Long-term multi-site eddy-covariance observations of carbon, water, and energy fluxes in West Africa

Hingerl, Luitpold; Guug, Samuel; Bliefernicht, Jan; Kadum, Hawwa; Sy, Souleymane; Sawadogo, Windmanagda; Quansah, Emmanuel; Oussou, Francis E.; Amekudzi, Leonard; Salack, Seyni; Diallo, Belko Abdul-Aziz; Ayamba, Michael; Ramde, Emmanuel; Steinbrecher, Rainer; Neidli, Frank; Kunstmann, Harald

doi:10.5194/essd-2026-379

Preprints

https://doi.org/10.5194/essd-2026-379

Preprints

22 Jul 2026

| 22 Jul 2026

Status: this preprint is currently under review for the journal ESSD.

Long-term multi-site eddy-covariance observations of carbon, water, and energy fluxes in West Africa

Luitpold Hingerl, Samuel Guug, Jan Bliefernicht, Hawwa Kadum, Souleymane Sy, Windmanagda Sawadogo, Emmanuel Quansah, Francis E. Oussou, Leonard Amekudzi, Seyni Salack, Belko Abdul-Aziz Diallo, Michael Ayamba, Emmanuel Ramde, Rainer Steinbrecher, Frank Neidli, and Harald Kunstmann

Abstract. Long-term ecosystem–atmosphere flux measurements are crucial for understanding land–atmosphere interactions, ecosystem functioning, and climate variability. In West Africa’s Sudanian savanna, such data are limited despite rapid land use change and high climate sensitivity. We address this by providing a harmonized, multi-site eddy covariance dataset of carbon, water, and energy fluxes from six savanna ecosystems in Burkina Faso and Ghana.

The dataset provides half-hourly and daily turbulent CO₂, latent and sensible heat fluxes, net radiation, meteorological variables, and, where available, soil moisture and soil temperature. Measurements span the period 2013–2025 across contrasting land-use types ranging from forest savanna to cropland and degraded grassland. All data were processed with a uniform EddyPro-based workflow and further post-processed to ensure physical plausibility, temporal consistency, and harmonized variable definitions. Quality flags accompany all turbulent fluxes for user-defined filtering.

Archived in the PANGAEA repository and openly available (DOI: https://doi.org/10.1594/PANGAEA.994143), the dataset’s temporal continuity, explicit quality information, and multi-year coverage support diverse applications, including land surface and Earth system model evaluation, land–atmosphere coupling studies, satellite and machine-learning product validation, and analyses of seasonal to interannual variability and hydroclimatic extremes. These long-term, quality-controlled flux observations from a climate-sensitive, underrepresented region provide a key resource for advancing process understanding and model development in West African savannas.

Received: 19 May 2026 – Discussion started: 22 Jul 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

Status: final response (author comments only)

RC1:
'Comment on essd-2026-379', Kseniia Ivanova, 09 Sep 2026
General Comments
The dataset presented is unique and valuable. West Africa is significantly underrepresented in EC measurements. Providing data from six sites along a land-use gradient from 2013 to 2025 is a much-needed resource for the community.
The technical validation is reproducible.
Main reasons for a major revision:
There is a systematic calculation error in the Evapotranspiration (ET) values across all the stations.

The data completeness percentages in Table 3 are calculated from the number of existing records rather than the full time-series length, creating a misleadingly high impression of data coverage.

Several statements in the manuscript contradict the actual contents of the archived files and each other.

Key methodological procedures are not documented, including the numerical plausibility thresholds, the ET calculation, and the rules for daily aggregation.

It is not clearly stated which parts of the dataset are being published for the first time and which have already been used in previous publications.

Conclusion: Overall, this dataset deserves publication in ESSD after these issues are addressed.

