Then, just like Spot.us raises funds to support the reporting on a particular topic, we’ll be looking for donations to support our research team. We’ll be aiming to raise micro-donations in 99 cent increments.

Head over to the Questions page to share (rank) questions and feel free to click the Amazon donate button to throw your 99 cents behind a question.

Over the past few days, these questions have been submitted via the website:

• How Can Cars Run on Hydrogen?
• How Are Fruit Trees Pollinated?

We welcome your feedback and ideas on this new project.

Can a machine be taught to determine the sentiment of a Twitter message about weather?  With the data from over 1 million crowd sourced human judgements the goal was to use this data to train a predictive model and use this machine learning system to make judgements.  Below are the highlights from the research and development of a machine learning model in the cloud that predicts the sentiment of text regarding the weather.  The following are the major technologies used in this research:  Google Prediction API, CrowdFlower, Twitter,  Google Maps.

The only person that can really determine the true sentiment of a tweet is the person who wrote it.  When the human crowd worker makes tweet sentiment judgements only 44% of the time do all 5 humans make the same judgement.  CrowdFlower’s crowd sourcing processes are great for managing the art and science of sentiment analysis.  You can scale up CrowdFlower’s number of crowd workers per record to increase accuracy, of course at a scaled up cost.

The results of this study show that when all 5 crowd workers agree on the sentiment of tweet the predictive model makes the same judgement 90% of the time.  When you take all tweets the CrowdFlower and Predictive model return the same judgement 71% of the time.  Both CrowdFlower and Google Predictions supplement rather than substitute each other.  As shown in this study, CrowdFlower can successfully be used to build a domain/niche specific data set to train a Google Prediciton model.  I see the power of integrating machine learning into  crowd sourcing systems like CrowdFlower.  CrowdFlower users could have the option of automatically training a predictive model as the crowd workers make their judgements.  CrowdFlower could continually monitor the models trending accuracy and then progressively include machine workers into the worker pool.  Once the model hit X accuracy you could have a majority of data stream routed to predictive judgments while continuing to feed a small percentage of data the crowd to refresh current topics and continually validate accuracy.  MTurk hits may only be pennies but Google Prediction ‘hits’ cost even less.

Weather Sentiment Prediction Demo Application:

LIVE Demo Link: www.SproutLoop.com/prediction_demo 

Note, this demo uses server side Twitter feed that is throttled, retry later if you get no results. Contact me regarding high volume applications and integrations with full Twitter firehouse.

Match Rate/ Accuracy Findings:
Below are the highlighted match rates of CrowdFlower human judgements to Google Prediction machine judgements.  A match rate compares the resulting predicted sentiment labels from one method up to those from another:

• Google Prediction API matching CrowdFlower Sentiment Analysis  = 71% match rate
• Mirroring DialogueEarth Plus filtering of lowest 22% confidence scores  = 79% match rate
• Tweets sentiment can be confusing for humans and machines.  Google predictions of only the tweets in which all the crowd workers agreed (CrowdFlower confidence score = 1)  =  90% match rate

About Google Predication API:
During the May, 2011 Google IO conference Google released a new version of their Google Prediciton API with open access to Google’s machine learning systems in the cloud.  The basic process to creating predictive models is to upload training data to Google Cloud Storage and then use Google Prediction API to train a machine learning model from the training data set.  Once you have a trained model in the cloud you can write code with their API to submit data for sub second (avg 0.62 sec per) predictions.

