Climate-Aquatics Blog #27: Part 1

Post Reply
NaepcA1
Site Admin
Posts: 1766
Joined: Tue Oct 27, 2009 11:50 am

Climate-Aquatics Blog #27: Part 1

Post by NaepcA1 »

Hi Everyone,
This time out we’re embarking on a bit of a diversion from our regularly scheduled programming as we start a 3 part mini-series on spatial statistical models for stream networks. It’s out of the planned sequence since these blogs will ultimately be filed under the “Climate-Aquatics Cool Stuff Module” but I talked to the blog’s publisher and he agreed that their content was significant enough that it warranted an exception & because recent developments are about to make these models much more accessible for broad application to streams anywhere, so without further adieu…

For anyone that’s ever tried to build something & get by with poor quality tools or tools not specifically designed for the job at hand, you know the frustration. Something can usually be made to work but it’s often a painful process and the thing built could have been much better with the right tools. Well, I’d argue that those of us working with streams and rivers and their critters have long lacked important tools for analyzing and understanding these systems at one of the spatial scales that is fundamentally important for planning and prioritizing conservation efforts. That scale is the network scale—spanning 100’s – 1000’s of stream kilometers and 10’s – 100’s of individual streams. It’s this scale at which multiple stakeholders across a river basin coordinate and make choices about how & where resources are spent to protect and restore habitat or conserve populations. From a biological perspective, this scale is also relevant because it’s one at which metapopulation dynamics and the exchange of individuals between populations occur to lend populations some resilience and facilitate the persistence of species within landscapes. That being the case, the network scale is where “smart maps,” informed by lots of data to show the status and trends of important aquatic resources (blog #26), would be powerful tools for facilitating coordination efforts among resource professionals and different agencies to make sure we do the smartest things in the smartest places.

The good news is that there are literally armies of us running around the woods measuring all manner of stream things at discrete points on those networks and 100’s – 1000’s of such measurements now exist in many river basins. Those data are the fundamental elements that could be stitched together to build the necessary “smart maps” but doing so also presents a few issues that have been difficult to address until recently. Those issues involve such things as accurate geo-referencing of sample locations (now easily overcome with inexpensive GPS units), computational horsepower for dealing with large databases (more & cheaper horses all the time), and having standardized protocols to obtain unbiased measurements of commonly collected stream and biological attributes (this one’s still evolving but rapidly improving with development of inexpensive sensors, genetic assay techniques, and bioassessment protocols. And in practice, having bias in some measurements may be acceptable if it better informs parts of a network-scale “smart map” that would otherwise be devoid of information).

The fourth major issue, and it’s a biggie, is having analytical techniques that can properly deal with those 100’s – 1000’s of nonrandom, clustered measurements that our uncoordinated army is generating (graphic 1). Those nonrandom site locations, especially in areas with high data densities, create a problem called spatial autocorrelation wherein there’s a certain amount of spatial dependence in the samples and redundancy in the information they provide. This redundancy violates one of the basic assumptions inherent to many statistical methods, that samples are independent of one another. Analyses that ignore the way sites are located in space and potential spatial dependence among them run the risk of providing biased parameter estimates (and poor maps if used to predict) when applied to these sorts of databases. To deal with the issue, one could simply discard some of the data until the redundancy was reduced to an acceptable level, but whose data do you throw away, how much, and where?

Much better to use it all, if possible, since someone went to the time and expense of collecting the data. Plus, each sample location contains some valuable information if the data can be properly weighted to account for the degree of spatial autocorrelation with nearby sites. Spatial statistical analyses were developed, in part, to address this issue, and can be used to derive unbiased parameter estimates and valid inferences from clustered databases. These analyses have long been widely available for, and applied to, terrestrial systems but they simply haven’t existed, other than a few basic descriptive measures, for data on stream networks. I know because I searched high & low for them to no avail during my Ph.D. 15 years ago as I struggled to analyze my own autocorrelated datasets. Although many (myself included) have tried work-arounds that apply spatial terrestrial analyses to stream data, we’re back to that wrong tool for the job thing. It works but it’s ultimately fitting a square peg in a round hole because the way that information moves through space in a stream network differs fundamentally from a 2-D terrestrial system.

Key to resolving the issue was that somebody had to develop a covariance structure that accounted for the unique topology of stream networks and the spatial configuration of samples measured on those networks. For example, streams flow in one direction, are linear features, have flow connected and unconnected sites, and tributary confluences where big changes occur over small distances. All these factors have consequences for how spatial patterns in networks manifest and how nearby sites influence one another. These network characteristics had to be embedded in the DNA of a new type of statistical model for streams if accurate renditions of spatial patterns in data measured on networks were to be made (graphics 2 and 3). I’m happy to say that that atom has finally been cracked and so this week & in a few subsequent blogs we’re highlighting work by two pioneers in the field of stream statistics, Jay Ver Hoef and Erin Peterson. This duo has published steadily on the topic over the last 6 years and have developed a rich literature and solid statistical foundation for their spatial stream network models (attached are a couple recent examples, a more complete bibliography is provided in graphic 4). I can’t do the work proper justice here since most of the math is over my head but I encourage you to spend some time with it and for those with especially strong quantitative skills and statistical bents, you’ll be in hog heaven.

