse regional hydroclimate impacts. 80 In addition to knowing the geographic sources of a region's precipitation, identification 81 of the surfaces from which that terrestrial ET is sourced can further enhance understanding of 82 regional hydroclimate and assist water resource management. Land surface ET is a 83 combination of transpiration (T), canopy evaporation (C), and ground evaporation (E). The T 84 component alone accounts for nearly 64% of total global ET (Good et al., 2015), making 85 vegetation critical for land moisture recycling. However, the degree to which a region relies 86 on T for precipitation depends on proximity to dense, high-transpiration plants, and/or 87 alignment to the prevailing winds that flow over those plants (Keys et al. 2016). Regions that 88 rely more heavily on T may exhibit less variability in precipitation than regions that rely 89 largely on C or E, as plants with deep rooting systems are able to tap into water deep below 90 ground providing moisture to the atmosphere even during relatively dry periods (Tueling et components' role in precipitation and local recycling individually is necessary for a complete 95 understanding of the hydrologic cycle. 96 Additionally, there is much uncertainty regarding the partitioning of ET in future 97 climates (Kirschbaum, 2004 Here we present a new unified framework to study regional moisture recycling and the 105 sourcing of precipitation into its individual ET components (T, E, and C). Using a climate 106 model with water tracing capability, we identify the reliance of each location in North 107 America on precipitation sourced from transpiration, canopy evaporation, and soil/lake 108 surface evaporation, highlighting regions most susceptible to changes in vegetation type, 109 coverage, and physiology. We then estimate precipitation recycling on a regional scale, 110 providing an assessment of precipitation sensitivity to local and remote land use/land cover 111 changes (whether land management-or climate-driven). Lastly, we combine the water tracing 112 analyses to quantify the breakdown of local and remote terrestrial moisture recycling into the 113 T, E, and C components.  where and are vectors of domain-scaled P and E of size n, I is the identity matrix, is 216 the domain-scaled export matrix with diagonal entries " and all non-diagonal entries equal to 217 0, and is the convergence matrix with non-diagonal entries equal to !" and all diagonal 218 entries equal to 0. 219 220 221  and is an appropriate model to study North American precipitation characteristics.  ( Figure 3c). Though the maximum grid cell error occurs during the spring season, the 297 average of the absolute value of the errors continent wide peaks in the summer (0.14 mm 298 day^-1) compared to the winter (0.05 mm^day-1), spring (0.12 mm day^-1), and fall (0.08 299 mm day^-1) seasons. One potential source of error in climate model ET estimates is directly 300 related to errors in the modeled precipitation (Mueller & Seneviratne, 2013). If precipitation 301

Comparison of iCESM Precipitation to GPCP Observations
is overestimated (underestimated) in the model, more (less) moisture is available for ET. 302 Though the potential for this bias exists across the continent given the errors in modeled other error sources are responsible for the disagreement between ET in the model and 309 observations. Unlike total ET, the error in the individual ET components (T, C, and E) are 310 seasonally consistent. Continent wide, iCESM generally underestimates T and overestimates 311 C and E (Supplemental Figures 1-3). This behavior is present across CMIP5 models and is 312 attributed to numerous potential land surface models errors including errors in leaf area index 313 (LAI), interception loss, root water uptake, light-use efficiency (LUE), and water-use

Seasonal Land-Based Precipitation Signals
The average percent contribution of the global land surface to total precipitation over 345 North America follows a seasonal cycle with the largest contributions in the summer and the 346 smallest contributions in the winter. This is consistent with the reanalysis-based findings of  The land surface contributes less moisture for precipitation during the winter season than 374 in any other season ( Figure 4a). However, moisture recycling, defined as precipitation in a region 375 that is sourced from ET within that region, still contributes between 2-4% of total precipitation 376 across much of the southern United States, and over 4% in parts of Central America (Figure 5a, 377 Supplemental Table 1a). Some areas near the Great Lakes also receive between 2-3% of their 378 total precipitation from local moisture recycling, though this is likely from the Great Lakes 379 themselves since lake evaporation is part of the E variable. While the contribution of local recycling to total precipitation is greatest in the south, local recycling comprises a considerable 381 fraction of total winter land-based precipitation (total precipitation that is sourced only from the 382 land surface) along the US west coast (Figure 6a, Supplemental where L is the percent recycling, a is the land area of each region, and ‖ ‖ represents the 394 Frobenius norm. This scaling ensures that regions with the largest land area are scaled down so 395 the percent recycling is not solely a function of domain size. Both the PNW and SWC regions as 396 well as the SMM region continue to have very high rates of local recycling when analyzing land-397 based precipitation only, even when considering the normalized values ( Figure 6b, Supplemental 398 Table 1c). Overall, results indicate that local moisture recycling is an important source of 399 precipitation for much of Central America and the SCP region during the winter season. 400