About The Data:
Much has been written about DialogueEarth.org’s Weather Mood Pulse system.  Pulse has collected 200k+ dataset of tweets that have been assigned one of five labels regarding the tweets sentiment related to the weather.  This labeling of tweets is crowd sourced with the CrowdFlower system that presents each tweet with a survey for the workers in the crowd to decision.  CrowdFlower has quality control processes in place to present the same tweet to several people in the crowd.  DialogueEarth’s crowd jobs were configured so that each tweet was graded by 5 different people.  CrowdFlower uses this 5 person matching and each person’s CrowdFlower “Gold” score to calculate a confidence score between 0 and 1 for each tweet.  About 44% of the tweets have a confidence score of 1 or 100% of the graders agreed on the sentiment label for the tweet, while some of the other tweets have low scores like 0.321 meaning very little agreement in tweet sentiment plus some influence from each of the graders Gold scores.  The Pulse system has chosen to use only the tweets that have a CrowdFlower confidence score equal or greater than 0.6.  Dozens of models where build using various segments of CrowdFlower confidence score ranges.  Testing showed that the best model used the full confidence range of CrowdFlower records.    

Weather Sentiment Tweet Labels From CrowdFlower:
The CrowdFlower scored tweet data contains the tweet text, weather sentiment label and the CrowdFlower confidence score.  The tweet data set was randomized into two segments: 111k(~90%) rows used to train model and 12k rows held out for testing the model.

Sentiment	Modeled Tweets 90%	Test Tweets 10%	% of Total
TOTAL	111,651	12,389	124,040
negative	23,976	2,578	21.4%
positive	21,688	2,384	19.4%
not weather related	34,232	3,780	30.6%
neutral	29,333	3,340	26.3%
cannot tell	2,422	307	2.2%

CrowdFlower Confidence Scores Correlation to Google Prediction Match Rate / Accuracy:

Running Match Rate – Ordered by CrowdFlower Confidence, Google Prediction Confidence

• X- axis shows the distribution of CF confidence scores in 10% random test data set 12k rows
• Google is better at predicting Tweets that have a higher CrowdFlower confidence score
• The Google confidence score correlates with accuracy/match rate, on average higher Google confidence = higher accuracy of matching

Google’s Prediction Confidence Score and Correlation to Match Rate Accuracy
Running Match Rate and Google Prediction Score – Ordered by Prediction Score, Random

• Higher Google confidence scores correlate with higher matching/accuracy rate.
• Filtering results at Google conf score > 0.8290 will result in 80% accuracy and filtering/loss of 24.41% of data.

Accuracy/Match	Google Confidence	% Data Filtered	Rows out of 12390
98.47%	score = 1	86.88%	1627
90%	score > 0.99537	55.16%	5543
85%	score > 0.95688	38.78%	7586
80%	score > 0.82900	24.41%	9367
78.93%	score > 0.79122	21.60% **	9715
75.00%	score > 0.61647	11.06%	11021
70.91%	score > 0.25495	0.0%	12390

** Note, 21.6% is the current % of data that Pulse filters by excluding CF conf. scores <= 0.6

Effect of Model Size on Match Rate / Accuracy

• The larger the data model training data set the higher match accuracy
• 95% (5k of 111k) decrease in dataset set size decreases match rate by 7.1% (64% – 71%)
• The classificationAccuracy returned from Google’s model build was between 5%(model 5k) to 0.06%(model 111k) different than tested accuracy rates.

Formatting Tweet Text for modeling:
When preparing text for modeling Google recommends removing all punctuation because “Punctuation rarely add meaning to the training data but are treated as meaningful elements by the learning engine”.  The tweet text was lowercased, stripped of all punctuation, special characters, returns, tabs,.. These were replaced by a space to prevent two words from joining.  With the unique 140 character limit and the use of emoticons it might be interesting to replace emoticons with words like replacing  :’-(  with something like a specific replacement  ‘ emoticon_i_am_crying ‘ or general  ‘emoticon_negative’ before building the model.  Here is an example of tweet hygiene below:
BEFORE:  83 degrees n Atl @mention:59pm :> I LOOOOVE this …feels like flawda
AFTER:  83 degrees n atl mention 59pm i loooove this feels like flawda

Training The Google Prediction Model
Here are the basic steps for training a model.  The training time can take a few hours for a 100k/10MB training data set, this seems to depend on the sever load.  When Google is finishes building the model the trainedmodels.get method will return a confusion matrix and also the model’s “classificationAccuracy” score.  Note, Google’s classificationAccuracy score and the testing match rate  accuracy scores below are statistically the same (0.71 vs 0.709).

