Rainfall response

[sumanager56]

1. Very little rainfall response in anything except very shallow soil (3.5 cm)
2. Possible reasons are low rainfall intensity, Leaf interception or ET losses from shallow zone before soil moves through

[sumanager56]

Just figured out that there's no soil moisture fluctuations at deeper depths (60 cm and 100 cm in the plots below) All the parameters are kept same as our last results on the "optimization with pest" issue.

[sumanager56]

Apparently at depth 60 cm (which is the second soil layer) soil seems to be at -15000 cm which is the PWP (there's still significant amount of soil moisture so this layer is probably being treated as clay) and no there's no root water extraction Excluding 60 cm and looking just into 4 depths - no water stress at these depths Root extracting water only from the top soil layer

Tried generating similar plots for the CN case study for corn. Thought this would be helpful to compare

[sumanager56]

One possible reason behind less response of sw in deeper depths could be root water extraction which directly controls transpiration. Root water extraction is calculated using root density distribution (Bouten, 1992) and sums up to total transpiration when integrated over the rooting depth. The crop input file (image below) shows the values for root density distribution which is similar to the one used in CN-case study for corn.

The calculated root uptake are reduced to suboptimal soil conditions: too wet or too dry and controlled by parameters shown in the image below. Athough these parameters are relative sensitive, changing these parameters didn't improve the swc response.

Interception for agricultural crops is calculated using Von Hoyingen-Hune equation. Depends on the coefficient COFAB (0-1 cm). Lower values (0) resulted in no interception while higher values (0.9) resulted in just 0.25 cm of intercepted water out of 28cm of ET. Didn't have an impact on our soil water content response at deeper depths.

[julieshortridge]

Thanks Suman - have you tried adjusting the root density distribution to encourage more uptake from deeper horizons?

[sumanager56]

Yes, Dr. Shortridge. Changing root density distribution so that more uptake from deeper horizons would happen did seem to improve the result. First image below is root water uptake from the best optimized version (alpha and npar) which shows very less water uptake values and no uptake below 40 cm (first soil layer in our discretization). Root water uptake and pressure head (very low at 60 cm) before changing root density distribution

When I increased the root distribution values on deeper soil layers, the overall root water uptake increased and there's some uptake from 60 cm soil layer as well. The uptake behavior of different layers is somewhat close to the CN module example. Pressure head and root water uptake with updated root distribution values (higher values in mid layers)

In the above plot, the water uptake values still seem to be very low (comparing to uptake in CN example). I checked the transpiration and evaporation ratio, and found it to be less (15cmT:10cmE). I reduced the CFBS ratio in the soil evaporation section (image below) from 0.5 to 0.25. This reduces the portion of potential soil evaporation from the PET (and ultimately actual evaporation values as well). By doing so, and running the optimization process again for new alpha and npar, the transpiration and evaporation ratio was (23T:9E) which seemed more reasonable to me. This also increased overall crop water uptake values but still deeper depths do not look better. Pressure head and root water uptake with updated root density distribution and increased transpiration using CFBS=0.25

The latest alpha and npar values (optimized) are shown below

It can be seen that alpha values for the 3rd layer (55-85 cm) is relatively higher. This is the layer where root water extraction is very low (should contribute higher than 3cm soil if we look into CN example). Lowering the alfa values or changing Npar is highly sensitive to changing pressure head and root water extraction. For example when I reduced alpha in layer 3 from 0.15 to 0.015, please see the image below (we can compare this to the last two images above and see around july and mid august, water extraction and head is improved after lowering alpha in this layer). Alpha manually reduced from 0.15 to 0.015 in soil layer 3 (I feel there's a need to include one of the deeper layer (40-60cm) in the overall optimization process so that better alpha values and npar would be obtained for deeper layers as well)

The same way, if I manually adjust Npar value for layer 3 from 1.81 to 1.21 to the above final plot (after root, CFBS, alpha change), the root extraction gets even better. Overall I think my next step should be including at least one deeper soil layer (>40cm) in the optimization process and try to find best alfa and npar values that can encourage better root water uptake from all depths as expected I am also including the latest version of soil water content plot that I have right now (roots and CFBS updated). These are all using just the 20cm depth observed soil water content for optimization (2 parameters-alpha and npar) SSE=0.166; R2 = 0.51; NSE=0.5 for 20cm soil water content

[sumanager56]

@julieshortridge Dr. Shortridge, I was working on the met data to create a detailed rainfall intensity file for input into SWAP (SWRAIN=3 option). I am able to do so pretty easily for WeatherSTEM data since they were cumulative readings - For Hobo data, (which fills missing weather data for September) I have to do it separately since rainfall data are not cumulative. I am still working on that and will soon merge it with the Weather STEM data and upload it into SWAP. While working with the Weather STEM data, I figured out that I was using the rainfall data wrongly. Instead of using the total rainfall, I was actually using the mean of the rainfall data for any particular day. I had that correct in one my previous codes but I am not sure how I ended up with the mean. Also, I might have been mistaken because cumulative rainfall value occurs at midnight, which is actually the next day that didn't make sense to me when I compared the total rainfall for a particular day and last rainfall value for that day (expected it to be same but is not the case). With the corrected rainfall data (Weather STEM from Jan1-Aug31 + Hobo data from Sep1-Oct10 + WeatherSTEM from Oct11-Dec31), following are the plots of soil moisture at different depths after optimization of alpha and npar values.

