Data It

A Conceptual Design of Visualization of temporal vegetation data

Sarmistha Banerjee

Research Associate, National Institute of Design, Gandhinagar

*********@***.***, *********.********@*****.***

Shefali Agarwal

Scientist, IIRS, Dehradun

Dr.Connie Blok

Associate Professor, ITC, The Netherlands

Introduction:

A strategy of designing visualization is to transform the data in such a way that it

should create an easy, comprehensible environment for obtaining information of

the real environment. The sophisticated computing technology has opened up

numerous approaches to create visual displays which ultimately facilitate thinking

and problem solving. The aim is to create knowledge construction through the

designing of data visualization. When this concept of visualization deals with

geographic data it becomes geovisualization. Geovisualization involves

transferring the information about characteristics and nature of data to human

brain by exciting the sensory system through the proper use of graphics. Hence

to study the geospatial datasets in detail geovisualization is one of the ways.

Geovisualization has make room to represent the geospatial data through

innovative scene construction which is essential to study the complex

phenomena of natural and social sciences. It includes techniques of catography

and geographic information science. Cartographic techniques and methods are

mainly applied to translate the data into graphics, which include map like

products (Kraak, 1998).

Traditional maps only display the geospatial data but interactive maps offer more

exploration options. Interactivity enables the user to study the data from different

perspectives like combine map with other graphics, diagrams, photographs and

videos. The multivariate representations of geospatial datasets by different

creative graphics have more impact than traditional mapping methods. In this

context it can be said that maps are the tools for thinking about the

characteristics of geospatial data.

In the world many geographic data are dynamic in nature (e.g. weather data,

urban growth, vegetation growth) and many users are interested to view the

changes over time. The users not only want to view the changes but also to

analyze the phenomena. This calls for interaction with the data. Series of

interactive static maps is one of the ways to visualize such changes, but in this

case the sense of dynamics is very limited, it has to be constructed in the mind of

the user. Animation seems to be a suitable technique to represent the dynamism.

It is a subject of great interest among the computer graphics. It facilitates the

function for seeing the whole process of change. The study aims at integration of

graphics with animation and, in effect, developing an effective tool for exploring

temporal NDVI data.

Characteristics of spatio-temporal data:

On the earth surface each phenomena involve number of complex physical

processes. These processes vary spatially and temporally. To study these

complex processes advanced statistical and computational modeling were used

by geographers to explore and understand how geographic system function. In

today s world effort is being made to collect the information about these complex

processes through earth observation satellites. To monitor, explore and analyze

different geographic phenomena, spatio-temporal data plays a major role. The

analysis of spatio-temporal data helps to predict the future, analyze the changes,

do trend analysis, model generation and effect estimation (Blok, 2000). Hence

comprehensive understanding of spatio-temporal phenomena is necessary

before representing the complex geographic phenomena.

According to Peuquet spatio temporal data involves space (where), time (when),

and objects (what) (Peuquet, 1994, cited in Andrienko et al., 2002). It can be

explained as, during a specific time at a particular place an object can change its

characteristics or an object can change from time to time at different places. The

other one is object and location both is changing from to time. It can be explained

as:

When + where what

When + what where

Where + what when (Andrienko et al., 2003)

Based on this, the questions can be framed as:

What is the object present at location l during the time t1 ?

Where is the location of object o at time t1 ?

When the object o occupied the location l ?

All these questions are very common while dealing with the analysis of changing

spatial distributional pattern. Hence to answer these questions a proper

visualization is essential which can be helpful for monitoring the dynamics of the

object.