Google Model building confusion matrix with a classificationAccuracy of 0.71

SENTIMENT	negative	positive	not weather related	neutral	cannot tell
negative	1627	191	261.5	351	88
positive	200.5	1631.5	220	236	40.5
not weather related	239.5	163.5	268	1965.5	66
neutral	262.5	163.5	268	1965.5	66
cannot tell	5	3	1	8.5	3.5

Testing 12k hold out tweets against the model above with a match rate / accuracy of 0.709
Crowd Flower Actual   

SENTIMENT	negative	positive	not weather related	neutral	cannot tell
negative	1824	232	294	413	94
positive	196	1732	216	200	61
not weather related	251	200	2967	460	64
neutral	300	215	303	2263	86
cannot tell	7	5	0	4	2

Author : Kevin Cocco @kcocco – SproutLoop.com

Drinking water that recently was flushed down a neighbor’s drain is a tough concept, pardon the pun, to swallow. However, as the piece in the Times correctly points out, we are headed into times in which resources like water are likely to be scarcer. To my mind, this community is a shining example of people—who rightly had very strong feelings on an issue—being willing to accept what the science community had to offer. Jerry Sanders, the mayor of this San Diego community, put it this way: “If science is behind you and you can prove that, I think people are willing to listen.” Here, here!

Image in post from Climate Watch. Thumbnail image on home page from the NY Times article.

In the meantime, I have been reflecting more on whether Quora really is the place to build up a library of great Q&A about the environment. Specifically, I am most interested in questions that are answerable, yet good, clear answers are not available on the Web.

Here’s a flowchart that describes the “sweet spot” that I believe Dialogue Earth could target with the EarthQ project. Some explanations follow the chart below.

Starting off, we are focused on issues related to the environment. I take a broad view of what is connected, but at some point we need to set limits.

It should take more than a couple of minutes and a good search engine to answer the question. The folks on Quora, for example, are okay with questions that a quick search would solve (see this Q for example). That’s fine, but I propose that is outside of the EarthQ niche. I see EarthQ as a place for asking questions that are tough to answer well.

The questions should not be impossible to answer (e.g., are there other life forms in space?) or the type that require an opinion, even if from an expert (e.g., should hydraulic fracturing be permitted in location xyz?). Squarely in the EarthQ sweet spot would be questions that require expert knowledge to answer. Finally, a key will be that “someone” is willing to pay for a great answer. I put someone in quotes, because this could be the person who poses the question, it could be Dialogue Earth in cases where there is a ton of interest in the question, it could be a philanthropic interest who wants to support these types of Q&A, and it could be an advertiser or underwriter.

A final note about the EarthQ sweet spot is that the typical answer will typically require some multimedia to help explain it. Maybe just a nice, clear graphic and a paragraph or two of text. But maybe the case requires more, such as a short video, like some of those we have prototyped here at Dialogue Earth (visit our YouTube channel).

Thoughts?

The sentiment we mapped was culled from tweets gathered from four weeks’ worth of data starting on May 22nd, 2011. This time period was chosen to coincide with Memorial Day, a holiday during which many Americans travel by car. Our team was curious to see whether there would be an uptick in either the volume of tweets about gas prices during this period or a noticeable change in sentiment about these prices.

We found that during the month-long period we monitored Twitter users’ sentiments, feelings about gas prices were overwhelmingly negative and changed little on a week-by-week basis (see the maps below). The maps are from our interactive Pulse tool, which displays the Twitterverse’s sentiments about gas prices on the day before Memorial Day (deeper orange colors indicate negative sentiment; greens indicate positive sentiment). Additionally, we did not see clear distinctions in sentiment between different regions in the United States. We did see an apparent decline in tweets related to gas prices over the period, although the data set is not sufficiently long to report a clear spike in volume associated with the Memorial Day weekend travel period. Please check out our larger, interactive data set having to do mood about daily weather on the Pulse platform.

