Unraveling Local and Upwind Controls on Terrestrial ET-Runoff Partitioning in China: Grid-Scale Analysis and Explanatory Diagnostics

Luo, Zhiwei; Ji, Ling; Xie, Yulei

doi:10.5194/egusphere-2026-3497

Preprints

https://doi.org/10.5194/egusphere-2026-3497

Preprints

03 Sep 2026

| 03 Sep 2026

Status: this preprint is open for discussion and under review for Hydrology and Earth System Sciences (HESS).

Unraveling Local and Upwind Controls on Terrestrial ET-Runoff Partitioning in China: Grid-Scale Analysis and Explanatory Diagnostics

Zhiwei Luo, Ling Ji, and Yulei Xie

Abstract. How water is divided between evapotranspiration and runoff controls regional water availability, but this partition is usually attributed mainly to local climate and land-surface conditions. Because part of precipitation is supplied by evaporation from upwind land areas, downwind partitioning may also reflect the ecohydrological state of terrestrial moisture-source regions. Here we analyze monthly evapotranspiration-runoff partitioning across mainland China from 2008 to 2017 using a geographically weighted random forest that separates local hydroclimatic and land-surface covariates from upwind land-surface covariates aggregated along atmospheric moisture trajectories. The evapotranspiration share varies non-monotonically with aridity, remaining near 0.80–0.82 from hyper-arid to dry sub-humid regions before declining to 0.62 in humid regions. Local controls dominate the explainable variability nationally, with mean cross-validated R2 increasing only from 0.468 to 0.471 when upwind covariates are added. However, the terrestrial upwind signal is geographically and seasonally concentrated, being most evident in dry sub-humid transition zones and in autumn. Upwind vegetation density, soil wetness, and precipitation show inverted-U relationships with the downwind green-water share, whereas upwind evaporative demand shows a monotone-negative relationship. These patterns persist across alternative evapotranspiration and runoff product combinations, but their interpretation remains limited by gridded product uncertainty, the terrestrial-only definition of upwind covariates, and the observational attribution design. The results suggest that assessments of blue-green water partitioning may benefit from considering upwind land-surface conditions alongside local hydroclimate, particularly in dry sub-humid transition zones.

Received: 15 Jun 2026 – Discussion started: 03 Sep 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 1905 KB)

Supplement (2034 KB)

Download & links

Zhiwei Luo, Ling Ji, and Yulei Xie

Status: open (until 19 Oct 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-3497', Anonymous Referee #1, 07 Sep 2026 reply
This manuscript examines the relative roles of local hydroclimatic conditions and upwind source-region characteristics in explaining ET-runoff partitioning across China. The topic is relevant to HESS, and the combination of geographically weighted modelling, moisture-source information, and model interpretation is interesting. The manuscript is generally well organized, and the main finding that local factors dominate overall while upwind signals become more apparent in some regions and seasons is presented with reasonable caution. I think the study has merit, but several methodological and interpretative points should be clarified before publication.
Major comments
The GW-RF implementation and validation procedure should be described more clearly. The comparison among the local-only, upwind-only, and all-factor models is central to the paper, but it is not fully clear how bandwidth selection and RF settings were handled relative to the temporal cross-validation. Some additional information on these settings and the validation workflow would make the analysis easier to reproduce. Since the improvement after adding the upwind variables is rather small at the national scale, I would also like to know whether this pattern is reasonably stable across the validation procedure.

I am not sure that the current description of the BG metric fully reflects what is being measured. The ratio (ET/(ET+R)) provides a useful measure of the relative partitioning between evapotranspiration and runoff, but it is not equivalent to a complete blue-green water balance because several storage and human regulation processes are not represented. I suggest defining it more explicitly as an ET-runoff partitioning index or flux-share proxy and briefly discussing its main limitations.

One point that was not completely clear to me is what is included in the upwind source region. If the analysis retains only source cells located within China and outside the receiving province, this should be stated explicitly, because the current wording may suggest a broader terrestrial source domain. It would also be helpful to clarify that the remaining source weights are renormalized after spatial filtering and that the resulting variables represent environmental conditions in extra-provincial terrestrial source regions within China rather than the total terrestrial moisture contribution to the receiving grid cell.

Some physical explanations in the Discussion should be stated more cautiously. The seasonal differences in the upwind signal are interesting, but interpretations involving monsoon retreat, continental moisture transport, or changing oceanic influence appear to be plausible explanations rather than processes directly tested in the present analysis. These interpretations can remain, but the distinction between model-based diagnostics and physical mechanisms should be made clearer, particularly when discussing PDP, permutation importance, and SHAP results.

Specific comments
Please check the consistency of the manuscript and Supplement titles, particularly the use of “Explanatory Diagnostics” versus “Causal Diagnostics”.

The range used for the PDP curves is described differently in the figure caption and Supplementary Methods. Please make these consistent.

The spatial processing of LAI appears to be described differently in different parts of the manuscript. Please check whether bilinear interpolation or area-weighted averaging was actually used.