In this study temporal NDVI images have been used for entire INDIA for the

month of November, December, January, February, March, April, May, October

and December. Those are used along with the landuse/ landcover map for the

same region generated from the above NDVI images. It is derived from SPOT 4

VEGETATION data for the year 1999-2000. It has four spectral bands blue

(0.43-0.47mm), red (0.61-0.68mm), infra-red (0.78-0.89mm) and short wave

infrared (1.58-1.75mm) at a spatial resolution of 1 km and temporal resolution of

1 day. The plant pigments chlorophyll a and b absorbs wavelengths of 0.66 mm

and 0.65 mm respectively in the red region and infrared wavelength reflected in

the wave region 0.78 0.89 mm is not interfered by water absorption. This gives

precise information of amount of green vegetation on the earth surface calculated

as ratio of red and infrared wavebands in NDVI. It represents the ratio of

The values range from 1 to +1. Negative values indicate water, zero indicates

bare soil and positive value indicates healthy vegetation. In this study NDVI

values are rescaled from 0 to 255. These are the general characteristics of the

data.

The specific characteristics of temporal NDVI data are:

It is a ratio type data.

Vegetation growth is a discrete phenomena.

Vegetation growth takes longer time so the duration of this change event

is long.

Magnitude of change is different at different places.

Magnitude of change is different in different season.

Number of physical factor is responsible for changing for example climate,

physiography, soil

In this study the major concern is to represent the specific characteristics of

temporal NDVI data with a proper visualization technique which can be helpful to

answer the three fundamental queries of user like where? When? What? Apart

from the study of data characteristics, user study is constantly required to design

any kind of data visualization.

User s requirement for exploration of spatio-temporal data:

To study the spatio-temporal data users have different and distinct needs. It

varies between presentation and exploration of data. The needs are made up of

series of tasks or a list of processes that are undertaken and invoked within the

confines of the users thought processes (Ogao, 2002). For example,in the field

of forestry a user would like to use the spatio temporal NDVI data for monitoring

the vegetation growth, for understanding the phenological pattern of different

vegetation species and to discriminate them, for calculation of biomass, to

classify the land cover and to identify the cropping pattern etc. (From a

discussion with faculty and researchers working in the field of forestry at IIRS,

Dehradun).

In this case the user may use different statistical operations in order to achieve

the above objectives. In the field of geovisualization to explore these events

numerous techniques could be applied for the exploration of spatio-temporal

characteristics. For example to explore the temporal datasets static techniques

may not be able to fulfill the user s task. Alternative techniques (like linking

between maps and different datasets, interactive manipulable maps etc.) could

be applied for this type of datasets. Identify, locate, associate and compare form

the basic visualization operations that are widely used for analyzing spatio-

temporal datasets (Ogao, 2002).

Hence user s requirement for studying the dynamic events using spatio-temporal

datasets is an important aspect along with data characteristics during the

designing phase of visualization tool. For example, if a tool is developed for

visualization of temporal NDVI data using animation, without considering some of

the user s need like extensive query operations related to locational and

attributinal changes, linking with other climatic database, then the visualization

tool would be simply an attractive animation movie with all kinds of basic display

functionalities but will be of no use to the user, in terms of applicability.

Different aspects of spatio-temporal vegetation data:

The data used in this research contains the vegetation reflected value. The

values are not constant but changing from time to time and area to area. The

values are indicated by NDVI. For example in case of time t1 the pixel values of

an area is same, in time t2 a portion of the area has changed( in the figure-1 it

showing by x1 value). In time t3 more changes can be seen like in the figure1

from x, x1, x2.

Figure1: Pixel values are changing from time to time

In each successive time period changing values are not uniform and changing

pattern is not uniform. In spatio-temporal data the changes are generally

locational, thematic and attribute in nature. In present temporal data the changing

pixel values indicate the locational changes and amount of values indicates the

attribute changes. In the above figure nine cells are representing nine pixels. It is

a case of SPOT-4 images. Each pixel are representing 1 km area. If one area

have same type of species then it will reflect same values (e.g. in the figure 1,

four cells are showing x1 values in t2.) and that values may change from time to

time (e.g. in t3 the previous x1 values become x2) due to the chlorophyll

concentration. It is applicable for other areas also. In reality different types of

vegetation cover can be there within 1km area. Therefore growth rate also varies

within one kilometer area. All these changes are depend upon physiography,

species, climate and chlorophyll concentration. In this research the main

emphasis is to represent the locational changes and attribute changes through

animation.

