Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
Do you really want to delete this prezi?
Neither you, nor the coeditors you shared it with will be able to recover it again.
Make your likes visible on Facebook?
You can change this under Settings & Account at any time.
Copy of Cubing Template
Transcript of Copy of Cubing Template
Assessment of the Predictive Reliability of a SWAT Flow Model and the Evaluation of Runoff Generation and BMP Effectiveness in a Shale-Gas Impacted Watershed Using a Modeling Approach
THE BIG PICTURE
DISCUSSIONS AND SIGNIFICANCE
UPSURGE IN SHALE-GAS EXPLORATION AND DEVELOPMENT ACTIVITIES
It is also known that the clearing of vegetation and the use of heavy exploratory equipment contribute to changes in runoff generation (Seguis et al., 2004; Entrekin et al., 2011).
(Seguis et al., 2004)
The overarching concern in this study is Runoff Generation
Predictive reliability depends on land-use land-cover data.
1. To evaluate the predictive reliability of a calibrated SWAT stream-flow model set-up with high-resolution (1 m) NAIP LULC data classified with object-oriented image analysis technique and low-resolution (28.5 m) LULC data classified with pixel-based maximum likelihood method.
2. To quantify the overall LULC change relative to shale-gas related infrastructure from 2006 and 2010 using NAIP aerial imagery classified with Object-oriented image analysis and assess their contribution to the generation of the storm-water runoff and stream-flow in the most active (in terms of shale gas activities) 10-digit HUC sub-watershed of the Little Red River watershed.
3. Employ a modeling approach to evaluate the effectiveness of the implementation of storm-water BMPs in mitigating runoff generation identified in a high shale-gas activity watershed.
The watershed is completely located within the Fayetteville Shale play and is approximately 4668 sqkm
Roughly 70% of this area being classified as mixed forest land
Average annual precipitation range:1270 – 1320 mm; winter and summer average temperatures of 2 deg C and 30 deg C respectively
A 2006 LULC
70% forest land, 16% pasture, 2% cropland, 3.69% Urbanized, 3.33% water and 5% herbaceous
Perform Object-oriented Image analysis on 2006 high-resolution (1 m) NAIP imagery
Perform comparative accuracy assessment with 2006 CAST LULC data
Calibrated SWAT stream-flow model set-up with high-resolution (1 m) NAIP LULC data classified with object-oriented image analysis technique
Calibrate SWAT stream-flow model set-up with low-resolution (28.5 m) LULC data classified with pixel-based maximum likelihood method.
Evaluate the predictive reliability of the 2 SWAT Models
Predictive reliability in terms of P and R-factors (talk more on them later)
KEY POINTS IN OBJECT 1
To evaluate the predictive reliability of a calibrated SWAT stream-flow model set-up with high-resolution (1 m) NAIP LULC data classified with object-oriented image analysis technique and low-resolution (28.5 m) LULC data classified with pixel-based maximum likelihood method.
OBJECTIVE 1 AND HYPOTHESIS
High-resolution LULC data obtained through OOIA yields a SWAT runoff model of higher predictive ability than the same model created with a low-resolution LULC data derived through pixel-based maximum-likelihood classification.
Pixel-based Maximum Likelihood
Object-oriented Image Analysis
For each class the function calculates the probability of a specific pixel being a member of the class
Also takes into account the mean and covariance of training sets
Classification is achieved by assigning each pixel to the class with highest probability
Assumes a multivariate normal distribution of the pixel within class
A discriminant function is built for each class
Object-oriented Image Analysis
Allows for the decomposition of the image into many homogeneous objects known as segments
SWAT is a:
physically based and
semi-distributed parameter model
(Arnold et al., 1998).
KIA for Low-Res = 0.85
KIA for Hi-Res = 0.82
To compare the two method by their error matrices (pairwise Z-value)
Z-value = 0.0432 which is < 1.96 (taking a 95% CL)
This shows that there is no difference in accuracy for the two methods (at least statistically)
Model Calibration Evaluation Criteria
Pre-Calibration Results for Models with different LULC data
DISCUSSION AND CONCLUSION
In pixel-based spatial continuity of the classified classes are not as strong as that of the object-oriented
Similar overall accuracy values obtained from both methods of classification represent a marked difference from literature
“Salt and pepper effect” is eliminated by the introduction of segmentation in object-oriented method.
Low and high-res models confirm literature results (Cheng et al., 2005; Bosch et al.,2004)
In spite of higher HRU discretization and slightly higher NSE values, the slightly lower classification accuracy for the high-resolution LULC data could be a factor in the model’s inability to account for a greater percentage of the observed discharge data over the low-resolution data model
Differences in model behaviour vis a vis classification methods and LULC data resolution are easier to identify when periods of high and low stream-flow regimes are considered
It is conclued from results that a high-resolution imagery classified with the object-oriented method does not enhance the predictive reliability of a SWAT flow model.
The uncertainty band around the best estimation in each case was higher for the high-resolution model as shown by r-factors of 0.17 and 0.14 respectively.
Both models braketed less than 40% of the observation data; 32% and 37% for high-resolution models respectively.
To quantify the overall LULC change relative to shale-gas related infrastructure from 2006 and 2010 using NAIP aerial imagery classified with Object-oriented image analysis and assess their contribution to the generation of the storm-water runoff and stream-flow in the most active (in terms of shale gas activities) 10-digit HUC sub-watershed of the Little Red River watershed.
Shale-gas related activities have a significant effect on land-cover change and stream-flow from 2006 to 2010 in the South Fork of the Little Red River watershed.
BACKGROUND AND SIGNIFICANCE
There are various environmental implications of shale-gas activities
However, this study investigates the specific problems of LULC change and runoff generation as identified in the SFLRR
Currently, no study can be located done to quantify the differential impact of shale-gas on the overall LULC change and subsequent runoff generation in a shale-gas watershed.