Spring Season Recycling 409 410
The maximum contribution of the land surface to total precipitation shifts to much of 411 Canada  During the summer, the land surface is a major source of moisture for much of the North 438 American continent (Figure 4c). The contribution of local moisture recycling to total 439 precipitation also increases for the entire continent during this season (Figure 5e, Supplemental 440 Table 3a). The highest contributions of local recycling occur in the western US with 27% of 441 UPR and 22% of SWW precipitation coming from recycled moisture. Across the central and 442 southern US, local recycling accounts for 12% to 17% of total summertime precipitation. 443 Recycled moisture is also an important moisture source for WIP, EIP, and NCA contributing 21, The spatial pattern of land-surface contribution to precipitation is very similar between the 460 spring and fall seasons, though the magnitude of the contribution is slightly lower in the fall 461 (Figures 4b, 4d). Similar to the land-surface contribution, the contribution of local recycling to 462 precipitation is also spatially similar between the fall and the spring (Figures 5c, 5g, 463 Supplemental Tables 2a, 4a

Divergence and Convergence of North American Land Moisture
Land surface evaporation that does not precipitate locally is exported out of its evaporative 480 source region and is available for precipitation elsewhere. Equation (5)  The predicted precipitation using Equation (5) captures both the spatial variability and the 492 magnitude of CESM-simulated North American land-based precipitation (Supplemental Figures  493   1-2). The spatial patterns of (from hereon called the divergence term) and (from 494 hereon called the convergence term) indicate seasonal shifts in the key land moisture source and 495 sink regions ( Figure 10). During the winter season, the magnitude of the divergence term is 496 highest in the southern US and Central America, aligning closely with the evaporation fields 497 (Figures 10a and 10b). Though higher evaporation totals potentially allow for higher amounts of 498 moisture divergence, our framework developed in Section 2.2 only considers moisture 499 divergence that later converges within North America. Differences between the evaporation 500 fields and the divergence fields are attributed either to high amounts of internal moisture 501 recycling or to atmospheric circulation features that may export evaporation from some land 502 regions out of the North American domain. The evaporation is highest in the SMM region, but 503 the SMM divergence term is relatively equivalent to that of the CMM region. Given the high 504 amounts of local recycling during the winter in the SMM region (Figures 5a-b, 6a-b), local 505 recycling likely accounts for the lower value of divergence. Despite the concentration of high 506 divergence values in the southern portions of the continent, the magnitude of the convergence 507 term during the winter is highest along the eastern coast of the continent extending from the SCP 508 region up to ALC. This suggests that atmospheric circulation features transport supplies of 509 terrestrial ET from the south to the north for precipitation. 510 511 3.5.2 Spring Transport of ET 512 513 The divergence field shifts north from the winter to the spring season ( Figure 10e). Both 514 the US west coast and the US central/southern plains exhibit high divergence term magnitudes 515 during the spring season relative to the rest of the continent. The US central/southern plains also 516 exhibit high levels of evaporation allowing for their high divergence values (Figure 10d). Both 517 the SSE and SMM regions also experience high levels of evaporation, but their divergence terms 518 are relatively weaker. Consistent with the winter season, local recycling in the SMM region 519 contributes a considerable amount to total precipitation (Figures 5c-d, 7a-b) likely reducing the 520 amount of evaporation available for export. However, local recycling in the SSE region is of a 521 similar magnitude as the SCP region indicating atmospheric circulation likely exports SSE 522 moisture off the continent to the Atlantic Ocean. Similar to the divergence field, the magnitude 523 of the convergence field increases across much of the northern US and southern Canada from the 524 winter to the spring season (Figure 10f). The spatial variation of these two terms during the 525 spring season indicates that both the US west coast and central plains are important sources of 526 land-based moisture for much of the continent. 527 528 3.5.3 Summer Transport of ET 529 530 The maximum divergence shifts to the north again from the spring to the summer season 531 resulting in the highest divergence values across the US Northwest, US central/northern plains, 532 and the Canadian Prairies (Figure 10h). The corresponding evaporation field closely aligns with 533 the divergence values except in the OHV, SSE, and NEE regions (Figure 10g). During the 534 summer season, the SSE receives 46% of land-based precipitation from internal local recycling 535 (Table 3b) During the fall season, the divergence field is more spatially diffuse across the continent 549 than in the summer season (Figure 7k). The highest magnitudes of divergence are present across 550 much of the southern and eastern US, largely consistent with the evaporation field ( Figure 7j). 551 While the evaporation and divergence fields are consistent for most regions, the OHV, NEE, 552 CMM, and SMM regions all have higher evaporation fields than their divergence fields suggest. 553 Consistent with the summer season, the OHV and the NEE regions likely transport their moisture 554 out of the North American domain given the low amounts of internal recycling (Figures 5g-h, 9a-555 b). However, both CMM and SMM rely on internal local recycling for large fractions of their 556 precipitation (Figures 5g-h, 9a-b) likely lowering the available moisture for export. Unlike the 557 divergence field, the convergence field is more concentrated to the northeast, aligning closely 558 with the maximum percent contribution of the land surface to total precipitation (Figure 4d). 559 Though the SSE and NMW regions exhibit the highest magnitudes of the divergence term, the 560 spatial diffusivity of the divergence term in the fall season in comparison to the other seasons 561 suggests the high convergence in the northeast is sourced from a wide range of regions 562 continent-wide. 563

