the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Feb 2025
Characterizing precipitation and soil moisture drydowns in Finland using SMAP satellite data
Abstract. Precipitation (P) and soil moisture (SM) are critical components of the global water, energy, and biogeochemical cycles, yet their patterns and interrelations in the Arctic are poorly understood. Due to the sparse in situ measurement network, satellites are the only way to observe P and SM in high-latitude regions. This study uses NASA’s SMAP satellite to analyze the relationship between SM and P, assess the feasibility of estimating P from SM, and examine SM drydown patterns in Finland from April to September over 2018–2019. The analysis reveals a notable spatial and temporal variability in SM, with a weaker correlation between P and SM in spring due to snowmelt and a stronger relationship in summer and fall. Water bodies complicate the SM retrieval causing the SM retrievals to saturate. Using the SM2RAIN algorithm, we estimated P from SM data. The algorithm shows promising results, detecting the area of rainfall accurately in most cases but estimating the intensity of the rainfall is more challenging, particularly for light and very heavy rain. We analyzed SM drydown patterns by fitting an exponential model to each SM drydown period, from which we estimated the exponential decay time scale (τ) and the lower bound of SM (SMmin). τ does not show much spatial or temporal variability. The distribution of τ is positively skewed, with a mode of 1.6 days and a median of 4.0 days, consistent with other studies. The distribution of SMmin is also positively skewed, with a mode of 0.14 m3 m-3 and a median of 0.17 m3 m-3. SMmin exhibits another lower peak at 0.02 m3 m-3, the lower limit of SMAP SM retrievals, possibly causing an artifact in the results. SMmin shows spatial variability, with the lower bound being slightly higher near water bodies but also showing a more prominent peak at 0.02 m3 m-3. Grid cells with dense vegetation and low vegetation agree better with each other, indicating that water bodies particularly affect and complicate SM retrieval. The promising results suggest that the method could be applied across the entire Arctic region.
- Preprint
(5143 KB) - Metadata XML
-
Supplement
(1078 KB) - BibTeX
- EndNote
Status: open (until 31 Mar 2025)
-
RC1: 'Comment on egusphere-2025-245', Anonymous Referee #1, 18 Mar 2025
reply
The paper reads well and is interesting. It covers a topic not always fully assessed but I never the less have questions about a few facts. For this reason I think maybe the manuscript should be resubmitted (hence my reject suggestion) to account for the issues raised in the first two points. But I agree it could be only major revisions should the authors be able to address the 2 issues
Maybe the most striking is why is it important to assess rainfall in Arctic environment where – to my understanding – most of the water comes under the form of solid precipitation. In other words what is the real impact of assessing exactly liquid precip and how the errors in liquid precip relate to uncertainties in solid precip.
A second big question is about the neglecting of run off. (line 169). To me, but I might be wrong especially for Arctic areas, most of the run off occurs during rainfall or immediately after, so it cannot be negligible.
I also have question on some aspects of the range of soil moisture. Maps show SM in excess of 0.7 m3/m3 but (figure 4) but I doubt the field capacity is higher than 0.5 m3/m3. So even the range seems excessive. It could corresponds to flooded areas but then the range being minimal, it would correspond to water bodies. But surely SMAP sees and flags water bodies (otherwise SM estimates are bound to be wrong) so what is it exactly? The authors might want to elaborate on this as it is most intriguing.
I am not sure also I understand the lower limit of 0.02 m3/m3 for SMAP Does it mean that SMAP SM estimates are never lower than this value? And if yes what is the rationale for this.
I have also a question for the authors on the choice of SM2RAIN. As here are some limitations linked to the assumptions of the approach (such as the one made above). So why use such algorithm? Why not use more robust approaches assimilating SMAP data in a simplistic model to infer rainfall? Of course such approaches require a initialisation through a first guess precip usually from systems such as IMERG. Did the authors consider such approaches which should be more reliable and why did they or did not? I am thinking of Pellarin et al for instance.
Line 112 a point of detail, the native spatial resolution of SMAP is not 36 km (it is an ellipse) but rather 39x47 km according to the SMAP handbook. So this statement is somewhat misleading. As very rightly indicated the 9 km grid corresponds to oversampling but this is not clear in table 1 where the resolution is indicated as being 9 km. The latter should be corrected.
Citation: https://doi.org/10.5194/egusphere-2025-245-RC1 -
RC2: 'Comment on egusphere-2025-245', Anonymous Referee #2, 18 Mar 2025
reply
Review of Characterizing precipitation and soil moisture drydowns in Finland using SMAP satellite data
In this paper, SMAP SM data are analyzed in the Arctic region (specifically in Finland) to assess their capability in estimating SM in this complex environment, explore the relationship between P and SM, estimate precipitation from SM using SM2RAIN algorithm and examine SM drydown pattern from April to September in the years 2018-2019. The analysys shown promising results, with good correlation between SM and P, excluding frozen and snowmelt period; limitations of SM quality in the proximity of water bodies and good spatial accuracy of the SM2RAIN rainfall product (but low intensity accuracy, with overestimation of low rainfall and underestimation of the heavy rainfall). Moreover, the length and the exponential decay time scale of the SM drydown were analyzed, showing stable results both in time and space.