Specific Comments
There is a systematic error in Evapotranspiration (ET) values. The ET units do not match between Fig. 6 (mm/h), Table 4 (mm) and the PANGAEA metadata (mm at half-hourly resolution). When comparing the published ET column with the value recalculated from Qe at Δt = 1800 s, a systematic discrepancy of approximately a factor of two is observed at all stations checked; this corresponds to the use of an hourly time step instead of a half-hourly one. As the daily ET is obtained by summing the half-hourly values, the discrepancy is also carried over into the daily products. Please check the calculation, provide the formula and the value of λ used in the text, and ensure that the units are consistent across the table, the figure and the repository.

Table 3 is misleading regarding data completeness. The percentages in Table 3 are calculated from "Total n" (the number of available records) rather than the theoretical length of the time series. If calculated from the full grid, the resulting values are substantially lower. Please add a column showing the maximum possible number of half-hourly periods and show the percentage based on that. Specify the exact operational period for each station in the table.

Soil heat flux (G). The text mentions three HFP01SC plates at 8 cm depth (L166), but there is only one G column in the data files.

Contradiction regarding QC = 2 (L202 vs L209-212). Line 202 states that fluxes with flag 2 are excluded from the dataset. Lines 209–212 state that filtering was not applied and QC = 2 data were only removed for temporal aggregations. The archive includes data with QC = 2. Please correct the text to reflect this.

Distinction from previous publications. Previous works by the group are cited (L52–53), but it is never explicitly stated which part of this dataset is being published for the first time. Additionally, on the PANGAEA landing page, the "Supplement to" field points to a preprint by Sy et al., not to this manuscript.

Exclusion of CH4 data (L172–174). CH4 was measured at Janga and Mole but was removed "for future separate analyses," even though Guug et al. 2025 regarding CH4 is already published and cited. The ESSD review criteria state that datasets should not be split to increase the number of publications. Please either justify this exclusion or include the CH4 data.

Undocumented daily products. Appendix A1 only mentions "daily averages or sums derived from the half-hourly data", and L596 says "derived accordingly." Crucial details are missing: what is the minimum data coverage required to calculate a daily value? Which variables are averaged and which are summed? Additionally, there are no quality flags in the daily files (L568–569), meaning users cannot filter the daily data based on quality.

Broken reference (L217–218). The text refers the reader to the "Data availability" section for information on the structure of data gaps. However, that section contains no such information.

Precipitation. The text states that three sensors were used and cross-validated (L163–164). However, only one "most complete" series is released. It is unclear which sensor this is (and it might be different for different stations).

Data availability section (L470–480). A full bibliographic reference is pasted directly into the text instead of a standard citation. Only the parent DOI is provided, but each station has its own DOI.

Categorisation of Sumbrungu/Gorigo (L485 vs L139–151). In the conclusion (L485), Sumbrungu/Gorigo are listed with a slash as a single site, whereas in Section 2.1.3 and Table 1 they are listed as different land-use types with different installation heights. The wording using a slash suggests a continuous series from 2013 to 2024, whereas there is a gap between them, involving the relocation of the station and a change in height from 2.65 to 4.3 metres. Please separate them in the conclusion.

12 The metadata files described in Appendix A1.1 are missing from the PANGAEA collection(campaign_metadata.tsv, event_metadata.tsv, parameter_metadata.tsv)

Technical Corrections
L76: Unremoved editor/author note: "(include Quansah/Nadolski)".

L690: EddyPro v7.0.8 is listed in the references, but v7.0.9 is mentioned in the text and metadata.

Table 4: Air pressure is listed as kPa, but in the archive in PANGAEA it is in hPa (values are 963–997).

Fig. 3 Caption not self-explanatory.The explanation for a)–e) should be in the caption, not just the legend. Missing details: time period, filtering criteria applied, number of points, and type of regression used. It is not stated why only 5 out of 6 stations are shown (please repeat in the caption as well).

