Forecasts of Opportunity for Northern California Soil Moisture
Round 1
Reviewer 1 Report
The authors investigated the predictability of soil moisture in northern region of California using Linear Inverse Modeling method. The paper is logically written and the results are discussed sufficiently. I would recommend this for publication after some minor adjustments.
The representation of figure 8 needs improvement. It might be because of the distributed pdf format. If not, please try to add a legend to this figure and change the labels, tick marks, and station codes format to the figure 7 font size.
Please add a label to the Figure 9 colorbar.
Author Response
Response to Reviewer 1:
Figs. 8 and 9 are now approximately the same size in the manuscript. Unfortunately, we were unable to use the same software to generate them both, but the new version of Fig. 8 should look much more similar to the former Fig. 7 than it did.
Labels have been added to the color bars in Fig. 9.
Reviewer 2 Report
Paper review on “Forecasts of opportunity for northern California soil moisture”. Soil moisture is an essential state variable of land and atmosphere interactions. This paper mainly investigated the capability of the Linear Inverse Modeling (LIM) which is a data-adaptive technique, to forecast soil moisture in northern California by using soil moisture and temperature measurements from 10 stations. The results indicate the LIM maintains skill in soil moisture hindcasts for lead times of 1-2 weeks. The topic is important and the presented results are potentially interesting to the soil moisture research community. The paper will be suitable for publication in Land after the following issues being addressed properly. 1) The predictability of soil moisture was tested by LIM using ground soil moisture measurements from 10 stations. In addition to ground data, several satellites/sensors (e.g., SMAP, SMOS, AMSR2, ASCAT) can provide soil moisture information (i.e., soil moisture products) worldwide. It is not clear if the LIM method can directly applied to satellite soil moisture for forecasting or other assumptions are need? It would be very interesting if the authors can do some test regarding this issue since using satellite soil moisture products can avoid reliance on ground measurements and can forecast soil moisture on a larger scale. I will leave the decision on this point up to the authors. 2) Aside from precipitation and temperature, the spatial variability (even temporal evolution) of soil moisture is also affected by other factors, such as land cover (e.g., vegetation coverage), climate type, soil texture and topography. Since the predictable time of different stations is different, did you investigate this issue concerning the environmental factors? This may help you understand the different performance of LIM in different locations. Other minor comments: 1) Line 60: a typo here, should be “assumption” rather “assumptiion”. 2) Fig.1: add longitude and latitude information for this figure. It is also better to add the land cover for the study area. 3) Table I: there are some formatting problems, please correct them. 4) Equations: Please use the correct format for all equations.Author Response
Response to Reviewer 2:
Point 1:
We have added several paragraphs to the Introduction in answer to Reviewer 2’s Point 1. Specifically, we point out that using satellite data may indeed be more consistent with previous applications of LIM. However, we have applied LIM to station data for several compelling reasons. Most importantly, as workers in government laboratories, it is our mandate to supply operational entities with methods of expanding products already provided to stakeholders. Forecasters at the American National Weather Service and California operational hydrologists are heavily invested in using the station data analyzed in this study. Thus, it was imperative for us to investigate the feasibility of applying LIM to these point observations. For reasons now discussed in this manuscript, we do envision applying LIM to satellite data in the future; however, that will be addressed in a future study more concerned with investigating the physical nature of the hydrologic system rather than the monitoring of drought and flash floods.
Point 2:
There are many variables that affect soil moisture that were not included as predictors in this application of LIM. Given the short time series available to us, we needed to keep the predictor set as parsimonious as possible. Fortunately, the fact that the tau test (Fig. 3) is passed as well as it is (for comparison, please see Fig. B1 is Penland and Sardeshmukh 1995, cited in the manuscript) leads us to believe not only that the dynamical system is linear, but also that most of the prediction signal in these other fields at the depth we consider (10cm) is contained in the soil temperature and moisture. In particular, we do not consider precipitation as an explicit predictor. Rather, any predictable signal in precipitation is integrated through the soil and included implicitly in the soil moisture. Of equal importance, of course, is the unpredictable signal in precipitation, which is a large part of the stochastic forcing. This is discussed in the manuscript, both in the Introduction and in section 2.3, where we discuss the results of the tau test.
Minor points: Thank you for pointing out the typos. We have replaced Fig. 1 and believe the formatting problems in Table 1 have been corrected. When we typed the equations, we used either the “Insert equation” tab provided by Microsoft Word or, if no symbols were required, we simply typed them from the keyboard, as directed in the instructions for authors. Please do let us know if any equations are unreadable. Thank you.
Round 2
Reviewer 2 Report
The authors have addressed my comments properly, and the paper can be published.
There are still some formatting problems in Table 1 and the equations in the revised version, and I suggest the authors correct them when submitting the final version.