The paper is well organized and well written, the language is fluent and clear.
However I have a few important points that need to be answered before the paper is accepted for publication. Specifically
Major comments:
Lines 179-172 “Using…. 2014)”: Not clear. Hence the SM2RAIN algorithm was not calibrated? This approach is not correct: SM2RAIN parameters need to be calibrated to each pixels, as they are dependent from the soil characteristic of the pixels. All the works of Brocca et al. after 2013 relied on this, to overcome the limits of 2013 paper. SM2RAIN should be calibrated against observations (without considering SM data obtained in frozen conditions). If the Authors prefer to not calibrate the data, they should still use some regionalization procedure to obtain SM2RAIN weight for each pixel (e.g. Filippucci et al 2021; 2022, here applied to ASCAT or S1 data). The parameters can change greatly according to the area. Moreover, this approach do not consider SM2RAIN temporal filter T. This would be important to assess the real performance of SM2RAIN in the area.
Moreover, SM2RAIN use interpolated daily data, normalized between 0 and 1. Did the Authors applied interpolation and normalization? It should be written and, if not applies, please explain why.
Regarding SM2RAIn quality, the error introduced by the snowmelt was already analyzed in Brocca et al., 2019, which applied mask to sm data during melting (0<T<3) in order to avoid this kind of noises. This should be implemented also here, to assess the very capabilities of NASA SMAP data.
Line 419 “while dense….challenge”: I don't think that this conclusion is supported by this analysis. Dense vegetation show a realistic pattern, but it is not verified that the SM beneath the vegetation is actually sensed. One way to obtain a confirmation would be to use SM2RAIN to evaluate precipitation time series for all the pixels and then assess its performance per pixel (with indices as RMSE or Pearson correlation or BIAS). If the results in the densely vegetated area are similar to those around, then this indicates that the capacity of SMAP to estimate SM under vegetation are good. But I expect to find worse correlation. In general, I understand that the lack of SM data allow to validate SM just in few areas, but precipitation data are available for all the study areas, hence I encourage to use SM2RAIn to gain more information regarding the SM data quality
Minor comments:
Lines 55-57 “While these….2019)”: This is not completely true, as, for example, brocca et al. do not exclude artic region, just periods in which the soil is frozen. Check and correct
Lines 167-169 “This….site”: Indeed, the discrepancies in SM between the two sites could be related to different precipitation pattern in the large SMAP pixel. However, it could be also related to error in SM measurement. Since Radar data have 250 meters spatial resolution, you could compare the radar Precipitation obtained in the pixels nearest to the Hyytiälä stations, to ensure that they are correlated with it and therefore there is no error in the observed SM measurements.
Line 363 “Figure…. on SM”: Not clear, please expand the reasoning behind the sentence
Lines 186-187 “Our analysis….Fig. 7),”: To be checked after proper SM2RAIN calibration (see major comment #1)
Citation: https://doi.org/10.5194/egusphere-2025-245-RC2 -
RC3: 'Comment on egusphere-2025-245', Preet Lal, 18 Mar 2025
reply
This study investigates the relationship between precipitation and soil moisture in Finland using satellite-based data from NASA’s SMAP mission, supported by ground-based radar and in situ measurements. It applies the SM2RAIN algorithm to estimate precipitation from soil moisture and characterizes soil moisture dry down patterns by fitting exponential models. The analysis covers two years (2018–2019) and highlights spatial and temporal variability in soil moisture, as well as challenges posed by snowmelt and water bodies in SM retrievals. The findings are valuable for understanding Arctic hydrological processes and suggest potential for broader application across the high-latitude regions.
The study is scientifically sound, methodologically thorough, and clearly communicated. It effectively combines multiple data sets and provides insight into soil moisture dynamics in a challenging environment. The use of dry down modeling and SM2RAIN adds value, and the discussion acknowledges key limitations transparently. Here are comments for further improvement before publications.
- The SM2RAIN algorithm is applied assuming that evapotranspiration and runoff are negligible during precipitation events. While this assumption is common, it may not always hold true in regions such as Finland, particularly in wetland areas or under light rainfall with significant canopy interception and latent heat fluxes. It would strengthen the study to assess the potential bias this simplification introduces, possibly by incorporating flux tower ET data or ERA5-Land evapotranspiration estimates.