Contribute to literature
Selected for stream-flow impact analysis due to extensive shale-gas activities compared to other subbasins
90% of Subbasin area is forested
Use OOIA to classify 2006 and 2010 NAIP (1m) imagery of study area
Analyze and evaluate differential impact of shale-gas change on stream-flow
Set-up and calibrate baseline and impact scenarios for SFLRR models with LULC 2006 and 2010
Perform land-cover change analysis for LRRW and SFLRR
Two main scenarios for each LULC models: SFLR10W, SFLR10, SFLR06W AND SFLR06
Baseline scenarios: those without well pads (replaced with FRS cover type)
Perform analysis for a projected 10-year period
All models calibrated and validate for 2000-2006 and 2007-2009 respectively
Simulated average annual flow depth for validated models for both scenarios increased from 226.33 mm to 249.61 mm (change of ~10.3%)
Runoff marginally changed from 226.51mm to 249.81 mm, runoff is predicted to change 23.3 mm (~10.3%) from 2010 to 2020
To account for well-pad impact, (2020) forecast scenarios were analyze assuming cover stays same
However, all cover classes had corresponding changes from 2006 to 2010, therefore 10% change is ...??????
Since the two scenarios represent 2006 and 2010 cover, it implies a change of 78% well-pad cover corresponds to a 10 increase in runoff
LIMITATIONS AND CONCLUSION
The distributions of 2006 and 2010 user's accuracies for all classes within their respective matrices have standard deviation of 0.13 and 0.34 with means 0.88 and 0.61 respectively. Indicative of inconsistency in classifications among individual classes the two datasets
Lack of uncertainty analysis brings ambiguity in model interpretation owing to equifinality
Well-pads significantly increased in land-cover from 2006 to 2010 by 630.55 acres (0.65% of the total SFLRR area).
Urban land also increased in coverage; by about 5% of the sub-basin area. Agriculture and forest also decreased by 2.39%
For the 10-year forecast, runoff is projected to increase by 10% assuming the current situation stays the same (78% well-pad land-cover)
Employ a modeling approach to evaluate the effectiveness of the implementation of storm-water BMPs in mitigating runoff generation identified in a high shale-gas activity watershed.
The SWAT model can be used to guide the choice and evaluate the effectiveness of BMPs meant to control storm-water runoff in the South Fork of the Little Red River watershed
BMPs are used as part of the regulatory framework for storm-water impact mitigation through NPDES
Also BMPs have been suggested by USFWS, AOGC, ADEQ, etc.
However, oil and gas activities are not regulated under the NPDES
So suggested BMP guidelines are meant for voluntary implementation by producer
No study or framework is established to evaluate the effectiveness of these BMP
EMPLOY HYDROLOGIC MODELING TO AID CHOICE AND EVALUATION OF BMP EFFECTIVENESS
3 BMPs were selected based on recommendations by USFWS and USEPA
Set Calibrated and validated models as baseline scenarios
Introduce the BMPs in the models to form three other scenarios
Effectiveness of BMPs related to well-pads alone were evaluated by taking the differences in simulated flow rates from model runs with and without well-pad cover types
BMP effectiveness = ((Simulated baseline flow - Simulated BMP flow)/ Simulated Baseline flow) * 100
Check Dam (NC Forest Service)
Grassed waterway (USDA NRCS)
Well-pad flow was determined to be 0.23 cms
Well-pad flow rate appears insignificant
However this is was found to represent a 10% increase in flow depth in objective 2
BMP effectiveness change when looking at overall storm-water mitigation then when only wel-pad flow is considered
Modeling approach holds promise for BMP choice and effectiveness evaluation
However, check dams are most effective in well-pad flow runoff mitigation
In this study, overall runoff is most effectively mitigated with wetlands while check dams is less effective
Effectiveness of chosen BMP depends on cover type (causative factor)
BMPs do have value in shale-gas stormwater runoff mitigation
Is a SWAT flow model reliability improved with an OOIA classified high resolution LULC data?
Yes; also it is imperative to evaluate runoff cause of interest
Insignificant impact but when projected into the future... in study, a 78% increase in shale-gas cover infrastructure is predicted to cause a 10% increase in runoff depth
Can a modeling approach be adopted to manage the choice and evaluate the effectiveness of BMPs designed to mitigate the noted changes in flow?
Does a noted change in shale-gas land-cover impact runoff generation and by how much??
A lot of work still need to be done;
Repeat in other watersheds
Expand to more BMPs
By: Kwasi Asante (ENDY)
Dissertation Adviser/Committee Chairman: Dr. Jackson D. Cothren
Dissertation Committee Members: Dr. Ralph K. Davis, Dr. Greg Thoma, Dr. John Van Brahana
To tackle this with a hydrologic model, it imperative to determine the predictive reliability
Whose accuracy depends on classification methods and data resolution
My Advisor, Dr. Cothren and Committee members, Dr. Davis, Dr. Thoma and Dr. Brahana
Dr. Mauro DiLuzio (Blackland Research and Extension Center) Dr. Mansoor Leh, Bruce Gorham, Dr. Steve Boss
My Family; Mom and Dad, Dr. and Mrs Albright, most importantly my wonderful wife Bethney
JoAnn, Mary-Gail and Lisa
"Since all models are wrong the scientist cannot obtain a "correct" one by excessive elaboration. On the contrary following William of Occam he should seek an economical description of natural phenomena. Just as the ability to devise simple but evocative models is the signature of the great scientist so overelaboration and overparameterization is often the mark of mediocrity". - George E.P. Box (1976)
"All models are wrong, but some models are useful. The practical question is, how wrong they have to be to NOT be useful" - George E.P. Box