In this dataset geometrical aspect is not considered. NDVI images showing

complex phenomena where shape and size is not easy to identify. The main

focus is to study the changing spatial distribution of objects. Changing spatial

distribution is related to three basic questions. According to Pequet (Peuquet,

1994, cited in Andrienko et al., 2002) these are when, what and where.

As vegetation growth rate is a continuous phenomena and its changing aerial

extent is not definite therefore in this study aerial changes is not considered in

case of representing the locational changes. The study concentrates mainly on

about the representation of changing vegetation growth rate at different locations

due to chlorophyll

concentration. Hence pixel

wise changing growth rate

has to determine. The

concept is when one is

able detect the landuse

changes by changing pixel

value then it becomes

easier to identify the types

of changing landuse by the

changing nature of pixel

value of particular land

class. Thereafter

measurement of changing

vegetation growth rate is

important to explore the

datasets.

The design concept is Figure 2: Conceptual Framework for the design of Prototype

described in flow chart

Figure 2. The concept is

detection, identification and measurement simultaneously should be

incorporated with animation though the significance of these three tasks are

different. Through animation locational and geometrical change (landuse change)

can be recognised. While attributinal change (vegetation type, species, landuse

class) can be recognised through statistical information and changing property of

object (changing chlorophyll concentration) can be visualised through direct look

up technique. An option should be there that functioning of these three steps

(mentioned in Figure 2) also present in static mode. Thus the combination of

dynamic and static approach can reveal more information through visualization.

Prototype Design Concept:

The overall prototype design is based on data characteristics and user s

requirement for the exploration of NDVI images.

In the prototype nine datasets have been used. Some data are missing. Datasets

of July, August and September are not used due to presence of cloud cover. To

detect and identify the changes simple animation has been used. It is a frame

based animation and is built in Macromedia FlashMX. Action script is used to

make all basic functionalities like zoom, pan, play, back, stop and other buttons.

No interpolation is made for making animation. Legend in the map is also

interactive. By clicking on any of the colour it will show the corresponding pixel

value. Movie showing state wise changes, selection of static images and graph

movie is also made in flash using action script. The prototype is named as

TEMPVIZ .

In the TEMPVIZ spectral profile, histogram and textual information are the main

graphics linked with animation. A single pixel for all nine images is important for

change detection and identification for the present datasets. On the other hand

information of a single pixel value and how it changes with time is useful for

studying the phenological variation of the particular landuse category, however to

take into account the variability of the pixels within a particular class the pixels in

the spatial neighbourhood should also be considered. Hence an option to

generate the profile using the nearby 3*3 pixels is also provided.

The main objective being that while animation is running user can interact by

clicking at any pixel and it will give information in terms of spectral profile of a

particular landuse class/pixel. This mouse clicking functionality is present with

static mode as well as dynamic mode. A separate window is also present in

TEMPVIZ which provide rgb value and xy coordinate of points. (In order to

relate the pixel to the geographic location). To increase the user interactiveness

a window provides a facility so that the user can select any image and is able to

get statistical information like mean, median, mode and the histogram plot of

whole month. It can help to compare the statistical information from month to

month. Thus the combination of animation and static view of same datasets can

help to analyse the phonological variation in detail.

Figure 3: Functionality of TEMPVIZ

Figure 4: Main Display Window

Multiple views are useful for

visual comparison of different state of change. In animation mode user is able to

detect the changes and on the other hand static view of same image can help

him/ her to compare the changes visually. Animation within animation (state wise

animation) has made to facilitate the view of local changes as per the user s

interest. It can be a debatable issue. But when an animation is showing changes

of large area then to view the changes of small area this type of representation

could be effective.