- The use of fixed empirical parameters (Z, a, b) from Brocca et al. (2014) across all grid cells is practical but may not adequately capture the spatial variability in soil hydraulic behavior. Given the known heterogeneity in Finnish soils, vegetation, and hydrological settings, a spatially stratified parameterization based on land cover or soil texture (e.g., from SoilGrids or FAO datasets) would be more representative. Alternatively, local calibration using available in situ or radar-based precipitation data could be tested in a subset of locations to explore the potential for improved accuracy.
- The exponential decay model used for drydowns is based on a first-order assumption that may oversimplify soil water loss processes in certain conditions, especially where rooting depth, vegetation type, or soil layering affects moisture dynamics. Exploring alternative drydown functions—such as bi-exponential or piecewise linear decay—on a subset of events could help determine whether the exponential model remains robust across regimes.
- The study uses the SMAP L3 enhanced product oversampled to a 9 km grid, though the native resolution remains approximately 33 km. This discrepancy could lead to spatial artifacts or smoothing, especially near land-water boundaries or heterogeneous terrain. Clarification is needed on whether any additional spatial processing or filtering was applied to the SMAP data prior to analysis. It would also be useful to discuss the potential influence of footprint overlap on statistical independence in the drydown fitting.
- Only the descending (6 a.m.) SMAP overpasses are used, which is justified in terms of stability and retrieval quality. However, excluding ascending passes may overlook sub-daily variability—particularly relevant in summer months when diurnal SM cycles can be significant. The authors are encouraged to briefly quantify potential information loss from this exclusion or justify it further using SMAP data quality metrics for ascending overpasses.
- The approach of averaging in situ soil moisture within SMAP footprints is appropriate, but spatial heterogeneity within grid cells—especially in complex environments such as fens versus forests—could introduce representativeness errors in the validation. A quantitative assessment of subgrid variability (e.g., standard deviation or interquartile range across in situ probes) and its impact on satellite comparison would strengthen the validation component of the study.
- Surface classification based on SMAP retrieval flags is a useful step; however, the interpretation of SMmin and τ values could be improved with more granular land surface information. Adding categorical breakdowns by vegetation type or soil class could help explain observed spatial patterns and potential retrieval biases, particularly in regions affected by water bodies. Additionally, it would be helpful to explicitly state the proportion of SMAP observations excluded by quality flags and assess whether this introduces systematic seasonal or regional sampling bias—e.g., due to frozen conditions or persistent cloud/snow cover.
- The springtime SM–P relationship is notably weaker due to snowmelt contributions to SM increases. While this is acknowledged, a more detailed discussion or methodology for separating snowmelt-induced SM changes from rainfall-driven changes would be beneficial. This could involve the use of modeled snowmelt fluxes from ERA5-Land, or analysis of coincident snow cover and air temperature data.
- The study is limited to two years of data (2018–2019), which are suitable for identifying seasonal patterns but may not capture interannual variability. It would be useful to briefly contextualize whether either year exhibited anomalous precipitation or temperature relative to the climatological baseline or mention how a longer time series could help address questions of variability or trend detection.
- Given the stated potential for Arctic-wide application, the manuscript could benefit from a short discussion on the method’s suitability for permafrost-affected regions. Permafrost soils introduce unique SM dynamics, including seasonal freeze-thaw cycles, active layer constraints, and limited drainage. Whether this methodology could be adapted or tested in northern permafrost zones (e.g., in northernmost Finland or elsewhere) is worth exploring.
- The use of the term “saturation” to describe SMAP retrievals near water bodies may be misleading for readers unfamiliar with remote sensing. Clarifying whether this refers to sensor saturation, retrieval floor/ceiling effects, or actual physical saturation of the soil column would improve clarity.
Citation: https://doi.org/10.5194/egusphere-2025-245-RC3
Data sets
SMAP Enhanced L3 Radiometer Global and Polar Grid Daily 9 km EASE-Grid Soil Moisture, Version 6 P. E. O'Neill et al. https://doi.org/10.5067/4DQ54OUIJ9DL
Finnish Meteorological Institute Open Data Finnish Meteorological Institute https://en.ilmatieteenlaitos.fi/open-data
The Sodankylä in situ soil moisture observation network J. Ikonen et al. https://doi.org/10.5194/gi-5-95-2016
Hydrological data measured in Lompolonjängänoja (LJO) catchment J.-P. Nousu et al. https://doi.org/10.5281/zenodo.10820563
SMEAR Värriö and Hyytiälä SmartSMEAR https://smear.avaa.csc.fi/
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 225 | 33 | 7 | 265 | 27 | 5 | 5 |
- HTML: 225
- PDF: 33
- XML: 7
- Total: 265
- Supplement: 27
- BibTeX: 5
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 87 | 32 |
| Finland | 2 | 24 | 8 |
| China | 3 | 19 | 7 |
| France | 4 | 17 | 6 |
| Netherlands | 5 | 14 | 5 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 87