After optimization (only using 20cm depth right now) with alpha and npar (roots distribution I have kept same as default-CNmaize exaple for now)

[julieshortridge]

Excellent, thanks Suman! That certainly looks a lot better. Can you add in an line for observed soil moisture readings to your 60cm plot?

The soil moisture is looking much better. The only issue that jumps out to me is that the simulated 10cm VWC is not drying out as quickly as observed, and I wonder if that might be contributing to the relatively high root water extraction coming from shallow soil in your last plot. But I suspect that's a relatively small influence.

The other two things that jump out to me about the root water extraction plot: 1) It is still showing maximum root water extraction in the shallow soil - by June/July I'd expect to see something similar to your CN example plot where most water is come from deeper levels in the profile. 2) If you summed all the extraction across depths, there doesn't appear to be a peak in July-August which I'd expect to see (also apparent in the CN case study file).

What I would suggest at this point is keep your soil parameters as they are for these plots you just posted, and focus on the root density and ET parameters to see if you can get the root water extraction plots more reasonable. Then once we've done that, I think we can do the complete calibration (multiple depths) in PEST. What do you think?

Also to make your plots of root water extraction and pressure head easier to read, could you make the lines a little heavier in both the plot and the legend (usually lwd = 1.5 or 2 will help) and adjust the colors so they're easier to see (maybe the grey line could be blue), and then order the depths in the legend so it's by depth - eg from left to right 3 cm, 10cm, 20cm, etc..

[sumanager56]

Thanks, Dr. Shortridge. I have updated figures as per your comments! First, I tried the SWRAIN=3 option and the plots look identical, so we can just proceed with this option for now. Changing root density helps root extract more water from deeper depths compared to shallow but the overall water extraction value do not improve (please see the plot below after I increased density values at deeper depths). Changing ET parameters help change the ratio of E/T but do not increase the overall root water uptake values or the contribution by deeper depths. Updated density distribution with higher density values at deeper depths

Again these all plots are obtained using 20cm optimization with alpha and npar

[julieshortridge]

Thanks Suman - so comparing your most recent root water extraction graph with the CN case study graph towards the top of this thread, it looks like the raw values in your simulation are about an order of magnitude lower than the case study, and we're still missing the seasonality. To easily compare those two simulations, it would be helpful to see a couple things:

• For the root water extraction plot for your simulations (not the case study), can you extend the time period back to the planting date?
• Can you create a summary plot that shows a single line for total root water extraction across all depths through time for your simulation plus the CN case study?
• Can you add a table showing the seasonal water balance for your simulations and the CN case study?

[sumanager56]

Dr. Shortridge, Thanks for pointing out an error in my root extraction plots. The root water extraction values for our simulations (TWRC) are 1cm scale values, and those for the CN example are larger scale (~10 cm scale) since they have fewer compartments in their discretization. That's why our root uptake was around an order of magnitude lower. Here's the plot for total root extraction across all depths through time for both the studies. I have extended time period back to April 20 (planting date). Total root water extraction across all depths Water balance for TWRC The cropping period for both the studies are almost identical - end date is sep13 vs 23

Water balance for CN example. I just ran the CN example for the last year 1982 April 20-Sep 23.

[julieshortridge]

Ok, thanks Suman. This all looks pretty reasonable to me, the only thing I'm still a little curious about is why some of our water balance numbers (e.g. rainfall, transpiration and evaporation) are so much larger than theirs. However, I think the numbers in your water balance seem more appropriate for our location (for instance, 66 cm of rainfall = 26 inches of rain versus 9.6 inches in their example), so maybe we can chalk that up to regional differences.

Let's confirm with Dr. Stewart this afternoon, but based on what you've shown me here I think it would be reasonable to move forward with the full calibration in PEST.

In the meantime, can you show us what the yield for this simulation and the CN example are? I'm just wondering how it will compare to the the synthetic met data simulations you ran before. Looking back at some of the graphs you made before it looks like those were averaging around 100 bu/acre yield, so I'm curious if the improvements you've made in the soil moisture representation here change that at all.

[sumanager56]

Thanks, Dr. Shortridge. When I looked into the yield with current simulation, it was way too low compared to the potential (please refer to first plot below). There's an option to flag on or off the Nitrogen Simulation in SWAP -* flag for nitrogen in crop and soil flCropNut = .TRUE./flCropNut = .FALSE. When I turned off nitrogen simulation (flCropNut = .FALSE.), the yield looked fine - actual yield was just underneath the potential yield line (plot 2). Something was going on probably with N simulation parameters. I just checked briefly and there's a parameter NFIXF (fraction of crop nitrogen uptake by biological fixation) set to 0. Turning it to 0.1 increased the yield significantly (plot 3 below). flCropNut = .TRUE.**

flCropNut = .FALSE.

flCropNut = .TRUE.

flCropNut = .TRUE and NFIXF-0.05 gave more reasonable yield

CN case study example for 1982. Their NFIXF was set to 0