In the future, we hope to track mood about gas prices and related topics over longer time periods, which we expect would allow us to elaborate further on how sentiment about gas prices changes across time, as well as across geographic regions.

Mood About Gas Prices: Week of May 22, 2011

Mood About Gas Prices: Week of May 29, 2011

Mood About Gas Prices: Week of June 6, 2011

Mood About Gas Prices: Week of June 13, 2011

Mood About Gas Prices: Week of June 13, Northeast & Midwest

Note that pie charts report mix of positive and negative sentiment for metro areas, whereas the coloring of states report the sentiment averaged over the entire state.

Mood About Gas Prices: Week of June 13, Southern U.S.

Mood About Gas Prices: Week of June 13, Western U.S.

Mood About Gas Prices: How do we do it?

At the heart of our process are judgments made by workers from the “crowd.”

We provide our workers with a survey, and a set of instructions on how to judge a tweeter’s emotion about gas prices. Dialogue Earth’s research team has previously done extensive testing of this approach using surveys on topics such as weather (read a more detailed analysis about this process here).

We also work on the tweets before and after they are sent to Crowdflower, our partner for crowdsourced workers.

Upstream of Crowdflower, we collect tweets on a continual basis using Twitter’s streaming API. Our collection is based on several different keywords (for example, “unleaded” + “price,” “pump” + “price,” etc.). Then, for a given week, we randomly pull 200 tweets for each state and metro area (read more about our rationale behind the geographic regions chosen here). The actual number sent to the crowd is often a bit less, because we eliminate exact duplicates so that we are not coding them multiple times.

Working with Crowdflower’s API, our team sends tweets from our database to a Crowdflower job, then back to our database once the requested number of judgments have been made. We then apply the sentiment found for each state and metro area to the full dataset for those geographic units.

I was drawn to the talk specifically to hear what Administrator Jackson would say about the new program, which I had read about in the morning’s Minneapolis StarTribune. She did not disappoint, although the bit about this new program occurred just as the questioning period came to a close. It was clear that she has a good deal of enthusiasm about this new program and its prospects for helping to reduce pollution, such as the introduction of nitrogen into the Mississippi that can lead to low-oxygen conditions in the Gulf of Mexico.

The write-up on the Minnesota Department of Agriculture site states that the program “is designed to accelerate voluntary adoption of agricultural practices that enhance water quality.” Specifically:

Program details will be developed cooperatively in the months ahead, but the general concept is that farmers who implement and maintain approved conservation plans will be assured that their operations meet water quality goals and standards. So long as the participating farmers meet program obligations, they will not be required to implement additional water-quality practices for the duration of their certification. Conservation plans for participating farms would be developed by a qualified team of experts with consideration given to the unique characteristics of the farm, the farming practices, the surrounding watershed, and the region in which the farm is located.

A sampling of news articles and conversations with others at the talk suggest the following reactions. Farming interests will, as Administrator Jackson indicated, be thankful to have a sense of certainty that should result from the program, enabling better planning of investments and other business decisions. On the other side, those advocating for clean water will wish that the program would be mandatory and, importantly, that farmers would be held to limit runoff, rather than only being held to implement best management practices that should limit runoff.

If the management plans are tailored to each farm as indicated in the quote above, then this may well be a decent compromise. Time will tell on that. Having a brother who is a dairy farmer, I am all for solutions that are a win-win for farmers and the environment. Looking at it from another perspective, nobody should be happy that the Mississippi transports something like a billion dollars of nitrogen, a key plant fertilizer, into the Gulf of Mexico in recent years.

My feeling is that a plan like this new state-federal partnership will require great monitoring data to confirm that the the new management plans are having the desired impact on water quality. In the short-term, it may be sufficient to require the adherence to good management plans, but ultimately, we as a nation will need to know if the implementation of these plans is having an impact. Detailed monitoring of water quality will be an essential piece of the puzzle.

Photo credit: University of Maryland on Flickr.

