The SWEVIS R Package for Forecasting and Visualization of Snow Water - - PowerPoint PPT Presentation

The SWEVIS R Package for Forecasting and Visualization of Snow Water Equivalent Data

James B. Odei

The Ohio State University Joint Work With

Jürgen Symanzik

Utah State University

June 12, 2015

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 1

Outline

.

1

Introduction

2

Goals of this Presentation

3

Conclusions & Future Work

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 2

Introduction: Background

The intermountain region of the Western United States comprises of a variety of ecological and economic systems Snowpack – accounts for 50 to 70% of the annual precipitation in the intermountain regions (Serreze et al., 1999) Over 75% of its water resources results from snowmelt water Multi-year droughts in the Southwest have severely affected supplies according to a report from the National Climatic Data Center These droughts are among major natural risks this region’s residents and ecosystems are facing To forecast water resources, the National Weather Service (NWS) maintains a set

• f conceptual, continuous, hydrologic simulation models used to generate

extended streamflow outlooks, and flood forecasts

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 3

Presentation Goals

Goal 1: Developed Statistical Model to Forecast Snow Water Equivalent (SWE) Data (see Odei et al., 2014) Goal 2: New R Package for Visualization and Exploration of Spatial and Spatio-Temporal SWE Data Goal 3: To apply the Newly Developed R Package Using Utah SNOTEL Sites and Upper Sheep Creek Site in Idaho as Case Studies

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 4

Presentation Goals: Goal 1: A Bayesian Hierarchical Model – Result 1 Tony Grove SNOTEL Site, Utah – 2008 Water-Year

• Oct. 1
• Jan. 1
• Apr. 1
• Jul. 1

20 40 60 SWE (inches)

10 11 12 1 2 3 4 5 6 7 8 9

• Jan. 8, 2008 −−
• prev. data env.
• pred. CI: 50%
• pred. CI: 95%
• prev. data mean

current data months 5th & 95th perc.

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 5

Presentation Goals: Goal 1: A Bayesian Hierarchical Model – Result 2 Horse Ridge SNOTEL Site, Utah – 2009 Water-Year

• Oct. 1
• Jan. 1
• Apr. 1
• Jul. 1

10 20 30 40 SWE (inches)

10 11 12 1 2 3 4 5 6 7 8 9

• Jan. 8, 2009 −−
• prev. data env.
• pred. CI: 50%
• pred. CI: 95%
• prev. data mean

current data months 5th & 95th perc.

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 6

Presentation Goals: Goal 1: A Bayesian Hierarchical Model – Result 3 Little Bear SNOTEL Site, Utah – 2010 Water-Year

• Oct. 1
• Jan. 1
• Apr. 1
• Jul. 1

5 10 15 20 25 30 10 11 12 1 2 3 4 5 6 7 8 9

• Feb. 7 −−
• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 7

Presentation Goals: Goal 2: SWEVIS R Package

“Visualization refers not only to a set of graphical images but also to the iterative process of visual thinking and interaction with the images" (Edsal et al., 2000) Visualization can bring to light subtle patterns that may not be immediately apparent in strictly quantitative data analysis methods

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 8

Presentation Goals: Goal 2: Types of Spatial Data

Spatially continuous data (also called geostatistical data) Data sampled at fixed point locations with spatial variation in a variable varying continuously over the study area Areal data (also called lattice data) The variable of interest does not vary continuously, but has values only within a fixed set of areas or zones covering the study area Other types of spatial data – spatial point patterns and spatial interaction data

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 9

Presentation Goals: Goal 2: ESDA & Visualization Tools

Exploratory Data Analysis (EDA) techniques (boxplots, histograms, and scatterplot matrices) ignore special characteristics of spatial data like spatial dependence and spatial heterogeneity (Anselin, 1990) Exploratory Spatial Data Analysis (ESDA) provides a set of robust tools for exploring spatial data ESDA methods are used to detect spatial patterns of the data, formulate hypotheses based on the geography of the data, and assess spatial models

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 10

Presentation Goals: Goal 2: ESDA & Visualization Tools

For areal/lattice data – most widely used visualization techniques are based on choropleth maps A choropleth map in grey scale showing the proportion of non-white births in North Carolina, 1974–1978. Source: Bivand et al. (2008)

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 11

Presentation Goals: Goal 2: ESDA & Visualization Tools

For spatially continuous data – variogram cloud plot used to gain insight into the covariance structure and visualize the spatial association Squared-differences variogram cloud for the scallops data. Source: Kaluzny et al. (1998)

