the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 May 2025
Sensitivity of a Sahelian groundwater-based agroforestry system to tree density and water availability using the land surface model ORCHIDEE (r7949)
Abstract. The Sahel region is characterized by its semi-arid climate and open-canopy agroforestry systems which play an important role in global carbon dynamics by sequestering an estimated 0.4 Mg C ha-1 yr-1, contributing to a total potential sequestration of approximately 558 Tg C if the agroforestry systems reach their maximum extent. However, land surface models (LSM) used in global climate modeling struggle to represent carbon dynamics in these ecosystems due to the inadequate representation of deep-roots tapping groundwater during dry periods, key environmental control for many agroforestry systems such as the widespread parklands based on the phreatophytic species Faidherbia albida. This study explores the sensitivity of Faidherbia albida parklands to tree density and water availability (rainfall and soil water content in the capillary fringe of the groundwater table) using a new configuration of the ORCHIDEE LSM. To this aim, the ORCHIDEE LSM was modified to simulate the growth of Faidherbia albida by simulating its inverted phenology based on forced temporal series of soil water content of soil layers between 4 m and 5 m and water saturation below 5 m and by adjusting the photosynthesis and carbon allocation parameters for Faidherbia albida and associated crops. The model was evaluated against independent eddy covariance and meteorological data from the Niakhar agroforestry site in Senegal. Simulation outputs were analyzed in terms of leaf area index (LAI), gross primary productivity (GPP), latent heat (LE), sensible heat (H) and net radiation (Rn). The model simulated tree GPP of 4.08 ± 0.21 tC ha-1 yr-1 compared to observed GPP of 5.06 ± 0.49 tC ha-1 yr-1. For croplands, the model produced GPP of 7.97 ± 0.89 tC ha-1 yr-1 compared to observed values of 7.78 ± 1.75 tC ha-1 yr-1. Simulations revealed that tree density positively influenced annual carbon uptake but reduced crop harvest at highest tree densities, indicating a trade-off between carbon sequestration and crop yield. Sensitivity analyses showed that interannual variability in soil water content in the capillary fringe of the groundwater table and rainfall influenced differently crop, tree and ecosystem carbon and energy fluxes. Despite its strengths, the model exhibited limited responsiveness of tree productivity to soil water content variability in the capillary fringe of the groundwater table, highlighting the need for enhanced representation of water uptake by tree roots in the model. These findings emphasize the importance of accurately modeling both surface soil water and groundwater dynamics and phenology to predict the responses of semi-arid agroforestry systems to climate variability. This study enhances our understanding of carbon and energy flux partitioning in complex, water-stressed and groundwater dependent agroforestry systems.
- Preprint
(2135 KB) - Metadata XML
-
Supplement
(1205 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-1102', Toni Viskari, 21 Jul 2025
-
RC2: 'Comment on egusphere-2025-1102', Anonymous Referee #2, 05 Sep 2025
This contribution presents the evaluation of the ORCHIDEE model on an agroforestry system in the Sahelian zone, characterized by Faidherbia trees, showing inverted (i.e. leafy in the dry season) phenology. The model is further used to conduct sensitivity analyses of (1) the ecosystem productivity to tree density and (2) year-to-year anomalies of water content in the non-saturated (tapped by the crop) and the saturated zone (tapped by deep-rooted Faidherbia).
I acknowledge the effort of the authors to adapt the ORCHIDEE model to this Sahelian agroforestry system, and evaluate it against LAI and carbon, water and energy flux data. The sensivity analyses are also nicely conducted.
Such a study is classical in its form but all the more informative than developed over a largely understudied ecosystem. More studies of this kind are needed to document the current behaviour (and project the future) of tropical ecosystems.
I see no major flaw in the science and propose the manuscript for acceptance with minor revisions (see my comments below).
Specific comments:
L17: MgC or tC : choose one and stick to it throughout the text
L18: "558 TgC" is expressed in units of stock. I suppose you're right, but please double check you did not mean a flux (in TgC/yr).
L107: "pearl millet and groundnut" : is it possible to precise the species? (with latin name)
L128-138: Overall clear description of EC data processing, but the method to partition GPP between tree and crop is missing. Please explain.
L177: unsure "physiognomy" is the proper word. Maybe "plant type" or "morphology"?
L182-184: missing is a description of how plants compete for light and water in this configuration of the model.
From L212, I understand there is no (i.e. zero) tree water uptake above 4 m? Is it coherent with isotopic analses of water uptake by Faidherbia (Roupsard et al. cited above)?
Is Equation 2 calculcated on a daily basis (or finer time scale)? Because Wi is dynamic so eq. 2 will modify Rf at the time step of Wi variations, with implication on carbon allocation for the tree.
L231-232: unclear to me, please rephrase.
L250-ff (from "In contrast, Jmax...") : sentence unclear or wrong. Rephrase.
L277 ("the ratio"): according to next sentence, the ratio is insitu/CRUJRA, so please rephrase to "the ratio of the sum of monthly rainfall in the observed data to monthly rainfall in CRUJRA".
L358: why not a full 12-month period? (jul-sep missing)
L356-361: the rationale behind definition of periods for the different variables of interest is not clear.
L370: What are the determinants of LAImax in this version of ORCHIDEE?
Figure 3: What is rather surprising for tree GPP (and to a lesser extent for crop GPP) is that day-to-day variability in EC-derived GPP is lower than in simulated GPP. Usually day-to-day variations of GPP are caused by radiation. Is it the case in simulated GPP? how comes this is not the case in EC-derived GPP?