Fig. 4: "Stab / Neu / Unst" are not defined in the caption.
Citation: https://doi.org/10.5194/essd-2026-379-RC1
RC2:
'Comment on essd-2026-379', Ossénatou Mamadou, 06 Oct 2026
Review of Long-term multi-site eddy-covariance observations of carbon, water, and energy fluxes in West Africa (essd-2026-379)
This manuscript presents a harmonized, quality-controlled eddy-covariance (EC) dataset from six Sudanian savanna sites in Burkina Faso and Ghana (WASCAL observatory), spanning 2013-2025 and covering a land-use gradient from near-natural forest to cropland and degraded grassland. The dataset is a valuable contribution given the acute scarcity of long-term, standardized EC observations in West Africa. The technical validation (energy balance closure, turbulence/stability regimes, footprint analysis, event-based consistency checks, climatological patterns) is comprehensive. The processing workflow, with quality flags, is sound but needs clarification and correction. I recommend major revisions before publication, primarily to resolve a numerical inconsistency in the energy-balance closure results, clarify data availability for several site-years, and address some issues noted below.
Major Comments
The energy balance closure values do not match those reported in Figure 3 (Section 4.1, lines 273–274). The text states that the closure ratios range between 0.67 and 0.89 (line 273), corresponding to an unclosed fraction of approximately 10-30% of the available energy. However, Figure 3 reports regression slopes ranging from 0.66 (Figure 3e) to 0.83 (Figure 3a), while the cumulative EBC ratios range from 0.60 to 0.96, with R² values between 0.75 and 0.88 across the five panels. Neither of these ranges is consistent with the 0.67–0.89 values cited in the text, and the reported 10-30% unclosed fraction does not appear to correspond clearly to either metric. Since the percentage of non-closure can be derived directly from the slope of the linear regression when the intercept is constrained to zero, the slope values of 0.66-0.83 shown in Figure 3 would imply a substantially different degree of energy balance non-closure. The authors should therefore either constrain the intercept to zero and report the corresponding percentage of closure/non-closure, or clearly specify which EBC metric the text refers to and revise the reported closure percentages accordingly. This is an important technical validation result and should be presented consistently throughout the manuscript. In addition, the authors could discuss the observed energy balance non-closure in the context of previous studies, such as Wilson et al. (2002; 1016/S0168-1923(02)00109-0), which provides a well-established reference for discussing energy balance closure issues and their possible causes.

It would be useful to present the 30min dynamics of all fluxes across sites to highlight the clarity of the datasets in relation to the seasonal and temporal dynamics reported for other sites in the region.

I have some concerns regarding the use of the term “long-term” to describe the observation period across all sites, given that only three sites have more than 5-10 years of continuous eddy covariance (EC) measurements. The authors should consider using a more appropriate term in the manuscript title.

Site description section: The authors could provide the ecosystem type according to the IGBP classification for the Nazinga and Mole sites. This information would facilitate inter-site comparisons.

Line 31: The sentence "Existing EC studies in Burkina Faso and Benin have demonstrated pronounced diurnal, seasonal, and interannual variability..." cites Brümmer et al. (2008), Ago et al. (2016), and Mamadou et al. (2016). The authors should also compare their results with Mamadou et al. (2014) (https://doi.org/10.5194/hess-18-893-2014), which provided an extensive analysis of the seasonal cycle of a mixed cultivated area similar to the Kayoro cropland site analyzed here.

Line 35: This is not necessarily an issue, but the phrasing could be clarified to avoid implying an overlap with the geographic scope of the present dataset, especially since the following sentences (Ago et al., 2015; Quansah et al., 2015; Berger et al., 2019) already refer specifically to the Burkina Faso and Ghana network sites.

L41-45: There are some interesting studies related to the similarity assumptions, seasonal changes in surface roughness, and atmospheric stability in the region (see, Hounsinou et al., 2022; 2023).

Lines 146-151: The reclassification of Gorigo from ‘degraded savanna’ to ‘semi-degraded grassland’ is justified based on the temporary cultivation of maize in 2018-2019 and a slightly higher sensor height. This is a reasonable argument, but the paper would be stronger if it briefly states whether this reclassification affects comparability with the original Sumbrungu dataset for users conducting long-term trend analyses across the Sumbrungu/Gorigo transition.