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 12

Presentation Goals: Goal 2: Linked Brushing

Multiple visualizations through interactive linking and brushing provide more information than considering the component visualizations independently Linking shows how a point, or set of points, behaves in each of the plots In brushing, points to be highlighted are interactively selected by a mouse and the plots are dynamically updated (ideally in real time) Linked brushing – one of the most powerful interactive tools for doing exploratory data analysis using visualization

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 13

Presentation Goals: Goal 2: SWEVIS R Package

The newly developed SWEVIS R package provides the following features and plots Spatial data manipulation and utilities: input of SWE data in a designed matrix format Forecasting: using the statistical model discussed in Goal 1 Mapping: maps from RgoogleMaps, heat maps, and image plots in a linked environment EDA and ESDA: statistical graphics like histogram, box plot, scatter plot and variogram cloud plots linked to a map view Linked Brushing: connecting map displays from RgoogleMaps and EDA/ESDA graphics from iPlots Variogram cloud plot: one-to-two linking/brushing between statistical plots and map view

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 14

Presentation Goals: Goal 2: SWEVIS Functions

The newly developed SWEVIS R package consists of 16 main functions Functions to read/store/manipulate SWE data

– ReadSweData, ReadSweAsciiData – CalcSweSumStat, SimSweMCMCData

Plotting functions

– RawSweDataPlot, SweBoxPlot, SweHistPlot, SwePostPlot – SweVariogPlot, SweRgoogleMap, SweAsciiImagePlot

Interaction functions

– iSwePlot, iSweAsciiPlot – iSweBrushMapSingle, iSweBrushMap, iSweBrushPlot

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 15

Presentation Goals: Goal 3: Application of SWEVIS R Package

End users of the R package proposed in Goal 2 are:

– from environmental agencies – individuals interested in the daily amount of snow measurements

We present two case studies that make use of the functionality from our newly developed R package Will use SWE data from (i) the SNOTEL sites in Utah and (ii) the Upper Sheep Creek (USC) Watershed in Idaho

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 16

Presentation Goals: Goal 3: Utah SNOTEL Data

100 200 mi scale approx 1:4,800,000

1 2 3 4 5 6 7 8 9 10

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 17

Presentation Goals: Goal 3: Single SNOTEL Site Visualization

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 18

Presentation Goals: Goal 3: Multiple SNOTEL Sites Visualization

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 19

Presentation Goals: Goal 3: Upper Sheep Creek Watershed Data

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 20

Presentation Goals: Goal 3: Upper Sheep Creek Watershed Data

Topography and instrument locations within the Upper Sheep Creek Watershed (Previously published as Figure 1 in Flerchinger and Cooley (2000))

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 21

Presentation Goals: Goal 3: Upper Sheep Creek Watershed Data

Snow water equivalent measured in Upper Sheep Creek March 3, 1993. The dots represent locations of the grid stations where measurements of snow water equivalent (SWE) were taken. No grid stations are available at points 9N and 25D and point L10 was not measured (Previously published as Figure 2 in Luce and Tarboton (2004))

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 22

Presentation Goals: Goal 3: USC Watershed Image Plots

522400 522600 522800 523000 4774100 4774300 4774500 February 10 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

522400 522600 522800 523000 4774100 4774300 4774500 March 03 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

522400 522600 522800 523000 4774100 4774300 4774500 March 23 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

522400 522600 522800 523000 4774100 4774300 4774500 April 08 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

522400 522600 522800 523000 4774100 4774300 4774500 April 15 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

522400 522600 522800 523000 4774100 4774300 4774500 April 29 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

522400 522600 522800 523000 4774100 4774300 4774500 May 12 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

522400 522600 522800 523000 4774100 4774300 4774500 May 19 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

522400 522600 522800 523000 4774100 4774300 4774500 May 25 x y

(0, 0.5] (0.5, 1.0] (1.0, 1.5] (1.5, 2.0] (2.0, 2.5] (2.5, 3.0]

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 23

Presentation Goals: Goal 3: USC Watershed Difference Plots

522400 522600 522800 523000 4774100 4774300 4774500 1st−Difference 1 x y

(−1.2, −0.8] (−0.8, −0.4] (−0.4, 0.0) 0.0 (0.0, 0.4] (0.4, 0.8] (0.8, 1.2]

