Lake Travis Water Level Forecast (3/1/2022)

Lake Travis Water Level Forecast
(as of 03/01/22)

Author: David Ferreira (Resume)
Data Sources:  LCRA, Texas Water Development Board (WaterDataforTexas.org)

Forecast

Fig. 1: 6-month forecast with 80% and 95% confidence intervals.

Forecast Data

Table 1 shows our mean forecast (“Point Forecast”) and 80% and 95% confidence intervals. (Confidence intervals explained.) We encourage readers to view our forecast as a range, rather than a specific number. However, we are excited about the accuracy of our Point Forecasts and encourage readers to compare our mean forecast to the LCRA’s own median projections*.

Table 1: Mean forecast and confidence intervals by month (6-month forecast)

Period	Point Forecast	Lo 80	Hi 80	Lo 95	Hi 95	Actual	Delta	% Delta
Mar-22	660.9	655.6	666.3	652.8	669.1
Apr-22	660.7	652.9	668.6	648.7	672.7
May-22	661.3	651.5	671.0	646.3	676.2
Jun-22	661.2	649.8	672.6	643.8	678.7
Jul-22	658.8	646.0	671.7	639.2	678.5
Aug-22	656.4	642.2	670.6	634.7	678.1

 

* When comparing the accuracy of our forecasts to that of the LCRA forecasts for Lake Travis, please keep in mind these 2 important details:

1.        The LCRA’s median forecasts are for water levels on a specific date (1^st of each month?) while ours are forecasts of monthly average water levels.

2.        The LCRA holds the keys to the Mansfield Dam and actually controls the water level. We do not.

Notes:

We are not affiliated with the Lower Colorado River Authority (LCRA) in any way whatsoever.

This document is not intended for commercial use or publication. It is produced for demonstration and evaluation purposes only. It serves as a record of our forecast of Lake Travis water levels made on 03/01/22.

This original forecast is provided free of charge to interested members of the general public as a courtesy. No warranty is made as to the accuracy.

This forecast and document were created by an automated script running in R.  Raw data was gathered automatically from sources by RapidMiner and uploaded to our SQL database.  It was resampled in R to produce this (proprietary) composite forecast.

Let us help you automate your Data Science and Machine Learning processes.

Historical

Lake Travis is one of 6 lakes managed by the Lower Colorado River Authority – the “Highland Lakes”. It was formed in 1941 when the river was impounded by the completion of the Mansfield Dam. These lakes serve as storage (inventory) to satisfy electrical demand for consumers across Texas and irrigation water demand from downstream farmers.

While water levels of other LCRA lakes are managed to maintain a narrow range of water levels, the water level of Lake Travis is allowed to fluctuate in a much wider range than other Highland Lakes. This creates challenges for property owners, commercial users, and recreation users of Lake Travis.

In recent years, the lake has been managed in the range of 650’ to 690’ above mean sea level (MSL). However, a significant drought from 2010 to 2015 caused levels to remain below 640’ for several years.

Fig. 2: Lake Travis historical water levels and classical decomposition into component elements: seasonal, trend-cycle, and remainder (noise)

Fig. 3: Mid-range view of prior water levels and 6-month forecast

Fig. 4: Long-range view of historical water levels and 6-month forecast

Frequently Asked Questions

Q1: Why do we do these Lake Travis forecasts?

A1: We generate these forecast on a quarterly or semi-annual basis for the following reasons:

-          We have an ongoing recreational interest in the water level of Lake Travis.

-          Public Service: Private and commercial property owners, recreational users, and authorities all have a keen interest in knowing future water levels with maximum accuracy.

-          We derive entertainment and educational value from doing forecasts like this and seek other opportunities for similar challenges.

-          We seek to re-channel our successful software engineering career toward a hybrid career that will merge our enjoyment of software engineering with our passion for econometric analysis and forecasting.

Q1: Why don’t you forecast the levels of the other Highland Lakes above and below Lake Travis? Is that too difficult for you? Are you scared?

A1: We chose Lake Travis, in part, because of the challenge it presents and because of the utility of accurate forecasts for ourselves and other stakeholders. Predicting the levels of others lakes in this region might not be very useful since their fluctuations are minimal. For what it’s worth, we are considering adding other lakes to our forecast repertoire. But first we’d like to focus on integrating our Neural Network model of Lake Travis with the model used in this forecast.

Q2: Doesn’t LCRA provide forecasts of Lake Travis water levels?

A2: Yes, they do provide 6-month “projections” each month. They also hold the keys to the dam. They have both control and fiduciary interest in the data being forecast. We encourage readers to compare our forecasts to LCRA projections for accuracy. We would, however, caution the reader that LCRA projections are for water levels for specific dates (the 1^st of each month?). In contrast, our forecasts are for a monthly average water level for each month being forecast. Still, we believe the accuracy of our forecasts surpasses that of the LCRA projections.

Q3: Can’t we just use Excel to forecast things like this?

A3: You can. And that may fit your particular need quite well. Another method would be to print out a graph and draw a line through both the same month in the first year of data and the last year of data and carry that line forward. These methods might both suite your needs (no sarcasm intended). We encourage you to compare these forecasts against ours, those of LCRA, and observed results. We think you will find that not all methods are equally accurate over time.

Q4: So … y’all studied Economics / Econometrics. How does that qualify you to forecast a hydrological phenomenon like the water level of a lake?

A4: The water level of Lake Travis is indeed a hydrological phenomenon. But it is also an economic phenomenon – it is impacted by statewide electrical demand, and the irrigation demands of farmers downstream. Forecasting the water level of Lake Travis is very much an econometric problem that involves some hydrology.

 

About the Author

Dave Ferreira has a life-long passion for quantitative analysis, data-driven reasoning, and extracting insight and Truth from data (the bigger, the better).

He is currently enrolled in the MS Finance & Economics program at West Texas Univ. He is an experienced Software Development Eng., formerly of 3M, Dell, and Microsoft.

Dave is also a former Peace Corps Volunteer and in a prior career, was a Staff Economist for a national industrial trade association.