Table 3: Consider adding a column for momentum flux QC statistics, since u* plays a key role in the turbulence/stability analysis (Section 4.2) and in the usage-note recommendation for additional u*-based filtering of nighttime CO₂ fluxes.

Line 160: In the sentence, “For the energy balance closure assessment, the incoming and outgoing short- and longwave radiation are essential,” the authors should consider referring more broadly to all non-turbulent fluxes, including radiative fluxes, ground heat flux, and, where relevant, storage terms. This is particularly important given the characteristics of the study sites, where ground heat flux may represent a significant component of the energy balance at sub-daily timescales. Accounting for these terms would provide a more complete assessment of the energy balance closure. I suggest reformulating the sentence and saying "radiative balance" instead of "energy balance closure assessment."

Lines 291-298: The explanation attributing Gorigo's broader daytime u* distribution to sensor height and larger sample size is plausible but somewhat speculative. If roughness-length z0 estimates are available, a brief quantitative comparison (e.g., z0 estimates for Gorigo vs. Sumbrungu) would substantiate this interpretation more convincingly than the qualitative argument currently given.

Footprint analysis is presented only for Mole, Janga, and Gorigo (Figure 5), with Nazinga, Kayoro, and Sumbrungu deferred to Hingerl et al. (2025). Given that this is meant to be a standalone dataset paper, a one-sentence summary of the key footprint findings for the three deferred sites (rather than only a citation) would help readers assess representativeness without needing to consult a separate paper.

Authors said that ‘Although both Nazinga and Mole are classified as near-natural savanna sites, their turbulence and stability fingerprints reveal important contrasts. At Nazinga, dense, spatially homogeneous savanna vegetation promotes locally generated turbulence and produces a comparatively well-organized stability distribution that largely conforms to Monin–Obukhov similarity theory (MOST) expectations (line 306-309). How? Please clarify.

Section 5 (Usage notes): This section is helpful and appropriately cautious. Consider adding an explicit note recommending u* filtering thresholds (even approximate, site-specific ranges) given that the turbulence analysis in Section 4.2 already provides the necessary data to suggest reasonable starting values for each land-use class.

Section 4.5: How did the authors deal with the data gaps in order to aggregate the measurements and derive the daily CO₂ fluxes presented in Figure 7? Please describe the gap-filling procedure or methodology used. In addition, refer to the variable as CO₂ flux rather than NEE, for consistency with the ‘flux science’.

Section 3.1 (line 200): The quality flag description states "0 = high, 1 = moderate, 2 = low quality," consistent with the Spoleto agreement, which is good, but it would help to explicitly state whether flags 0 and 1 are additionally distinguished anywhere in the dataset beyond the QC ≤ 1 recommendation, since Table 3 already reports both QC ≤ 1 and QC = 0 separately.

Minor comments
Line 230: correct "obeservatory" to "observatory", same in line 232 for precipitation

In the title, please replace “carbon” with “carbon dioxide” and “water” with “water vapor”

Line 260: please refer to Table 2. Overall the heatmap... There is a missing comma after "Table 2."

Section 2.1.1 (line 99): "the region's historically high population density" consider citing a source for this claim, as it is used to justify the "near-natural" (rather than fully natural) classification of Nazinga.

Consistent use of "eddy-covariance" vs. "eddy covariance" (hyphenated inconsistently throughout

Line 86-87 : Regarding the sentence, “The dataset includes half-hourly and daily aggregates of CO₂, H₂O, and energy fluxes, ancillary meteorological variables, and energy balance closure diagnostics, processed following FLUXNET community standards,” this statement may overstate the extent to which the data processing follows the FLUXNET community standards, particularly since flux partitioning was not performed; the storage term was not calculated; etc.. The authors should therefore rephrase this statement to more accurately describe the processing procedures actually applied and avoid implying full compliance with the FLUXNET processing protocols.

