We thank the reviewer for his/her comments. The comments have helped us to state the contribution of the study/dataset more precisely, and to strengthen the physical and methodological motivation for several choices that were presented too briefly in the submitted manuscript. Below, we address each comment raised by the reviewer.

On the whole, we would like to note that ESSD's uniqueness criterion states that "any data set on a variable supposed or suspected to reflect changes in the Earth system deserves to be considered unique," and that its test for methodology is whether "a new or improved method should not be trivial or obvious." Compound dry–hot extremes are a variable of precisely the first kind. On the second, we hope the responses below make clear that the methodological steps involved are neither trivial nor obvious: two components of the index are fitted locally at every land pixel rather than assumed globally, the soil-moisture memory timescale τ, which drives the antecedent weighting in SASMI (Comment 3), and the copula representing the joint moisture–heat dependence, and the event catalogue is constructed through cell-by-cell threshold calibration with distributional testing rather than by thresholding the index (Comment 5).

Detailed responses are hereafter provided:

Comment 1

In the review criteria, ESSD mentions significance as the most important aspect. With regard to uniqueness, it is stated that "it should not be possible to replicate the experiment or observation on a routine basis". The current manuscript, in my view, fails to clear this hurdle representing a more-or-less straightforward application of a standardized index to openly available reanalysis data, with poorly motivated choices that strongly impact the applicability of the dataset (see also below).

We thank the referee for engaging directly with the significance criterion. We are not aware of any openly available dataset that offers a daily, globally complete, soil moisture centric   compound dry–hot index over six decades together with a statistically filtered, event-resolved catalogue. The open compound indices we know of are regional, cover shorter periods, are monthly or seasonal, or rest on precipitation-based drought proxies rather than soil moisture. It is this combination that is missing, and that is the gap we aim to fill. The two components also serve different users: the continuous SSTCI field is suited to process-level and attribution research, where a daily signal on a common grid is what such analyses require, while the event catalogue offers a discrete, comparable inventory that impact, risk and exposure studies can work with directly.

Furthermore, the product is not the output of a routine index calculation. At each step, we parameterize locally at every land pixel rather than making global assumptions. Chiefly, the soil-moisture memory timescale τ, which drives the antecedent weighting in SASMI, and the copula linking the moisture and temperature components, which represents their joint dependence and the extreme event catalogue. Our companion paper (Aftab et al., 2026a), which focused on the methodology’s development, demonstrated the SSTCI methodology at a single basin using one uniform τ; a reasonable simplification for one homogeneous region, but not one that transfers to the globe. We set out the physical basis for both choices in our responses to Comments 2 and 3. The catalogue is likewise not a thresholding of the index; the event-extraction procedure and its statistical basis are described in our response to Comment 5.

Each of these steps carries its own computational cost, and they compound. That is, τ is fitted at every land pixel, a copula is fitted at every pixel, the index is evaluated daily across 63 years, and the event identification and its distributional tests are run cell by cell at 0.1° over the global land surface. Reproducing the product is not a matter of running a script over ERA5-Land. We would also note that ESSD has recently published Dheed (Weynants et al., 2025), a global catalogue of dry and hot extreme events built entirely from ERA5, which suggests the significance criterion turns on what a product delivers to users rather than on whether its inputs are public. We release our code alongside the data, and sparing users the burden described above is itself part of what the dataset is for. In the revision we will state this combined contribution.

Comment 2

In their approach, the authors use soil moisture averaged over the top 7 cm of the soil profile. This seems rather arbitrary, as there is no physical reason why the top 7 cm soil moisture would influence air temperatures more than the top 2 cm or 10 cm, or even deeper. In many more humid climates, soil moisture averaged over the root zone would be the best representation of soil moisture impacts on T through land-atmosphere interaction... The authors do not provide a physical justification of this arbitrary choice, nor do they provide a sensitivity analysis.

We thank the referee for this important point and welcome the opportunity to clarify the reasoning behind the choice, which follows from the objective of the index rather than being arbitrary.