570
Using the expanded water tracing capabilities of CESM, we are able to split the land-based 571 precipitation into precipitation from T, C, and E moisture individually. This allows us to directly 572 investigate the varying contributions of each ET component to precipitation fields, moisture 573 recycling, and moisture transport. 574 575

Annual T/E/C Contributions
The annual climatology of the percent contribution of the land surface to total precipitation 578 is spatially consistent with the spring ( northern Mexico where T is limited due to a lack of vegetation and the arid environment. The 587 proportion of land-based precipitation originating from C moisture (moisture evaporated directly 588 from plant surfaces) ranges from 15-25% across much of the continent (Figure 8c). Only Alaska, 589 northwestern Canada, portions of the southeastern US, and Central America receive 25% of 590 more of their land-based precipitation from C moisture. These same areas receive the lowest 591 contribution from E moisture (moisture evaporated directly from ground and lake surfaces), with 592 totals ranging from 15-25% (Figure 8d). In contrast, the highest contributions (45%) from E 593 moisture occur in the western/southwestern US and portions of far northern Canada. The Great 594 Lakes also show high E contributions, though this is likely a result of lake evaporation. 595 596

604
During the winter season, E moisture is the most prominent source of land-based 605 precipitation for most of the continent, contributing between 40-60% of the total precipitation 606 originating from the land surface (Figure 9d). The lowest contributions of E are seen in the 607 Canadian prairies (30-40%), southeastern United States (30-40%), and much of Central America 608 (10-40%). The fraction of winter precipitation sourced from C moisture is also fairly uniform 609 across much of the continent, comprising 10-30% of the land-based precipitation over a wide 610 area (Figure 9c). The maximum contribution of C (30-50%) occurs across western Canada, 611 specifically within the prairies. Unlike E and C, there is a clear gradient in the contribution of T 612 moisture to total land-based precipitation (Figure 9b). The contribution of T has a minimum 613 across central Canada (10-20%) and increases towards the coasts and towards the south. T 614 contributes most to winter land-based precipitation in parts of the southeastern US and Central 615 America, with fractional contributions between 40% and 50%.

625
In the spring, the role of T increases across the continent as leaf out occurs and vegetation 626 extent increases. T contributes 40-50% of the land-based precipitation for all of the central and 627 eastern United States and for Central America (Figure 10b). Across much of Canada and the 628 western US, the contribution of T moisture is lower at 20%-40%. Moisture contribution from C 629 is fairly uniform during the spring season (Figure 10c). Most of the continent receives 10-20% of 630 springtime, land-based precipitation from C. The contribution of C increases to 20-30% in the 631 southeastern portions of the continent and along much of the west coast. As the proportion of 632 T:ET increases from winter to spring, the contribution of E declines. In contrast with winter E, a 633 gradient of increasing percent contribution from the SE to the NW develops across the continent 634 in the spring (Figure 10d). The contribution of E has a minimum in the southeast (0-20%) and 635 increases gradually towards the north and west, with the maximum contribution of 50-70% 636 occurring across much of northern Canada and Alaska.

689
The dominant ET component for recycling is highly dependent on the season. During the 690 winter season when T contribution is low, the combination of E and C dominate moisture 691 recycling across the continent (Figure 13a, Supplemental Table 5). C is the main source of 692 recycled moisture across most of southern Canada during this season, while E dominates across 693 much of the US. The role of T in local recycling increases for every region from winter to spring 694 (Figure 13b, Supplemental Table 6), though the combination of E and C is still responsible for 695 the majority of recycled moisture. However, during the summer season, T becomes the largest 696 contributor to recycled moisture for much of the north and eastern portions of the continent 697 (Figure 13c, Supplemental Similar to the recycled moisture, the composition of transported land moisture varies by 704 the season. In the winter, E is the primary land moisture source for most of the continent ( Figure  705 14a, Supplemental Table 9). In the southern US and Central America, there is a fairly even split 706 between transported E and T moisture, though E still exceeds T in all regions except for SWC, 707

729
The same framework developed in Section 2.2 used to examine the divergence and 730 convergence of total land surface ET in Section 3.5 can be used to examine the behavior of the 731 individual ET components. Equation (5)  where P, E, F, and T are defined the same as in Section 2.2 except now they apply only to T in 740 Equation (6), C in Equation (7), and E in Equation (8). Consistent with Section 3.5, 741 represents the divergence of T-sourced moisture (from here on referred to as the T-divergence 742 term), represents the divergence of C-sourced moisture (from here on referred to as the 743 represents the divergence of E-sourced moisture (from here on 744 referred to as the G-divergence term), represents the convergence of T-sourced 745 moisture (from here on referred to as the T-convergence term), represents the 746 convergence of C-sourced moisture (from here on referred to as the C-convergence term), and 747 represents the convergence of E-sourced moisture (from here on referred to as the E-748 convergence term). These terms allow for the direct investigation into the varying behavior of 749 individual moisture flux sources rather than assuming all the components behave in a similar 750 manner as total ET. 751 752 3.8.1 Winter Transport of T, C, and E 753 754 The predicted precipitation for each moisture flux using Equations (6)-(8) captures both 755 the spatial patterns and the magnitude of CESM-simulated North American precipitation sourced 756 from T, C, and E individually (Supplemental Figures 3-8). The winter divergence fields reveal 757 disparate behavior for each moisture flux source (Figures 15-17b). T-divergence is confined 758 primarily to the southern regions of the continent where vegetation remains active during the 759 winter months and T is the highest (Figure 15a). Unlike T-divergence, both C and E divergence 760 extend from Central America to southern Canada. Consistent with the C field (Figure 16a), the 761 magnitude of winter C-divergence is highest in the PNW, PFC, and SMM regions and is 762 moderately high across much of the central and southeastern US. The magnitude of E-divergence 763 is also high in these regions, though the maximum magnitude of E-divergence occurs in the 764 Great Lakes region. While the E field closely aligns with E-divergence, the E in both the SMM 765 and OHV regions exceeds their respective divergence values (Figures 17a-b). As noted in 766 Section 3.5, using our framework, differences between the evaporation and divergence fields can 767 arise from high amounts of internal recycling or from moisture export out of the North American 768 domain. Though the percent of local recycling from E is low in SMM (Figure 13a), total local 769 recycling is high (Figures 6a-b). Conversely, local recycling in the OHV region is relatively low 770 (Figures 6a-b), but the contribution of E to local recycling is very high (Figure 13a). In both 771 cases, recycling and their proximity to the oceanic regions (leading to an increased amount of 772 moisture export leaving the North American domain) likely reduces their overall divergence 773 values. Despite varying behavior in the divergence fields, all three fluxes have the highest 774 convergence magnitudes along the eastern coast of North America during the winter season. 775 However, maximum C-convergence and E-convergence is confined further northeast than T-776 convergence (Figures 15-17c). This is likely a result of higher T-divergence in Central America 777 converging in the southern and southeastern US allowing maximum T-convergence values to 778 extend further south than E or C-convergence. This behavior is consistent with high percent 779 contributions of T to total imported precipitation during the winter season in the SCP and SSE 780 regions (Figure 14a). American domain. Local recycling is high in both regions, particularly in SMM (Figures 7a-b), 790 and T is the largest contributor to recycled precipitation in both regions ( Figure 13b). As with the 791 winter season E, the combination of local recycling and proximity to the coasts likely reduce the 792 T moisture divergence from these regions. Unlike the T-divergence, C-divergence remains 793 relatively unchanged between the winter and spring seasons (Figure 16e). Both the highest C 794 flux and the highest C-divergence remains in the PNW, PFC, and SMM regions, though C-795 divergence increases across the central/southern plains in the US during the spring. The spring E-796 divergence field is moderately high across the entire western half of the US with the highest 797 magnitudes along the US west coast and in the central/northern plains. The C and E evaporation 798 fields are fairly consistent with their respective divergence fields during the spring season, 799 though C is higher in the SSE and E is higher in the OHV than their divergence values ( Figures  800  16-17d). Though E contributes more to internal recycling in the OHV region than the other two 801 moisture fluxes (Figure 13b), local recycling remains low during the spring season (Figures 7a-802 b). The inconsistencies between the evaporation and divergence fields indicate both regions 803 export a considerable amount of moisture out of North America. Despite differences between the 804 divergence fields, the spring convergence fields are very similar across the three fluxes ( Figures  805  15-17f). The northern half of North America (with the exception of the immediate west coast) 806 generally has the highest values of convergence for all three fluxes. Additionally, the SCP region 807 has a high convergence magnitude for each flux, likely the result of divergence fields from 808 Central America. 