For the rest of us, suffice it to say that a really good tool now exists for analyzing many types of data measured on stream networks. Moreover, the Ver Hoef and Peterson models make it possible, even desirable, to develop integrated, interagency databases because huge amounts of new and accurate information can often be extracted from existing data at relatively small costs and over the span of a few years. Thus, the “smart maps” we need at network scales to describe historical and future stream temperature patterns could be developed from databases like NorWeST (blog #25) or others to inform species conservation efforts (graphic 5). And although these maps won’t be perfect, they’ll be significant improvements over what’s currently available, and in the climate change game, we don’t have the luxury of time, let alone the budgets, to design & execute the perfect monitoring strategy and then wait decades for the answers. In many cases, also, the spatial stream models perform significantly better than their traditional counterparts in terms of raw predictive power and are enabling new types of analyses and inference about streams that simply weren’t possible before. Next time out, we’ll explore a bit of this in more detail but it’s truly the start of a new era when fundamentally better spatial information about streams and their biotas is becoming a real possibility. That’s good because it’s still going to take a lot of work to generate that information and we’re going to need it as we strive to conserve aquatic biodiversity through this transitional century.

Until next time, best regards.
Dan











Welcome to the Climate-Aquatics Blog. For those new to the blog, previous posts with embedded graphics can be seen by clicking on the hyperlinks at the bottom or by navigating to the blog archive webpage on our Forest Service site at: (http://www.fs.fed.us/rm/boise/AWAE/proj ... _blog.html). To discuss these topics with other interested parties, a Google discussion group has also been established and instructions for joining the group are also on the webpage. The intent of the Climate-Aquatics Blog and associated discussion group is to provide a means for the 4,161 (& growing) field biologists, hydrologists, anglers, students, managers, and researchers currently on this mailing list across North America, Europe, and Asia to more broadly and rapidly discuss topical issues associated with aquatic ecosystems and climate change.

Messages periodically posted to the blog will highlight new peer-reviewed research and science tools that may be useful in addressing this global phenomenon. Admittedly, many of the ideas for postings have their roots in studies I and my colleagues have been a part of in the Rocky Mountain region, but attempts will be made to present topics & tools in ways that highlight their broader, global relevance. Moreover, I acknowledge that the studies, tools, and techniques highlighted in future missives are by no means the only, or perhaps even the best, science products in existence on particular topics, so the hope is that this discussion group engages others doing, or interested in, similar work and that healthy debates & information exchanges will occur to facilitate the rapid dissemination of knowledge among those most concerned about climate change and its effects on aquatic ecosystems.

If you know of others interested in climate change and aquatic ecosystems, please forward this message and their names can be added to the mailing list for notification regarding additional science products on this topic. If you do not want to be contacted regarding future such notifications, please reply to that effect and you will be removed from this mailing list.

Previous Posts
Climate-Aquatics Overviews
Blog #1: Climate-aquatics workshop science presentations available online
Blog #2: A new climate-aquatics synthesis report

Climate-Aquatics Thermal Module
Blog #3: Underwater epoxy technique for full-year stream temperature monitoring
Blog #4: A GoogleMap tool for interagency coordination of regional stream temperature monitoring
Blog #5: Massive air & stream sensor networks for ecologically relevant climate downscaling
Blog #6: Thoughts on monitoring air temperatures in complex, forested terrain
Blog #7: Downscaling of climate change effects on river network temperatures using inter-agency temperature databases with new spatial statistical stream network models
Blog #8: Thoughts on monitoring designs for temperature sensor networks across river and stream basins
Blog #9: Assessing climate sensitivity of aquatic habitats by direct measurement of stream & air temperatures
Blog #10: Long-term monitoring shows climate change effects on river & stream temperatures
Blog #11: Long-term monitoring shows climate change effects on lake temperatures
Blog #12: Climate trends & climate cycles & weather weirdness
Blog #13: Tools for visualizing local historical climate trends
Blog #14: Leveraging short-term stream temperature records to describe long-term trends
Blog #15: Wildfire & riparian vegetation change as the wildcards in climate warming of streams
Blog #23: New studies describe historic & future rates of warming in Northwest US streams
Blog #24: NoRRTN: An inexpensive regional river temperature monitoring network
Blog #25: NorWeST: A massive regional stream temperature database
Blog #26: Mapping Thermal Heterogeneity & Climate in Riverine Environments

Climate-Aquatics Hydrology Module
Blog #16: Shrinking snowpacks across the western US associated with climate change
Blog #17: Advances in stream flow runoff and changing flood risks across the western US
Blog #18: Climate change & observed trends toward lower summer flows in the northwest US
Blog #19: Groundwater mediation of stream flow responses to climate change
Blog #20: GIS tools for mapping flow responses of western U.S. streams to climate change
Blog #21: More discharge data to address more hydroclimate questions
Blog #22: Climate change effects on sediment delivery to stream channels

Future topics…
Climate-Aquatics Biology Module
Climate-Aquatics Management Module
Climate-Aquatics Cool Stuff Module
_________________________________________________________
Dan Isaak - Fisheries Research Scientist
Rocky Mountain Research Station; U.S. Forest Service
322 E. Front St., Suite 401; Boise, ID 83702
Phone: (208) 373-4385; Fax: (208) 373-4391; Email: disaak@fs.fed.us
Relevant Websites: Climate-Aquatics Blog, Stream Temperature Webpage,
Scientist Webpage
__________________________________________________________





This electronic message contains information generated by the USDA solely for the intended recipients. Any unauthorized interception of this message or the use or disclosure of the information it contains may violate the law and subject the violator to civil or criminal penalties. If you believe you have received this message in error, please notify the sender and delete the email immediately.
Attachments
Peterson & Ver Hoef 10 Covariance structures for geostatistical models on stream networks.pdf
(251.24 KiB) Downloaded 759 times
VerHoef et al. 06 stream networks models that use flow & distance.pdf
(317.84 KiB) Downloaded 713 times

Post Reply