SSTCI is designed to resolve the daily-scale onset and evolution of compound dry–hot conditions and the rapid land–atmosphere coupling that drives them. For that target, the near-surface layer is where the coupling is most directly expressed. That is, it controls direct soil evaporation and is most tightly tied to the daytime surface energy balance that governs latent–sensible partitioning, which is the mechanism by which moisture deficits amplify near-surface heat. This is consistent with our use of daytime (06:00–18:00 local time) fields, chosen to capture the window in which that coupling is strongest. The 0–7 cm layer (swvl1) is also the standard and most extensively validated near-surface soil-moisture layer of ERA5-Land, which supports comparability with the broader literature. We would add that the choice of a shallow layer does not leave the index blind to longer-timescale moisture conditions. This is what the SASMI component supplies, where the antecedent weighting with a grid-specific memory timescale carries the accumulated moisture history into the index, while the state variable itself retains the fast response needed to resolve sub-weekly transitions. The formulation is an attempt to obtain memory and responsiveness together rather than to trade one for the other, and the depth choice should be read alongside it rather than in isolation.

We nonetheless take the reviewer’s point seriously, and our formulation does not claim to capture deep vegetation water stress, which is optimized for rapid surface coupling and near-surface conditions. This is already stated in Sect. 3.9, where we note that reliance on top-layer (0–7 cm) soil moisture may not fully capture deeper root-zone constraints that govern vegetation stress and drought persistence, and where we identify root-zone incorporation as a priority for future development. We will make this trade-off more prominent in the revision and would note that the framework is fully extensible in this direction. ERA5-Land provides soil moisture at multiple depths on the same grid, so a root-zone or multi-layer variant of SSTCI can be constructed directly within a similar framework, as the authors believe (at least from preliminary analyses) that sub-surface development would require critical parameterization of the framework. Additionally, a systematic comparison across depths is a natural next step that the present, near-surface product is designed to enable. We also note that it was important for the authors to first receive community feedback based on the current form, based on which we would further develop the sub-surface products, which is currently being developed. Here, we are following the principled used by other major groups such as the H-CEL’s GLEAM which started from the surface to the root-zone, combining lessons from the community’s feedback.

Comment 3

The authors calculate the soil moisture part of their index based on SASMI which is using an e-folding timescale. The added value of this approach is not made clear. What is the benefit of using this method over using any other possibility, including using SSMI calculated using soil moisture from ERA5-Land directly? The choice again feels arbitrary, and in my opinion results in less realistic values than alternative choices.

We thank the reviewer for raising this, as it lets us clarify the physical motivation for SASMI, which we agree was stated too briefly in the original manuscript.

The added value of SASMI over an instantaneous standardized soil-moisture index is physical rather than procedural. A plain SSMI characterizes the soil-moisture state on a single day, which could be considered as a snapshot of anomaly. Compound dry–hot coupling, however, is not driven by the moisture state on any one day but by the accumulated antecedent depletion over the preceding period. Thus, it is the progressive drydown of moisture that suppresses evaporative cooling and shifts surface energy partitioning toward sensible heat, preconditioning the surface for heat build-up. Antecedent memory is therefore almost an inevitable refinement of the signal, since it is the mechanism through which soil moisture modulates temperature, while a memoryless snapshot omits precisely the quantity that governs the coupling. The 2010 Russian mega-event (Fig. 5) illustrates this directly. In that event, a sustained early Summer soil-moisture decline precedes and physically enables the subsequent heat extreme, which is a sequence that an instantaneous index would easily miss.