809 810 3.8.3 Summer Transport of T, C, and E 811 812 The convergence magnitudes for all three fluxes remain high across the northern half of 813 the continent during the summer while both the western and southern portions of the continent 814 have very low convergence magnitudes (Figures 15-17i). Both the T and T-divergence 815 maximums also shift north from the central/southern US plains in the spring to the 816 central/northern US plains and the Canadian Prairies during the summer (Figures 15g-h). Given 817 that the model configuration used in this study includes both crop management and irrigation, the 818 northward seasonal shift in maximum T-divergence could be related to agricultural harvesting 819 and irrigation patterns, though this is not addressed directly in this study. The T evaporation field 820 is consistent with the T-divergence field across the continent except for the OHV and NEE 821 regions where the amount of T far exceeds the divergence (Figures 15g-h). Although T is the 822 dominant moisture source for local recycling in both regions (Figure 13c), recycling is relatively 823 low (Figures 8a-b). The differences between T and T-divergence in the OHV and NEE regions 824 are attributed to atmospheric circulation exporting excess evaporation into the Atlantic Basin. 825 Similar to T-divergence, summer C-divergence is high across the Canadian Prairies and 826 central/northern plains, and is also high in the SCP, SSE, and SMM regions (Figure 16h). 827 Despite higher amounts of C in the SSE and SMM regions, the C-divergence is highest in 828 western Canada and the central US indicating more C-moisture exported from these regions 829 converges and precipitates within North America. Additionally, summer local recycling is very 830 high in both the SSE and SMM regions (Figures 8a-b), and C moisture comprises over 35% of 831 the recycled moisture in each region (Figure 13c, Supplemental Table 7b). This implies that both 832 regions export C moisture off the coast of North America and recycle large quantities of C 833 moisture, reducing their C-divergence. The magnitudes of E-divergence are highest across much 834 of central/western US and southern Mexico, and E-divergence is generally higher (lower) in the 835 regions where C-divergence is lower (higher) (Figure 17h). This behavior is consistent with 836 studies showing that as interception increases, soil moisture (and the resulting soil evaporation) 837 decreases (Lawrence et al., 2007). The maximum summer E-divergence occurs in the PNW 838 region, likely leading to the high E-convergence values in the WIP, EIP, and UPR regions 839 (Figure 17i). The E fields for several regions are inconsistent with the E-divergence during the 840 summer season (Figures 17g-h). The OHV, NCA, ALC, CMM, and SMM regions all have 841 higher E magnitudes than their respective E-divergence fields suggest. Each of these regions 842 borders at least one ocean (NCA, CMM, and SMM border two), so moisture export loss to the 843 oceans is likely. Additionally, summer recycling is very high in the CMM, SMM, and NCA 844 regions, further reducing the amount of E available for export (Figures 8a-b). The E and E-845 divergence fields are also inconsistent between the PNW and UPR regions. Both regions have 846 relatively equal E magnitudes, but the PNW E-divergence magnitude is much greater than the 847 UPR E-divergence (Figures 17g-h). This difference is largely attributed to the increased land 848 area available for PNW E moisture to precipitate given a traditional westerly atmospheric flow 849 across North America. This also allows PNW moisture to precipitate as a result of topography in 850 the UPR region, while much of the UPR E moisture may evaporate east of the model's resolved 851 topography, limiting the orographically-induced precipitation. 852 853 3.8.4 Fall Transport of T, C, and E 854 855 During the fall season, the highest convergence values for all of the fluxes cluster in the far 856 northeastern portions of North America (Figures 15-17l). T-convergence in particular stays 857 relatively confined to the northeast, while C-convergence has some moderately high values 858 further west in the WIP, EIP, and UPR regions. As vegetation begins to die off in the fall season, 859 both T and T-divergence maximums shift to the south and southeast (Figures 15j-k). The highest 860 values of T occur in the SSE and SMM regions, but only the SSE has a very high magnitude of 861 T-divergence. While both the SMM and SSE regions recycle a lot of land-sourced moisture in 862 the fall season (Figures 9a-b), the recycling in the SMM region exceeds that of the SSE by 863 approximately 8% (Supplemental Table 4b). However, the contribution of T to local recycling in 864 the SSE exceeds that of the SMM region by approximately 7% (Supplemental Table 7b). These 865 results suggest that the inconsistencies between T and T-divergence in these two regions are less 866 a function of recycling and more a function of atmospheric circulation. The T-divergence fields 867 indicate that much of the exported T moisture from the SSE region converges in the northeast, 868 while exported SMM T moisture is transported into the Atlantic. Unlike T-divergence, the E-869 divergence field remains relatively unchanged spatially from the summer to the fall seasons, 870 though the maximum E-divergence shifts from the US west coast to the Great Lakes region 871 (Figure 17k). This shift is consistent with the shift in E from the summer to the fall season 872 (Figure 17j). The largest inconsistency between fall E and E-divergence exists between the 873 NMW and OHV regions. Both regions have low recycling values during the fall (Figures 9a-b)