Despite all the functionalities the prototype TEMPVIZ has some limitations:

It lacks GIS functionality: queries at any spatial location cannot be carried

out in a dynamic mode.

No database is linked with this prototype.

Only NDVI values have been considered here. Other parameters like

rainfall, temperature, which are responsible for vegetation growth, have

not been used.

Pixel information is also available for some sample pixels not for the entire

image.

No other temporal datasets can be used in this prototype.

It is a conceptual design, which can be improved.

Analysis of the design:

The research prototype has been evaluated through user testing methods. In this

evaluation focus group in combination with questionnaire method has been

followed. User s view about the usability of the prototype TEMPVIZ is analyzed

quantitatively in terms of effectiveness, efficiency and satisfaction.

The overall response about effectiveness, efficiency and satisfaction are: 41.65%

participants expressed that the prototype is moderately effective and 24.85% of

participants stated that it is highly effective. While 44.43% of participant

expressed that it is moderate in terms of efficiency and 27.95% of participants

addressed that it is a highly useful tool. On the other hand 75.5% participants are

moderately satisfied with the system and 25.5% of participants expressed that

they are highly satisfied with system as it fulfils their requirement.

Animation with linked graphics technique is moderately useful in terms of

informativeness and coherence. It is one of the effective ways for representing

spatiotemporal data. Though some parts of the design need improvement but

overall the users in terms of attractiveness and applicability in the particular

discipline appreciated the prototype design. The positive response on the

development of such type of visualization tool is a supporting point for carrying

out further research and improvements for designing of such type of visualization

tool.

Conclusion and recommendation:

The study reveals that visualization of spatio-temporal data is mainly depend on

nature of the data and user s need. On the basis of that numerous

geovisualization techniques can be applied to design a visualization tool.

Traditional cartographic techniques are not always helpful to analyze and

interpret the whole scenario or phenomena. In this research it is found that

simple animation techniques are not that effective to visualize the locational

changes. Sometimes locational changes needs to be addressed along with

attributes and geometrical properties for better perception and analysis, in such

cases animation can be combined with linked graphics. Currently there is no

proper theoretical and methodological background for the development of such

type of visualization tools. The entire technique is based on user s need and

characteristics of datasets. Hence an attempt was made to conceptually design a

visualization tool for exploring and analysing the temporal NDVI datasets which

has utility in various application like vegetation monitoring, crop monitoring etc.

Finally, it has lot of scope to do further research like interface can be made more

user friendly in terms of adding sound, video, photograph particularly for this kind

of datasets. If dynamic variables are added with this technique then it will

increase more usability and interactiveness for the exploration of temporal

datasets. If database is also combined to it then it will be more useful to enhance

the utility of the visualization tool and increase the applicability of the prototype.

Like for extensive analysis and using the NDVI datasets other parameters like

rainfall, temperatures, which are responsible for vegetation growth, should be

incorporated. It is well known fact that different exploratory techniques are

required for different types of spatio-temporal data therefore to develop a useful

dynamic visualization technique this animation with linked graphics techniques

should be experimented by using different types of temporal data.

Bibliography:

Andrienko, N., Andrienko, G., Gatalsky, P. (2003), Exploratory spatio-temporal visualization: an

analytical review. Journal of Visual Languages & Computing, Vol.14, pp. 503-541.

Blok, C. (1998), Dynamic visualization in a developing framework for the representation of

geographic data. Bulletin du Comite Francais de Cartographie, 156, pp. 89-97

Geresy, E.B. & Jones, C. (2000), Models and Queries in a spatio-temporal GIS. In: Atkinson, P. &

Martin, D. (ed.), GIS and Geocomputation. New York: Taylor & Francis, pp.27- 39

Kraak, M.J. (1998), Exploratory Cartography: Maps as tools for discovery. Paper presented at the

International Cartographic Association Commission on Visualization Workshop in Warsaw,

Poland: May 21-23, 1998

Roy, P. S., Agrawal, S., