Subject: climate data + some help Posted: 9/11/2024 Viewed: 1533 times
Hello everyone,
I am currently working on a project involving a basin in Crete, Greece, and I would greatly appreciate any guidance or assistance.
I have precipitation data from four stations and am using the Simplified Rainfall-Runoff method. However, I am facing some challenges due to the uneven distribution of rainfall, which is influenced by significant elevation changes across the basin.
When I imported a NetCDF file with data from just one station, the modeled streamflow values fit my observed data very well. However, when I used precipitation data from a different station, it overestimated the streamflow. My question is: How can I solve this issue and ensure that the precipitation values are more representative so that my modeled streamflow better matches the observed data?
Additionally, my study area has a karst topography, but I do not have any groundwater model to input data from. I am unsure how to address this, as groundwater may play a significant role in the hydrology of the basin.
Finally, I am curious about how the demand sites influence the streamflow. In the Simplified Rainfall-Runoff method, the streamflow is modeled as inflow minus outflow. My understanding is that inflow comes from precipitation and outflow is largely represented by evapotranspiration (ET). I would appreciate clarification on how demand sites should be factored into the streamflow calculations.
Thank you in advance for any advice or suggestions you can provide. I look forward to hearing from anyone with experience in similar situations.
Mr. Doug Chalmers
Subject: Re: climate data + some help Posted: 9/30/2024 Viewed: 1091 times
Ioulia,
Apologies for the slow response. Here are some answers to help-
Were your catchments created by catchment delineation mode with different elevation bands? If so, the climate datasets within WEAP will predict different precipitation amounts at different elevations. Consider saving a different version of your model and then using the precipitation data from the WEAP datasets. You could then look at the ratio between the different elevation bands versus the elevation of your precipitation station. For example, if the WEAP datasets say you get 1000mm/year at 250m where your observed station is location and 1500mm/year at 500m, then you could set the elevation band in your model with the NetCDF to multiply the precipitation from the station * 1.5. You can also look to literature for more information on interpolating between climate stations.
Matching your modeled to observed streamflows is a larger question of calibration, which is tricky. Resolving your precipitation will help. For information on additional fixes to calibrate, please see the Catchment Calibration chapter of our WEAP Tutorial at: https://www.weap21.org/tutorial/.
To think about the influence of karst topography, the most effective tool to help you will be local data that has studied this. If you do not have this data, you will have to rely on literature or studies that have studied basins with similar geology. The information in the Catchment Calibration tutorial mentioned above is also relevant here.
Finally, in Simplified Rainfall Runoff, the basic equation is that Runoff = Precipitation - ET. To model the effect of demands, you can either create a demand site, or right click on your catchment and select "Includes irrigation". During calibration, it will be important to re-create the historical streamflow as best you can, including any historical demands that would have occurred in the observed flow record. This is different than the natural/unimpaired flow, which would not have any human demands.
I hope this all helps. Please let me know if you'd like additional clarification.
-Doug
Ms. IOULIA KOROPTSENKO
Subject: Re: climate data + some help Posted: 10/17/2024 Viewed: 940 times
Thank you for your previous response, Doug. It has been quite helpful.
I am facing a challenge with data availability, as I have five stations in my study area, but only one provides representative data for the basin. The other four stations have significant gaps in their data records. What would be the best approach to address this issue?
Additionally, I am considering creating a groundwater node in WEAP and linking it to my catchment node via an infiltration link specifically for modeling in a karst environment. Would this approach be effective? If so, how should the amount of infiltration in karst areas be calculated, and how can I determine the portion contributing to surface runoff versus groundwater recharge in such a context?
Furthermore, we have not yet defined any karst characteristics or incorporated recharge from the springs in our model. However, the results we have obtained -- based solely on precipitation and evapotranspiration (ET) data -- yield an R^2 of 0.65. How is this possible? We also have not implemented any crop coefficient (Kc) values yet. Could you provide some insights on this?
Thank you in advance for your help
-Ioulia
Mr. Doug Chalmers
Subject: Re: climate data + some help Posted: 10/17/2024 Viewed: 927 times
Ioulia,
A lack of quality data is a common issue when building a model. As modelers, we must advocate for new studies of needed data, but while we wait we must make the best use out of the available data and make assumptions when needed.
When calibrating to flow stations, in general you should start calibrating to flow stations that are upstream and are largely unimpaired without significant human diversion. Calibrate each of those stations with the best data available and then move downstream and adjust your calibration as needed to match the downstream gauges. You will have to match the different sets of years to the different gauges depending on the data available. This is common.
Yes, I would think that a karst topography would have significant percolation to the groundwater from precipitation. A catchment node connected to a groundwater node would capture this process. You should use the soil moisture method for your non irrigated catchments and calibrate the model to match your surface flows and produce realistic patterns of the soil moisture result. Completing the calibration without good data is a common challenge. In general, the best practice is to 1. Use available local data or literature, 2. Use data or literature from a different geography, but which you think is similar to your basin, or 3. Calibrate the hydrology parameters to reproduce gauge surface flows and create a realistic pattern of soil moisture in the shallow and deep groundwater layers
Sometimes, we can reproduce observed flows, but with the model not recreating the processes in the correct way. This can happen by luck, but then your model won’t work right for other climate years. Again, check the soil moisture to see if the hydrologic processes are realistic.
Hope this helps. Calibrating hydrology without good data in areas with complicated geography is a significant challenge.
-Doug.
Ms. IOULIA KOROPTSENKO
Subject: Re: climate data + some help Posted: 12/5/2024 Viewed: 467 times
Doug,
In our case, we initially tried the Rainfall-Runoff method without implementing karst topography and achieved an R² value of 0.65. For the Soil Moisture method, however, we currently lack specific parameters for our basin. We have a single catchment, and the largest area of this catchment includes irrigated areas with specific crops. However, we do not have precise data on the crop areas—only the estimates provided by the official governance authority for this matter.
Given these limitations, could you advise us on how best to approach this issue?
Looking forward to your insights!
- Ioulia.
Mr. Doug Chalmers
Subject: Re: climate data + some help Posted: 12/5/2024 Viewed: 453 times
Ioulia,
A couple of recommendations-
I would recommend to use a separate catchment for your irrigated area, and to use the MABIA method for this irrigated catchment. For calibrating your hydrology in your non-irrigated area, it is almost always the case that the specific parameters for the soil moisture method do not exist. You will need to calibrate these values using the steps I mentioned above and in the Catchment Calibration tutorial chapter. It is also often very common that imprecise estimates for demands and crop areas, such as what you mentioned above, are the best data source available. One thought is that you could compare your crop areas to the agriculture area found by automatic catchment delineation in WEAP.
In modeling, there is really no such thing as a "perfect" calibration. There are always uncertainties as to the accuracy of our input data, and as a result, there are uncertainties around the results of our model. It is often the case that the calibration process, not when we have achieved the ideal fit, but rather that we have simply run out of time! Think about the goals of your model, and ask yourself the level of accuracy and detail you will need for your model to achieve your goals. "All models are wrong. Some models are useful" is a common saying!
Good luck!
-Doug
Ms. IOULIA KOROPTSENKO
Subject: Re: climate data + some help Posted: 12/16/2024 Viewed: 330 times
Doug,
Thank you once again for all the valuable advice you have provided. I truly appreciate your insights, and I plan to thoroughly review and implement your recommendations as I continue working on my current project. However, as this process will require some time and comparisons with realistic data, I should mention that we are facing a limitation in the available data. Specifically, the in-situ data we have is only available for the two years, which may pose some challenges for accurate analysis.
Before proceeding with your suggestions, I would like to seek further clarification on a few points, particularly regarding the simplified rainfall-runoff method we have already implemented. We are already working with the simplified rainfall-runoff method, which we will need to present soon. Below are a few questions I would appreciate further clarification on:
1. Implementation of Infiltration Links:
As previously mentioned, our project involves a complex karst topography with numerous drilling wells and springs that contribute to our allocation system. Given the Karst characteristics, I understand that incorporating an infiltration link could be beneficial. However, I am slightly confused about whether it is necessary to add infiltration links to all of these nodes (springs and wells). Would this complicate the system further, or is it a necessary step to account for the karst features adequately?
Additionally, the only karst-related data available to us consists of annual values in percentage form, which apply to the entire system as a whole. This percentage represents the overall infiltration characteristics of the karst terrain, but we are uncertain how to apply it specifically to the springs and wells within the system. Other studies in similar regions have used Karst SWAT models to simulate groundwater conditions for the springs and wells, incorporating detailed infiltration characteristics. Unfortunately, we do not have access to a groundwater model like SWAT, so we are left with the challenge of manually implementing this infiltration data. Could you suggest any approaches or methods for how we might integrate this percentage-based infiltration data effectively into our model, especially for the springs and wells?
2. Springs and Groundwater Recharge:
The springs and groundwater recharge have been connected to the stream using transmission links. However, it has been observed that there is no significant change in the overall system balance. Specifically, the streamflow closely mirrors the precipitation levels, with only minimal differences. Could you kindly provide some insight into the potential causes of this issue? Are there concerns regarding the linkage of the springs and groundwater recharge to the stream within the model, or might other factors be influencing the results?
I would be grateful for your further input on these points. I look forward to your guidance as I continue refining the model.
Thank you once again for your help!
-Ioulia
Mr. Doug Chalmers
Subject: Re: climate data + some help Posted: 12/16/2024 Viewed: 322 times
Ioulia,
If you are using the simplified coefficient method with your catchments, then you will need to set the runoff fraction to your various infiltration links in order to specify the fraction of runoff going to surface vs. groundwater. Draw any infiltration links from your catchments to the river objects and to the groundwater objects for any catchments where you want to model the runoff and infiltration. Then, in the Data tree, go to Supply and Resources --> Runoff and Infiltration, find your links for the given catchment, and then set the Runoff Fraction parameter to specify the fraction of runoff going to each. You can set a single annual value, which may be related to your data, or a more complex expression related to monthly timing or precipitation. Setting a more complex expression could be manually done in a trial-and-error approach to aid your simulation, could be based on additional literature, or done based on a conceptual understanding that more runoff generally goes to groundwater or surface water during specific times in your basin.
To simulate surface-groundwater interaction, go to your Groundwater object in the Data Tree, and set the Method to Model GW-SW flows. Follow some of the instructions detailed within the Hydrology chapter of the tutorial on GW-SW interactions, and then you should see inflows from groundwater to surface water when groundwater levels are higher, and inflows from surface water to groundwater when groundwater levels are lower. A transmission link would divert water from the groundwater object to the stream when it is requested by a downstream demand. However, it would not simulate the hydrologic equilibrium of the water table- this is done with the above steps instead.
Good luck!
-Doug
Ms. IOULIA KOROPTSENKO
Subject: Re: climate data + some help Posted: 1/14/2025 Viewed: 50 times
Doug,
Thank you very much for your previous insights and recommendations. I have spent some time since then trying to implement your suggestions, but it appears to be more complex than initially anticipated.
The primary challenge is that we do not have any gauges installed in the area, nor do we have reliable historical data on streamflow. The data available is quite limited and lacks precision. Regarding soil moisture, we are unable to reproduce soil moisture patterns as this data is also unavailable. Similarly, implementing the fraction numbers you mentioned has proven to be quite difficult.
To provide some context, our basin relies on springs and withdrawals as the main sources of water supply. We have implemented demand and supply nodes in our model, and the data we currently have includes monthly spring inflows and withdrawal rates. However, we have not adjusted our withdrawal nodes to account for the karst topography, as we are not working with a groundwater model.
The information available for the area suggests that precipitation equals ET minus runoff and recharge. Based on literature, if the annual precipitation is approximately 665 mm and 65% is lost to ET, the remaining 21% is runoff and 14% is recharge. Additionally, ET varies significantly, ranging from 140 mm to 300 mm monthly. How does this balance work? If we consider the rainfall-runoff method, where runoff is calculated as precipitation minus ET, could we potentially break this down further into recharge and runoff separately?
Your guidance on how to approach this issue, considering these constraints, would be greatly appreciated.
Thank you once again for your help!
-Ioulia
Mr. Doug Chalmers
Subject: Re: climate data + some help Posted: 1/14/2025 Viewed: 39 times
Ioulia,
To calibrate your catchment to the information in the literature, you will have to adjust the Kc and ETRef so that the proportion of water lost to ET and water entering the model as runoff/infiltration matches the proportions in the literature. To set the proportion of runoff/infiltration going to groundwater versus surface inflow, you can make runoff/infiltration links that go from your catchment to the river and from your catchment to a groundwater object. In the Simplified Coefficient Method, you could then set the Runoff fraction for each of these links so that the proportions match the literature. In the Soil Moisture Method, the proportion of water going to surface versus groundwater is determined in a large part by the preferred flow direction.
I acknowledge your challenge in calibrating a basin with no observed streamflow data and with a complicated hydrology. This is likely outside of your scope, but it sounds important to push to start collecting more flow data in this basin. Flow data is the fundamental piece of information needed to verify the accuracy of a water model. Recently, there have also been efforts to predict flows in ungauged basins using artificial intelligence. Here is a recent article discussing this: https://edit.doi.gov/ai/data-driven-streamflow-forecasting. Given the importance of groundwater in this area with the karst topography, a groundwater model may help simulate the hydrology in the basin- though building a groundwater model is likely outside of your scope as well. See the WEAP tutorial for the WEAP-MODFLOW linkage for more information on connecting a groundwater model with WEAP. As modelers, we can only build the best model that the data will allow for- the quality of data will often limit what we can model.