Line 591: The stability parameter should be expressed as (z − d)/L rather than z/L. Please revise the corresponding equation and text accordingly

Figure B1: The authors should also consider including the momentum flux, as this would allow the results to be more directly comparable with other datasets

Line 172: Remove the sentence corresponding to the CH4 measurements since this is not the focus of this paper.

In Table 2: sonic air temperature should be ‘sonic dry air temperature’

Line 197: In the sentence ’Fluxes were corrected for density fluctuation using the Webb-Peanman-Leuning correction’; the correct reference is Webb et al., 1980; please remove the other references.

Table 3 : Complete with the QC statistic for the momentum flux

Table 4: This should be only the CO2 flux since the storage term was not taken into account. The stability parameter must be z-d/L, d the displacement height

In lines 275- 277: How can you explain the difference in EBC among the sites? It is very strange that closure at the forest sites (Nazinga, Mole) was better compared to the other sites. How can you explain these differences? The authors could compare these results with other findings in the Sudanian climate for the best energy closure balance above a mixed cultivated site (see, for example, Mamadou et al. 2014; Mamadou, 2014).

Line 282: In the sentence ‘This distinction is important flux not ... complete with the reference.

References cited
Wilson, Kell, Allen Goldstein, Eva Falge, et al. « Energy balance closure at FLUXNET sites ». Agricultural and Forest Meteorology, FLUXNET 2000 Synthesis, vol. 113, n^o 1 (2002): 223‑43. https://doi.org/10.1016/S0168-1923(02)00109-0.
Hounsinou, Miriam, Ossénatou Mamadou, Maxime Wudba, Basile Kounouhewa, et Jean-Martial Cohard. « Integral turbulence characteristics over a clear woodland forest in northern Benin (West Africa) ». Atmospheric Research 268 (avril 2022): 105985. https://doi.org/10.1016/j.atmosres.2021.105985.
Hounsinou, Miriam, Ossénatou Mamadou, et Basile Kounouhewa. « Turbulence Characteristics in the Atmospheric Surface Layer Over a Heterogeneous Cultivated Surface in a Tropical Region ». Boundary-Layer Meteorology 188, n^o 2 (2023): 321‑44. https://doi.org/10.1007/s10546-023-00815-z.
Mamadou, Ossénatou. « Etude des flux d’évapotranspiration en climat soudanien : comportement comparé de deux couverts végétaux au Bénin ». These de doctorat, Grenoble, 2014. https://theses.fr/2014GRENU019.
Mamadou, O., J. M. Cohard, S. Galle, et al. « Energy fluxes and surface characteristics over a cultivated area in Benin: daily and seasonal dynamics ». Hydrol. Earth Syst. Sci. 18, n^o 3 (2014): 893‑914. https://doi.org/10.5194/hess-18-893-2014.
Citation: https://doi.org/10.5194/essd-2026-379-RC2

Data sets

Long-term eddy-covariance measurements of energy, water, and carbon fluxes across a savanna land-use gradient in West Africa Luitpold Hingerl et al. https://doi.org/10.1594/PANGAEA.994143

Viewed

Total article views: 589 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
313	142	134	589	176	128

HTML: 313
PDF: 142
XML: 134
Total: 589
BibTeX: 176
EndNote: 128

Views and downloads (calculated since 22 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	113	44	7	164
Aug 2026	55	17	2	74
Sep 2026	98	59	101	258
Oct 2026	47	22	24	93

Cumulative views and downloads (calculated since 22 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	113	44	7	164
Aug 2026	55	17	2	74
Sep 2026	98	59	101	258
Oct 2026	47	22	24	93

Viewed (geographical distribution)

Total article views: 569 (including HTML, PDF, and XML) Thereof 569 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 14 Oct 2026

Short summary

This paper presents a long-term environmental observation dataset from six sites in Burkina Faso and Ghana, providing measurements of exchanges of carbon, water, and energy between ecosystems and the atmosphere. The data were collected across different land-use types and processed using a consistent workflow to ensure quality and comparability. The openly available dataset supports climate research, environmental monitoring, and improved understanding of environmental change in West Africa.


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