SASMI encodes this mechanism by weighting antecedent soil-moisture states with an exponential decay whose timescale is the grid-specific dry-down constant τ, estimated from each cell's own observed dry-down behavior. This is a deliberate improvement over both a raw daily snapshot and a uniform, globally fixed accumulation window. Fig. 1 in the manuscript shows why the latter is not defensible at global scale, where τ falls below 5 days across arid and semi-arid regions and reaches or exceeds 20 days in the Amazon, Central Africa and maritime Southeast Asia, with temperate and managed regions intermediate at roughly 7–15 days, a pattern consistent with the aridity and land-surface controls documented by McColl et al. (2017). A fast-draining semi-arid cell and a slowly buffered humid cell do not share a memory timescale, and thus, SASMI adapts the weighting accordingly rather than imposing one assumption everywhere. The authors believe that this choice is grounded in the observed physics of soil-moisture memory, which we characterize globally in Sect. 3.1, rather than being arbitrary.

On the concern that SASMI may yield less realistic values, our evaluation does not support that reading. SASMI agrees well with two independent, established daily drought metrics, SPEI and DEDI, across most global land areas (Fig. 2). We take this as evidence that the antecedent weighting preserves drought-signal fidelity, it recovers the drought behavior those metrics identify rather than distorting it, while operating directly on soil moisture at the grid cell's own memory timescale, which is what the compound coupling requires and what neither metric provides.

In the revised manuscript we will make this rationale explicit in Sect. 2.3: the distinction between an instantaneous state and antecedent memory, and the role of the grid-specific τ in setting it.

Comment 4

The authors claim that "Yet, most existing compound indices are provided at monthly or seasonal scales". This does not really do justice to the large body of literature on multivariate indices for heat-drought extremes. I refer here to Miralles et al, 2012 and Shan et al. 2024 as examples... The novelty of the study should be motivated better based on a more thorough literature review.

We thank the referee for this constructive pointe. We agree that the sentence in question was framed too broadly and did not adequately reflect the substantial literature on multivariate and compound heat–drought characterization We will revise the Introduction to give a fuller and more balanced account and to motivate our contribution against it. Nonetheless, the authors would like to clarify a couple of points about the references provided.

Miralles et al. (2012) worked on a foundational contribution to understanding soil-moisture–temperature coupling and the role of soil desiccation and atmospheric heat accumulation in mega-heatwaves. Their study thus directly reinforces the physical rationale for a soil-moisture-centric compound index, which is central to our approach, and we already draw on the closely related Miralles et al. (2014) for the land–atmosphere feedback perspective. We would note, though, that this line of work characterizes the coupling mechanism rather than delivering a multivariate compound index or an openly available compound dataset. More importantly, the goal of Miralles et al (2012, 2014) was to understand the role of soil moisture–temperature feedback during heatwaves, and not directly extendable to droughts. This is because the formalism of Pi requires a strong positive correlation of soil moisture deficit with air temperature to be significant. As a result Pi is not optimum in some cases for studying the full evolution of the compound event (personal correspondence with Diego Miralles). Thus, their study motivates our product rather than overlapping with it, and we will use it in that role in the revised Introduction.

Shan et al. (2024) is likewise relevant and closely related to our work, we adopt its removal–merging event identification procedure. It does not, however, provide a compound index in the integrated sense that SSTCI does. There, drought and heat are represented as two separate standardized series, a precipitation-based daily (SPI) and a standardized heat index (SHI), and compound events are defined through their temporal set relationships rather than by combining the drivers into a single quantity. SSTCI instead integrates the soil-moisture-based drought signal (SASMI) and the temperature signal into one continuous standardized compound variate through copula, so that the joint dry–hot dependence is carried within the index itself. The two approaches are complementary rather than equivalent.

Taken together, these references let us position the contribution more precisely, and we will expand the Introduction accordingly, acknowledging this broader literature while clarifying that, to our knowledge, no existing openly available product combines a daily, global, multi-decadal, soil-moisture-based compound index with an objectively constructed event catalogue.

Comment 5

The authors claim that a main novel aspect in their work is the catalogue of event timing and magnitude per location. I do not consider this particularly novel, since it is rather straightforward to do once the index has been defined and calculated. A more novel aspect would be to define unique spatio-temporal clusters of heat-drought events, and catalogue these rather than the just the temporal dynamics for each grid cell.