For a taste of the debate, check out the controversy a year ago surrounding the premier of Troubled Waters, A Mississippi River Story (link to view on Twin Cities Public TV). Among other activity, there was a heated back-and-forth on the opinion pages of the Minneapolis Star Tribune (first piece arguing that research for film was dubious, counter piece by producers about the research behind the film). The flare up nearly squelched the premier, although it ultimately took place.

Personally, I experienced how difficult it can be to establish a “consensus view” of how to describe nitrogen flows for the report on ecosystem indicators described in my previous post. There was a constant tension throughout the report process to “tell readers what to do with the information” rather than just “giving them the facts.” I believe we ultimately struck a good balance, however, we nearly had resignations of participants when we unintentionally pushed the line on the indicator dealing with the movement of nitrogen and got too close to pointing fingers.

For this post, I’m coming at the debate from a very different angle. Yes, there are multiple sources of the nitrogen in the Mississippi as it flows into the Gulf of Mexico. Sure, municipal sewage discharges should not be ignored, and I would argue they are not. Here’s a different question: if we had a way to dip a futuristic “nitrogen net” into the Mississippi just before it enters the Gulf, what would be the value of the recovered nitrogen (N)? A quick estimates suggests a lot: about one billion dollars worth of nitrogen!

Here’s the cost of anhydrous ammonia, the cheapest form of N-fertilizer (source: USDA’s Economic Research Service). These data are presented based on the cost of the N in the anhydrous (NH₃). As an example, the 2011 cost of anhydrous is reported as $749 per ton; because each pound of NH₃ has in it 82.4% N, then the cost of N per ton ends up being $910.

So, let’s use a value of about $1000 per ton of N…the trend in cost suggests we might be there before too long. Looking back at the previous post, we saw that there were about one million metric tonnes of N flowing into the Gulf each year—here’s the graph again.

For this rough analysis, we can treat a metric tonne as equivalent to a regular U.S. ton (the conversion is 1.1). So, that means there’s about a billion dollars worth of N flowing into the Gulf of Mexico each year.

In a forthcoming post, I’ll dig into this further. Regardless of the source of this nitrogen, it seems like a huge opportunity. Either we need to figure out how to tighten things up so that the costly nutrients never make into our rivers, or some industrious soul should figure out that futuristic “nitrogen net” and reap dollars by the bushel.

Thumbnail image credit: U.S. Army Corps of Engineers on Flickr.

The first indicator in the spotlight is one that describes the movement of nitrogen (N) into major waterways, through run-off as well as point discharges such as sewage treatment facilities (here’s a link to download a pdf of the indicator). Why is this an important indicator of the state of U.S. ecosystems? Our explanation in the report does a good job of answering this question:

Nitrogen is an important plant nutrient and is essential to all life. Nitrogen is an abundant component of the earth’s atmosphere, but it is unavailable to most life in gaseous form. In order to be used by plants and other organisms, nitrogen gas must be “fixed,” or converted to a “reactive” form, that plants can use, such as nitrate. Nitrogen is fixed and accumulates in ecosystems through natural processes, such as the growth of nitrogen-fixing plants like clover and soybeans. However, human activity has greatly increased the amount of reactive nitrogen added to ecosystems. The largest human-caused input of nitrogen to ecosystems comes from the conversion of atmospheric nitrogen gas into fertilizers. Additional reactive nitrogen gas is produced by the combustion of fossil fuels. Reactive nitrogen from all these sources can ultimately enter streams and rivers. Excess nitrogen transported to coastal waters by rivers can lead to low oxygen conditions, threaten fish and animal life, and degrade coastal water quality.

Below is the main graphic from the indicator. The lower panel is a map of the lower-48 states, with major watersheds colored based on the amount of nitrogen that enters waterways from runoff, discharges or other sources. The watersheds in black had the highest amount of nitrogen per area entering waterways; those in light blue had the least (there were no data available for those watersheds shown in gray).