Please check the entropy equation in the Supplement and the small numerical differences in several reported Δ R² values.

A short table summarizing the main datasets, variables, spatial resolution, and their role in the analysis would help readers follow the multi-source data framework.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-3497-RC1
CC1:
'Comment on egusphere-2026-3497', Allen G. Hunt, 09 Sep 2026 reply

I would like to draw the authors' attention to a recent paper. On lines 43-59, the authors address the recycling of continental moisture from the perspective of the moisture source, namely that some of the source of precipitation originates upwind over the continent. Then they begin the new paragraph on line 60, "Theoretical frameworks for water partitioning have not fully incorporated this continental perspective. The Budyko hypothesis links aridity to the long-term evaporative ratio at basin scale (Hunt et al., 2024; Peng et al., 2025; Rice and Emanuel, 2019)." Our publication (Hunt et al. 2024) does not, as the authors point out, address this topic. However, our subsequent publication (Hunt et al. 2025) does do so, albeit in a somewhat different manner. We point out that moisture recycling over larger basins should indeed increase the run-off ratio and decrease ET, since each cycle of evaporation and reprecipitation merely releases the same latent heat absorbed. A local perspective would count this ET multiple times. Our particular study found evidence of this scale-effect on ET in the more classic paper of Choudhury (1999) as well as in the lower evaporative index of continents than would be expected from its spatial average over land masses. Further, we also show that this recycling, as described in a classic review of Eltahir and Bras (1996), is directly related to the scaling of storage fluctuations with length scale, as revealed in studies of elasticity of streamflow. One may certainly correctly say that our publication does "not fully incorporate the continental perspective," but we believe that our analysis may help solve the same problem from a different perspective. Note that the important secondary references are all found in the paper cited below.
Allen Hunt
Hunt, A. G., Ghanbarian, B., Sahimi, M., and Duan, Qingyun, 2025 Scale effect on evapotranspiration: Predicting the continental and global scale water balance based on percolation theory and optimality principle, Accepted Nov. 17, 2025, by Water Resources Research. https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2024WR039658

Reply

Citation: https://doi.org/10.5194/egusphere-2026-3497-CC1
- AC1: 'Reply on CC1', Ling Ji, 10 Sep 2026 reply
  
  Dear Dr. Hunt,
  
  We thank you for drawing our attention to Hunt et al. (2025) and for explaining its relevance to continental moisture recycling. Our Introduction did not adequately acknowledge its contribution to understanding water partitioning across spatial scales.
  Your study combines percolation theory with the principle of optimality to examine the scale dependence of evapotranspiration and its relationship with continental moisture recycling. It also clarifies the different implications of temporal averaging and spatial aggregation, including the role of storage fluctuations.
  The relevance to our manuscript is therefore one of positioning and scope. Hunt et al. (2025) examines long-term evaporative ratios across spatial scales, whereas our study investigates whether the ecohydrological conditions of upwind terrestrial moisture-source regions add explanatory value for monthly ET–runoff partitioning at downwind grid cells beyond local hydroclimatic and land-surface conditions. This distinction also reveals an important limitation of our analysis: sensitivity to spatial aggregation remains untested at our single 0.5° resolution, and the monthly flux-share proxy contains no explicit storage term.
  Accordingly, we will revise the relevant passage in the Introduction to incorporate Hunt et al. (2025) and define the remaining research gap more precisely in terms of spatially resolved source–sink connections and source-region ecohydrological conditions. We will also add a brief statement to the limitation subsection of the Discussion, clarifying that the inferred relationships are specific to the spatial and temporal resolution examined here and that their sensitivity to spatial aggregation and storage change remains to be investigated.
  Thank you again for this constructive contribution to the discussion, which has helped us refine the manuscript’s conceptual framing and clarify the scope of our analysis.
  
  Kind regards,
  Ling Ji
  on behalf of all authors
  
  Reply
  
  Citation: https://doi.org/10.5194/egusphere-2026-3497-AC1

Zhiwei Luo, Ling Ji, and Yulei Xie

Supplement

https://doi.org/10.5194/egusphere-2026-3497-supplement

Zhiwei Luo, Ling Ji, and Yulei Xie

Viewed

Total article views: 331 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
237	55	39	331	38	34	31

HTML: 237
PDF: 55
XML: 39
Total: 331
Supplement: 38
BibTeX: 34
EndNote: 31

Views and downloads (calculated since 03 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	237	55	39	331

Cumulative views and downloads (calculated since 03 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	237	55	39	331

Viewed (geographical distribution)

Total article views: 331 (including HTML, PDF, and XML) Thereof 331 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 24 Sep 2026

Short summary

Rainfall can either return to the atmosphere or become runoff, shaping water availability. We studied this balance across China using monthly climate, land, and upwind moisture-source information. Local conditions explained most of the pattern, but land conditions in upwind source regions added a small signal in dry sub-humid areas and in autumn. The results show where water assessments may need to consider remote land conditions as well as local climate.


Total:	0
HTML:	0
PDF:	0
XML:	0