874
and E contributes a relatively equal percentage of recycling in both regions (Figure 13d).

875
Consistent with the fall T/T-divergence differences, the differences between E and E-divergence 876 are also likely a feature of atmospheric circulation conditions. Fall season C-divergence remains 877 high across southern Canada and the southern/southeastern US, though it drastically decreases 878 from the summer to the fall season in the central/northern US plains (Figure 16k). The main 879 disparities between the C and C-divergence magnitudes occur in the CMM, SMM, NEE, and 880 ALC regions (Figures 16j-k). Each region likely exports moisture out of the North American 881 domain given their proximity to the ocean. Additionally, the fall recycling is considerably higher 882 in CMM and SMM compared to most of the continent (Figures 9a-b) further limiting the C 883 moisture available for export. The contribution of C moisture to local recycling is slightly higher 884 in ALC than the NEE (34% vs 28%) potentially explaining the difference in the C-divergence 885 between these two regions ( Figure 13d, Supplemental Table 8b).   North America, where the influence of ocean evaporation is diminished (Figure 4). However, 917 distance from the coastline is clearly not sufficient to explain the heterogeneous spatial 918 patterns of land-sourced precipitation. Instead, these patterns reflect features of the regional 919 atmospheric circulation, topography, and vegetation type and distribution. 920 921 During winter, the contribution of terrestrial ET to total precipitation is relatively small 922 across North America due to a lack of photosynthetically active vegetation in the domain, 923 consistent with the relatively high contribution (40 -60% across much of the continent) of 924 bare ground and lake evaporation (E) to land-sourced precipitation (Figure 4a and Figure  925 9d). However, land-based precipitation accounts for 15-20% of total precipitation across a 926 southwest-to-northeast oriented swath from Central Mexico to the Great Lakes (Figure 4a). 927 This pattern reflects the substantial land-based moisture transport of ET from Central Mexico 928 and the South Central Plains towards the northeast within the prevailing winter southwesterly 929 flow (as shown in the moisture export analysis), highlighting the key role of these land 930 regions in shaping winter precipitation across a substantial portion of the Central U.S. 931 932 As mean temperatures rise in spring, the contribution of terrestrial ET to precipitation 933 increases across North America. In fact, terrestrial ET becomes the dominant source (50-934 60%) of precipitation in the northern Central Plains of the U.S. and central Canada by this 935 time (Figure 4b). This land-based precipitation maximum is equally sourced from E and T, 936 highlighting the emerging role of vegetation and transpiration by spring ( Figure  937 10b). Indeed, transpiration accounts for 40-50% of precipitation sourced from terrestrial ET 938 for most of North America as early as the spring season.

940
It is during the summer season however, that terrestrial ET, and in particular, vegetation, 941 becomes the dominant regulating source for much of North America's precipitation ( Figure  942 4c biomass for accurate prediction of summer precipitation (Figure 11c). For example, if instead 952 of landing on the canopy and evaporating quickly back into the atmosphere, precipitation 953 falls to the soil and infiltrates, the timing and magnitude of ET and subsequent precipitation 954 will be inaccurate. By fall, cooler temperatures and leaf senescence drive a reduced, though 955 still important role for terrestrial ET in shaping precipitation, similar to that in spring ( Figure  956 4d). However, compared to spring, the contribution of transpiration to fall precipitation is 957 smaller across much of North America, while the contribution of canopy evaporation to 958 precipitation is greater ( Figure 10 and Figure 12). In other words, the ratio of transpiration to 959 canopy evaporation in fall is smaller than that in spring. that relies on E-based or C-based recycling rather than T-based recycling may be more 978 susceptible to variability in recycling contributions to precipitation. This framework is used 979 to inform the examination and discussion of recycling at the seasonal scale.

981
During the winter season, appreciable recycling is confined primarily to southern North 982 America within the SCP, CMM, and SMM regions (Figures 5a-b). These three regions 983 largely comprise the southern half of the area of maximum winter land-surface contribution 984 to total precipitation (Figure 4a), further indicating the importance of local-ET for 985 precipitation in this portion of the continent. Given that E moisture dominates moisture 986 recycling in the SCP and CMM regions during winter (Figure 13a), a lack of precipitation 987 could quickly shut off the surface moisture available for local recycling, further reducing 988 precipitation. Since the SMM region receives most locally recycled moisture from T ( Figure  989 13a), recycling may persist during short periods of reduced moisture import/precipitation 990 from other regions, making internal recycling less vulnerable to sudden changes in 991 precipitation, potentially reducing land surface amplification of drying. evaporation transport also becomes apparent in spring as plants leaf out, contributing to 1143 remote precipitation on both sides of the continent (Figure 19f).