The graph in the upper panel rolls up the mapped data, showing the percent of land area in the different N delivery categories. The main take-home message from the graph is that the 600-1500 pounds per square mile per year category shrunk, with those acres moving into the lower category (10-600 pounds per square mile per year). While it is easy to get tongue tied explaining that an increase of land in a lower N delivery category is good news, that’s a reasonable conclusion from the data.

So, can we tell anything about how things have changed since the 2001 — 2005 period? It doesn’t look like the data shown in the graphic above have been updated, but we can tell a good bit from the third graph in the indicator: how much nitrogen is flowing from the Mississippi to the Gulf of Mexico.

In terms of flow of nitrogen to the Gulf of Mexico via the Mississippi-Atchafalaya river basin, there does not seem to be any improvement, and possibly an uptick in discharges. The graph below compares three U.S. Geological Survey (USGS) data sets: (1) the blue line is from the 2008 report from the Heinz Center and is only for nitrate discharges and is through 2004 ; (2) the orange line is for both nitrate + nitrate, which is typically a minor component, and is from a predictive model called LOADEST; (3) the yellow line, which tracks the orange line closely, is a composite of samples, and is not available from every year (note there is no data point during the spike witnessed in 1993. Note that data from the Heinz Center report can be accessed here, and the USGS data for the orange and yellow lines are available here.

So, not great news on the flow of nitrogen into the Gulf. What about the low-oxygen zone in the Gulf of Mexico? This is, after all, the main motivator for studying the movement of nutrients downstream. Here’s a map from the Hypoxia Research Team at the Louisiana Universities Marine Consortium showing the extent of low-oxygen bottom waters during their 2011 summer cruise (see their press release here):

The graph below, also from the Consortium, shows that the size of the zone with low oxygen levels (bottom waters below 2 mg/L dissolved oxygen) was somewhat smaller than in 2010, yet well above the goal set by the Hypoxia Action Plan—see the second graph below that reports data through 2010. Note that Consortium researchers describe how a recent tropical storm likely mixed waters before their survey during the summer of 2011, most likely causing their area estimate to be lower than it might have been days earlier.

Stepping back, it looks like discharge of nitrate into the Gulf of Mexico from the Mississippi-Atchafalaya basin has been more-or-less unchanged since the data reported in the 2008 Heinz Center ecosystem report. Plus, while there’s been fluctuation, the driving reason for concern—they zone of hypoxia in the Gulf of Mexico—remains well above the management target.

And, the first question is also up on Quora: What are the compartments in which carbon is stored on Earth? Do you have the credentials to answer this, or do you know someone who does? Please help out—I believe it will be a rewarding way to bridge the gap between the science community and the rest of society. Also, keep in mind that “upvoting,” adding comments for the author, or suggesting edits,  as soon as there is one or more answers will be a really great way to participate.

For anyone new to Quora, I’d encourage you to take a look at a new infographic and video that helps explain how to have a winning strategy within the Quora community.

Any and all feedback is welcome!

As an experiment, I created a short (3 min) video that walks you through the infographic. Note that there are a bunch of links to Quora questions throughout the video—ensure annotations are turned on to see them. Feedback welcome!

http://www.youtube.com/watch?v=9sF-_0bTdH4

And, the infographic…

A few weeks back when I dipped my cup into the Twitter river, which I need to do more often, I happened upon a press release from the U of Minnesota Extension having to do with a new spreadsheet tool designed by energy economist Doug Tiffany to help those considering an alternative vehicle purchase.

Tiffany’s tool allows the user to customize the data input, adjusting things like purchase price, down payments, expected cost of fuel, one’s personal opportunity cost (the cost to me for tying up cash for down payment and monthly car payments). The tool then produces several graphs, the first of which is a 15-year cumulative cost projection comparing several vehicle types. For the graph below, I used the default values in the tool, which can be downloaded here.

So, what this means, is that for the input parameters, the 15-year cumulative cost of ownership of the extended range vehicle, such as a Chevy Volt, comes out the highest, whereas the electric vehicle, like the Nissan Leaf, comes in with the lowest cost. The following graph shows the breakdown between car types at year 15.