We thank the reviewer for this thoughtful point. On the catalogue itself, we would like to note that its contribution lies less in conceptual simplicity than in what its construction requires and what it delivers.

The catalogue is not a thresholding of the index. Events are identified through an objective removal–merging procedure (Fig.S1 and refer to Aftab et al., 2026a, figure 4 for better understanding) in which minor spells are removed and mutually dependent spells merged, with thresholds chosen so that the retained events are statistically independent (exponentially distributed inter-arrival times) and belong to a genuinely extreme population (GEV-distributed severities). Critically, these are not global settings: the thresholds are searched and the distributional tests evaluated cell by cell, so that every land pixel receives its own calibrated pair of criteria and its own goodness-of-fit verification. Applied across the global land surface at 0.1°, this amounts to an iterative threshold search and two distributional tests at each of several million grid cells over a 63-year daily record; a substantial and time-consuming computation rather than a single pass over the index. We adopt the procedure from Shan et al. (2024) but apply it directly to the compound SSTCI variate rather than to separate drought and heat series, so that the independence and extreme-value tests are applied to the compound quantity itself rather than to its marginals, a distinction that matters, since well-behaved marginals do not guarantee that events assembled from them are independent or extreme as compound events.

What this yield is a product that is operationally ready in a way a raw thresholding list is not. Thus, for every land grid cell globally, a direct record of discrete events with explicit onset, termination, duration and marginal severity, the form of information most relevant to impact assessment and early-warning applications, where users need to know where and when an event occurred and how severe it was rather than reprocess continuous daily fields.

We agree that coherent spatio-temporal event objects are a scientifically richer construction, and we thank the referee for the suggestion. We would note that a similar product also exists (published in ESSD), where Dheed (Weynants et al., 2025) derive spatio-temporal clusters of dry and hot events from ERA5, and that the two constructions are in any case complementary rather than alternatives. Spatio-temporal clustering operates on per grid cell event records, which are the elements from which coherent event objects are assembled. The present catalogue is therefore not a lesser substitute for clustering but a natural input to it, and one in which the constituent events have already been filtered for independence and extremeness before any clustering is applied. Extending it in this direction involves consequential choices, spatial connectivity criteria, temporal overlap rules, minimum cluster size, that merit dedicated treatment rather than being appended here, and we will note it as a promising direction in the outlook.

We also believe that this allows the data to be easily and readily usable by non-scientific users who would need this for other applications beyond scientific research and operational services.

Comment 6

Given that many different indices exist, I would see potential value in a large ensemble of heat-drought indices, which would also allow quantification of some of the uncertainties involved through the ensemble spread. Such dataset would, in my view, classify as significant.

We agree that a multi-index ensemble quantifying uncertainty through ensemble spread would be a valuable contribution, and we thank the reviewer for highlighting it.

We would note, however, that such an ensemble presupposes the existence of well-documented, validated single-index products to combine. Members must be openly available, standardized, and provided on a common grid before any meaningful ensemble spread can be computed from them, and it is precisely this that is currently lacking for daily, global, soil-moisture-based compound indices. SSTCI is released openly with its code, expressed as a standardized compound variate, and provided on a uniform global daily grid over 1961–2023, properties that make it directly usable as an ensemble member rather than only as a standalone product.

On its own merits, the dataset also fills a gap that no existing openly available resource covers: a daily, global, multi-decadal, soil-moisture-based compound dry–hot index together with an objectively constructed event catalogue. We would suggest that the two things are not in contradiction, providing a well-characterized single product is how the community comes to have an ensemble at all.

In the revision we will position SSTCI explicitly as a ready to ensemble member of a family of compound indices and note ensemble based uncertainty quantification as a direction in the outlook.

Summary

We are grateful to the reviewer for the comments that have materially improved the framing of this work. We hope that, taken together with the clarifications above, the revisions demonstrate that the dataset makes a genuine and non-routine contribution and that it is appropriate for publication in ESSD.