Table 2 caption: about "n=24", precise this is a 6-year time series * 4 conditions (4 tree densities).
Figure 5a: When performing the SA to tree density, is all the surface not occupied by trees occupied by crops? I suppose yes, and this is the reason why in the "zero tree" condition, annual GPP reported here (about 10 tC/ha/yr) is higher than annual simulated GPP reported for crops (mixed with trees) in Table 1. This should be clarified in the MM.
Figure 6a: it's unclear to me how comes that SWCC anomalies are not distributed homogeneously along the annual cycle. It's said in the MM that the for computing the anomalies in environmental drivers (e.g. SWCC), the average annual cycle (e.g. of SWCC) was used. If that was the case, I would expected for a given day of year that the average SWCC anomaly is zero. This is apparently not the case. Let's take the example of the last day of Rainy season each year (vertical bar on October 1) : the negative anomalies of the first two years are not compensated by the positive anomaly in the 3rd year.
L522-523: failure to reproduce IAV of LAI are primarily due to lack of realism in the allocation part of the model (LAImax) rather than in phenological modules.
L545: "The model’s generality is estimated to have decreased in the developed configuration...". Rephrase to make clear that "the developed configuration" is the model version you are using in this manuscript.
L588: "competition for space"... and not competition for light?
L976: I suppose the reference for the paper of Vickers and Mahrt is wrong. Should be Vickers, D., & Mahrt, L. (1997). Quality control and flux sampling problems for tower and aircraft data. Journal of Atmospheric and Oceanic Technology, 14(3), 512-526.
Citation: https://doi.org/10.5194/egusphere-2025-1102-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,051 | 29 | 15 | 1,095 | 23 | 22 | 39 |
- HTML: 1,051
- PDF: 29
- XML: 15
- Total: 1,095
- Supplement: 23
- BibTeX: 22
- EndNote: 39
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This is a review for the manuscript “Sensitivity of a Sahelian groundwater-based agroforestry system to tree density and water availability using the land surface model ORCHIDEE (r7949)” submitted to Geoscientific Model Development by Gaglo et al. In this work, the authors introduce adjustments made to the ORCHIDEE model in order to simulate a PFT based on a dominant tree species in the Sahel region. The performance of the model is then compared against local flux measurements as well as tested with various scenarios to determine how sensitive it is to different conditions.
Overall, I was quite satisfied with the manuscript, as evidenced by the low number of line-by-line comments. It is a straightforward manuscript that focuses on a relatively simple implementation of a new model version with intentionally limited regional scope. Especially the calibration approach applied here is rather simplistic and carries its own effects that are not truly examined in the manuscript. However, since the examined system is in a semi-arid region that does not receive the appropriate amount of attention when discussing land surface modelling and touches on the various dynamics specific to these areas, I do consider it worthwhile contribution to the ecosystem model development discussion. As a sidenote, my apologies for that horror of a sentence there.
Now I am tempted to recommend acceptance with minor revisions as majority of the work here is easy to follow and comprehend, but the first section of the Discussion, specifically the realism/generality dichotomy here, makes me hesitate. I do comprehend the central idea of the matter and agree it is an important consideration when discussing model development. However, within the context of the work here, it came across as out of place considering that there isn’t enough experimentation here to ground majority of the claims there. For example, the argument of limited applicability at other locations is hindered by the fact that there is no experimentation how this PFT performs at those places even with simplified assumptions compared to using the existing PFTs that are not set for environments like this. And that is not even going into how much assumptions already exist in the general ORCHIDEE soil moisture implementation, so using that as a generalist comparison is debatable in itself.
My suggestion is to just remove the majority of the first Discussion section and focus completely on what you have shown here and be more concrete in explaining what the challenges in the larger implementation of the model are here. I understand that is partially the attempt here, but this is muddled by the chosen realism/generality/accuracy approach. Because of this I do think my recommendation is technically return for major submissions, but I do think it should be a relatively minor rewrite here.
Line-by-line comments:
Line 66: “In magnitude, the water use of Faidherbia albida trees at the plot scale was estimated to be less than 10 % of the annual amount of rainfall (Roupsard et al., 1999; Sarr et al., 2023). However, stable isotope tracing suggested a strong dependence of tree water use on groundwater (Roupsard et al., 1999).”
I was just a little bit confused by this pair of sentences. Is the argument here that the Faidherbia trees use little water, but also draw it from the deeper layers? As the current wording almost implies that the trees use approximately 10 % of the rain fall and in addition draw water from the deeper layers.
Either way, clarify the message here a bit.
Line 366: “…with an r2 between daily tree LAI simulation and observation of 0.81…”
Just to confirm that the correlation squared value was 0.81 between the observations and simulations? Which would indicate that the correlation itself was 0.9?
Nothing wrong with that, it is simply a staggeringly good fit. What caused me hesitations is that a few lines down the maximum measured LAI is also listed 0.81, which caused me a bit of confusion initially. Coincidences happen but still wished to check. Especially because that is a really high correlation to get when using the switch on phenological approach you seem to be using according to Figure 3.
Line 369: “However, 2 out of 3 years…” -> “However, during 2 out of the 3 evaluation years…”
Just to ease the reading a bit, although now I am wondering if I got the preposition right there.
Line 570: “As in our case study, their comprehensive evaluation of 13 models revealed that improving one dimension often compromises another, underscoring the difficulty of achieving optimal performance across all three.”
This is connected to my general comment, but the discussion preceding this does not really establish anything indicated here.