Long-term water stress and drought monitoring of Mediterranean oak savanna vegetation using thermal remote sensing

Drought is a devastating natural hazard, difficult to define, detect and quantify. Global meteorological data and remote sensing products present new opportunities to characterize drought in an objective way, and to extend this analysis in space and time. In this paper, we applied the surface energy balance model SEBS (Surface Energy Balance System) for the 15 period 2001-2018, to estimate evapotranspiration and other energy fluxes over the dehesa area of the Iberian Peninsula, with a monthly temporal resolution and 0.05o pixel size. A satisfactory agreement was found between the fluxes modelled and the measurements obtained for three years by two flux towers located over representative sites (RMSD= 21 W m-2 and R2 of 0.76, for all energy fluxes and both sites). The estimations of the convective fluxes (LE and H) showed higher deviations, with RMSD = 26 W m-2 on average, than Rn and G, with RMSD = 15 W m-2. At both sites, annual ET was very close to 20 total precipitation with the exception of a few wet years in which intense precipitation events, producing high run-off, were observed. The analysis of the anomalies of the ratio of evapotranspiration (ET) to reference ET (ETo) was used as an indicator of agricultural drought on monthly and annual scales. Hydrological years 2004/2005 and 2011/2012 stood out for their negative values, with the first one being the severest of the series, the impact observed on vegetation coverage and grain production. On a monthly scale, this event was also the longest and most intense, with peak negative values in January25 February and April-May of 2005, explaining its great impact on cereal production (up to 45% reduction). During the drier events, the changes in vegetation ground cover over the months, with a preponderant presence of grasslands compared with those in which only oak trees were active, allowed a separate analysis of the strategies adopted by the two strata to cope with water stress. These results indicate that the drought events characterized for the period did not cause any permanent damage https://doi.org/10.5194/hess-2020-190 Preprint. Discussion started: 26 May 2020 c © Author(s) 2020. CC BY 4.0 License.

1. Line 13: The sentence beginning "Drought is a ..." might be expressed more clearly as : "Drought is a devastating natural hazard that is difficult to define, detect and quantify." The sentence will be changed 2. Line 13: The sentence beginning "Global meteorological data ..." is oddly constructed. It might be more clearly expressed as" The increased availability of both meteorological and remotely sensed data provides an opportunity to develop new methods to identify drought conditions and characterize how it changes over space and time." The sentence will be changed 3. Line 26: The sentence beginning "During the drier ..." is unclear and needs revision.
The sentence will be changed to: "During the drier events, the changes in the grasslands and oak trees ground cover allowed a separate analysis of the strategies adopted by the two strata to cope with water stress". 4. Line 34: The sentence beginning "Drought is a ..." could be expressed more clearly if constructed as: "Drought, which is both a devastating natural hazard and globally widespread, has complex consequences across spatiotemporal scales and sectors." The sentence will be changed to the proposed construction.
Indicators of drought, it will be clarified in the manuscript.
7. Line 53: The sentence beginning "LST and VIs" reads oddly. The authors seem to be saying that by combining information about the surface temperature and vegetation, remote sensing-based models can provide accurate estimates of ET. But, rather than statin that explicitly, the coach it in terms of vegetation indices etc.
The sentence will be changed to 'LST and VIs have been used to provide ET estimations over agriculture … 8. Line 115: This paragraph is a bit unclear. The authors state the parameterization of green vegetation fraction and height are unique for the dehesa. Are the authors back calculating the leaf area index (L) using equations 8 & 9? If so, why? Also, there is no discussion of canopy height and how it's calculation is modified to better represent the dehesa.
Yes, we obtained Fc using eq. 8 and L is derived from Fc using eq.9. To clarify the procedure, we will modify eq. 9 to provide a more direct computation of L.
The computation of the canopy height is described in the manuscript, but the paragraph is unclear, and it will be modified. Considering that the tree stratum of the dehesa is quite homogeneous in composition, dominated by mature Quercus ilex sp., and that grassland canopy has a very high variability of low height herbaceous species, the ecosystem structure has been simplified to compute hc in the following way: A constant height of 8 m has been assigned to oak trees, which is multiplied by its ground coverage in each pixel. Oaks fc is computed annually using summer NDVI in eq. 8. During the summer the grasslands are dry, and the only photosynthetically active vegetation contributing to the NDVI signal are the oak trees. The grassland height is low (< 1 m), affecting the effective canopy height of each pixel less than the trees, and it is also difficult to computed based on monthly vegetation indices given the high species variability. For this reason, the grassland height has been discarded and only the contribution of trees was considered to compute hc. We are aware that this is a simplification of a complex system that will contribute to the error of modelled fluxes. However, it was an operative solution considering the scale of this study.