1145
The T-, E-, C-convergence results indicate that spring precipitation in the Canadian Prairies 1146 and Upper Rockies is likely vulnerable to springtime precipitation changes (those that 1147 greatly moisture, compared with E-or C-based moisture, is more likely to enter the atmosphere 1158 during periods of dry weather and is therefore less likely to precipitate quickly.

1160
In summer, ET-divergence magnitudes are highest in the central/northern US plains and in 1161 the Canadian Prairies indicating that these agricultural hotspots are major suppliers of land 1162 moisture for the rest of North America (Figure 7h). The low convergence values across the 1163 southern half of North America and along the west coast further indicate local ET is crucial 1164 for water supplies given most of these regions still receive between 30-50% of summer 1165 precipitation from land surface moisture (Figure 4c).

1167
The summer convergence & divergence fields for the individual ET components reveal 1168 several potential land moisture vulnerabilities for the North American continent. The T 1169 convergence field is clustered entirely in the north and northeastern portions of the 1170 continent (Figure 15i), while the T divergence field magnitudes are drastically higher in the 1171 US central/northern plains and in the Canadian Prairies (Figure 15h). These results suggest 1172 either land use/management changes (e.g., reduced irrigation) or severe/persistent drought The water tracers also allow for the direct investigation of moisture recycling from each 1248 defined land region across North America. During the warm seasons (Spring, Summer, Fall), 1249 moisture recycling is highest across the western half of the continent with some regions 1250 receiving up to 25% of total precipitation from internal recycling. Recycled moisture comes 1251 from all three ET components, but like the contributions to precipitation, the contributions of 1252 each component to recycling varies by season. E and C are the primary sources of recycled 1253 moisture in North America during the winter and fall seasons, while the summer is 1254 dominated by T, and the spring receives high contributions from all three components. 1255 Across much of western North America where local recycling rates are highest, the recycled 1256 moisture is comprised predominantly by E and C, which are both highly sensitive to changes 1257 in precipitation frequency. Our results indicate dry conditions in western and southwestern 1258 North America could quickly shut off the local recycling, amplifying drought conditions. 1259 1260 Using the water tracers and a matrix formulation of moisture transport, we identify key 1261 sources and sinks of land moisture seasonally across North America. In all seasons, eastern 1262 and northeastern North America import large quantities of moisture from other North 1263 American land regions, though the primary exports of land moisture vary seasonally. We 1264 identify potential key land surface teleconnections based on these moisture transport fields. 1265 Connections are found between Central America and the southern US, the southern/southeast 1266 US and the Northeast, the central plains and the Northeast, the Canadian Prairies and the 1267 Northeast, and the west coast and much of the interior western US/Canada. The individual 1268 ET component tracers reveal that in general, the connections to the western US/Canada 1269 primarily involve exports of C and E moisture, while the connections to the Northeast come 1270 where "! is the fraction of precipitation from partition that falls in partition . Note that for any 1343 region , the sum of all 's must equal one 1344 1345 1346 1347 1348 Placing equations (3)-(5) into equation (2), we obtain the following equality :  1349  1350  1351   1352  1353  1354 If we restrict our domain of interest to include # and $ only, several terms can be removed 1355 from equation (7). If our domain does not include % , both the evaporation from and the 1356 precipitation within % must be subtracted out of equation (7). To remove all of the evaporation, 1357 each term with '% is subtracted out, and to remove all of the precipitation, each term with "% is 1358 subtracted out. Recall that After subtracting the five terms from each side of the equality in equation (7), we are left with 1367 the following 1368 1369 1370 n j=1 λ ij = 1 (6) This straightforward result confirms the equality of E and P (using the assumptions introduced in 1382 Section 2) on limited subdomains. This result can also be easily expanded to include partitions 1383 of such that 1384 1385 1386 1387 resulting in 1388 1389 1390 1391