522400 522600 522800 523000 4774100 4774300 4774500 1st−Difference 2 x y

(−1.2, −0.8] (−0.8, −0.4] (−0.4, 0.0) 0.0 (0.0, 0.4] (0.4, 0.8] (0.8, 1.2]

522400 522600 522800 523000 4774100 4774300 4774500 1st−Difference 3 x y

(−1.2, −0.8] (−0.8, −0.4] (−0.4, 0.0) 0.0 (0.0, 0.4] (0.4, 0.8] (0.8, 1.2]

522400 522600 522800 523000 4774100 4774300 4774500 1st−Difference 4 x y

(−1.2, −0.8] (−0.8, −0.4] (−0.4, 0.0) 0.0 (0.0, 0.4] (0.4, 0.8] (0.8, 1.2]

522400 522600 522800 523000 4774100 4774300 4774500 1st−Difference 5 x y

(−1.2, −0.8] (−0.8, −0.4] (−0.4, 0.0) 0.0 (0.0, 0.4] (0.4, 0.8] (0.8, 1.2]

522400 522600 522800 523000 4774100 4774300 4774500 1st−Difference 6 x y

(−1.2, −0.8] (−0.8, −0.4] (−0.4, 0.0) 0.0 (0.0, 0.4] (0.4, 0.8] (0.8, 1.2]

522400 522600 522800 523000 4774100 4774300 4774500 1st−Difference 7 x y

(−1.2, −0.8] (−0.8, −0.4] (−0.4, 0.0) 0.0 (0.0, 0.4] (0.4, 0.8] (0.8, 1.2]

522400 522600 522800 523000 4774100 4774300 4774500 1st−Difference 8 x y

(−1.2, −0.8] (−0.8, −0.4] (−0.4, 0.0) 0.0 (0.0, 0.4] (0.4, 0.8] (0.8, 1.2]

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 24

Presentation Goals: Goal 3: USC Watershed Interactive Plots

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 25

Conclusions & Future Work: Conclusions

Our new R package developed

– provides great potential for use in various environmental agencies and individual uses – provides additional insights into snow water equivalent data – allows for linked-brushing connecting map views and statistical graphics from iPlots – allows for analyses of SWE measurements for single and multiple SNOTEL locations

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 26

Conclusions & Future Work: Future Work

Extension of our newly developed R package

– Linked to a database with SWE data at the operational level – Interaction tools like zooming and animation – Spatially lagged scatterplots to access local instability in spatial association – Consider cut-off distance in the variogram cloud plot – Consider making USC image plots and difference plots interactive – Upgrade to iPlots eXtreme (also known as Acinonyx)

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 27

. .

Thank you!

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 28

References

.

Anselin, L. (1990). What is Special About Spatial Data? Alternative Perspectives On Spatial Data Analysis, in D. A. Griffith, ed., Spatial Statistics, Past, Present and Future. Institute of Mathematical Geography, Ann Arbor, Michigan, pp. 63–77. Bivand, R. S., Pebesma, E. J. and Gomez-Rubio, V. (2008). Applied Spatial Data Analysis With R, Springer, New York, New York. Edsal, R. M., Harrowering, M. and Mennis, J. L. (2000). Tools for Visualizing Properties of Spatial and Temporal Periodicity in Geographic Data. Computers & Geosciences, 26, 109–118. Flerchinger, G. N. and Cooley, K. R. (2000). A Ten-Year Water Balance of a Mountainous Semi-Arid Watershed. Journal of Hydrology, 237, 86–99. Kaluzny, S. P ., Vega, S. C., Cardoso, T. P . and Shelly, A. A. (1998). S+SPATIALSTATS, User’s Manual for Windows and UNIX, Springer, New York, New York. Luce, C. H. and Tarboton, D. G. (2004). The Application of Depletion Curves for Parameterization of Subgrid Variability of

• Snow. Hydrological Processes, 18, 1409–1422.

Odei, J. B., Symanzik, J. and Hooten, M. B. (2014). A Bayesian Hierarchical Model for Forecasting Intermountain Snow Dynamics, Environmetrics, 25(5), 324–340. Serreze, C. M., Clark, P ., Amstrong, R. L., McGinnis, D. A. and Pulwarty, R. S. (1999). Characteristics of Western U.S. Snowpack from Snotel Data Water Resources Research, 35, 2145–2160.

• J. B. Odei (OSU) & J. Symanzik (USU)

The SWEVIS R Package for Forecasting and Visualization of SWE Data June 12, 2015 29