9. Line 172: It would be helpful if the authors included a histogram and an estimate of the distribution skewness for ET and relative ET. From the description given here it appears quite small.
In the figure below (Fig1-commentR1), we present the histograms (one for each month) of both variables. For both variables most months presented an approximately symmetric distribution, with skewness between -0.5 and 0.5, three of them were moderately skewed and only one month (for ET) and two months (for ET/ETo) were slightly above one. We will elaborate this point in the manuscript. However, given the limited number of available points, these graphs only provide preliminary information and more data is required to confirm this point. For this reason, we prefer to include these graphs as supplementary information and not as part of the paper. 10. Line 182: Replace "presented a general good agreement" with "generally showed good agreement".
It will be replaced 11. Line 184: Why the greater discrepancy for the turbulent fluxes compared to the C2 nonturbulent fluxes? Is this linked to imperfect closure for the flux measurements? Errors in partitioning the available energy between H and LE?
The imperfect energy balance closure is certainly a reason, as well as the discrepancy in the footprints of the different sensors (radiometer, soil heat flux plates, and the instruments for measuring the turbulent fluxes) and that of SEBS estimates. However, the different complexity in the formulation and computation of the radiative and the turbulent fluxes (the net radiation equation is a kind of linear representation, while the equation to estimate the sensible heat flux is highly non-linear), and the factors that influence each component (Rn is influenced by LWD, SWD, albedo and LST; H is influenced by LST, Ta, wind speed, NDVI, fc and LAI) also influence the final error. A small bias in LST, Ta, and vegetation information can cause a high bias in H (and thereby LE, compute as a residual). The soil heat flux has usually a low RMSD, but generally this comes with a higher relative error, due to the reduced magnitude of this flux.
12. Line 206: The sentence beginning "Very low runoff ..." is redundant and could be omitted.
It will be deleted 13. Line 207: Why isn't the relationship shown? Although it reasonable to suspect these two quantities would be correlated, a "close" relationship is a bit of a surprise. It would be useful to show this relational.
We show below (F2-commentsR1) the relationship between annual run-off measured at the Sta.Clo catchment reservoir and the annual aridity index (Budyko, 1974) estimated for the same catchment on the left, and the same relationship with the run-off (Q) also normalized by precipitation on the right. The shape of these relationships shows how variations in climate, as represented by variations of P and ETo, impact runoff and could provide a mean to assess the effects of a changing climate on water availability in this watershed. The budyko model represented in (b) was derived using Zhang et al. (2008) eq.9 with an adjusted value for a parameter equal to 0.54. It shows a mean to estimate long term annual run-off values in this catchment. Although these are interesting relationships, useful to complement the drought assessment, it's a little outside of the topic and might disrupt the flow of the results, so we prefer to present it as supplementary material.
F2-commentsR1. (a) Relationship between annual run-off measured at the Sta.Clo catchment reservoir and the annual aridity index (Budyko, 1974) estimated for the same catchment and (b) Relationship between run-off coefficient measured at the Sta.Clo catchment and the annual aridity index.
14. Line 207: Numerous metrics and indices have proposed been proposed over time to quantify aridity. It would be helpful to add a sentence or two to describe this index.
We will include in the text the following definition: "These annual run-off measurements followed a close relationship (Figure 4) with the annual aridity index (Budyko, 1974), estimated at Sta.Clo following Arora (2002), as the ratio between potential evaporation and annual precipitation. This is a correct observation. The difference is caused by the big difference in precipitation, as indicated in Fig. 3, precipitation at Sta.Clo (683 mm/a) is about twice that at ES-LMa (338 mm/a).
16. Line 253: it worth point out that the peak in the autumn is much weaker than the one earlier in the year.
Yes, it will be pointed out in the revised manuscript.
17. Line 299: The phrase "and the more ..." also refers to ES-LMa, which was already discussed.
Yes, it will be deleted 18. Figure 5: The word "fraction" is misspelled.
It will be corrected