Cost may not be the only motivator, however, at a time when many are concerned about greenhouse gas (GHG) emissions. Thus, it may be important for many to compare the estimated emissions of greenhouse gases from the different vehicle types. Tiffany’s tool produces this graph based on various inputs.

As would be expected, the conventional gasoline-powered vehicle has the highest emissions. There’s not as much difference between the other types, with the hybrid having somewhat higher GHG emissions than the all-electric, both of which do somewhat better than the extended range electric.

Tiffany’s tool reminded me of a widget available from my colleagues over at Climate Central, which is embedded below. It allows you to get a good sense of potential fuel and GHG emission savings between two different vehicles. It dynamically updates fuel costs for your state, which is a nice feature.

Together, these tools provide a great resource to anyone considering an alternative car purchase. Perhaps these groups could team up to create a tool that blends together both annual costs / savings and economics over the vehicle’s lifetime.

Why might storage play a key role in a grid that has a lot of wind turbines?

In early October, Mark Ahlstrom, the CEO of WindLogics, gave an excellent presentation as part of Frontiers in the Environment series at the U of MN’s Institute on the Environment (view the video here). He explained how our variable demand for electricity over the course of a typical day is met by a range of generation sources, from base load to “peakers.” Base load is met by nuclear, coal, and hyrdo—depending on location. Base load plants are meant to run 24/7, and adjustments to their output need to be scheduled well in advance. On the other end of the spectrum are peakers, which are small electrical generators that can be turned on with little notice and do not need to run for a minimum period of, say, a day. The figure below is based on a slide from Ahlstrom’s slides.

Ahlstrom went on to explain how things get interesting when a substantial amount of wind generators are added to the mix. In the figure below, the bright green line represents the reduced daily load that results by adding a number of wind turbines to this hypothetical typical energy demand. He goes on to explain how the rest of the electricity generators need to adapt to this new load curve, which can have more spikes and other challenges, such as a faster ramp-up in the early morning hours—challenges that the current energy system wasn’t set up to tackle.

Here’s a good piece by the ClimateWire group that landed in the NY Times: Fickle Winds, Intermittent Sunshine Start to Stress U.S. Power System. It dives into detail about the policy challenges that blending intermittent energy sources with our traditional electrical generation system.

Back to storage. A recent piece in the NY Times described how Batteries at a Wind Farm Help Control Output. In the largest battery installation connected to the grid in the U.S., they’ll use over a million batteries to provide storage for a few minutes of generated electricity from a large wind farm in West Virginia. The idea is that this stored electricity can be fed into the grid to help smooth things out when output from the wind turbines drops off momentarily. It is not designed to even out longer periods when the wind is calm. Significant storage on the time scale of hours and possibly days would most likely require a solution such as pumping water uphill or compressing air when the wind blows, and then using this stored energy to run a generator when the wind is quiet (here’s a project underway to study pumped hydro in conjunction with wind farms in northern Minnesota). There may also be options to store electricity in electric vehicles integrated with a smart grid of the future.

Finally, another NY Times piece that gave me some pause had to do using water heaters and electrical space heaters to store excess electricity brought about in part from excessive winds in the Pacific Northwest. The idea is that automated, shorter-term storage of excess electricity in homes that already have electric hot water heaters and heat with electricity could bleed off excess electricity pouring into the grid during storms when wind generators are running at maximum output (note that the situation was further complicated because hydropower operators were unable to reduce output from their generators for fear of creating conditions that might kill fish).

At a gut level, this strikes me as a good way to avert disaster, but probably not a great strategy from an energy efficiency standpoint. My sense is that using electricity for water heating and space heating is not nearly as efficient as, say, natural gas. The standard reason for arguing this point is that the efficiency of traditional energy plants is much less than 50%, whereas a high-efficiency hot water tank can exceed 90%. However, maybe this standard reasoning needs to be updated in a situation where electricity comes from wind or solar. This is a topic that merits further consideration.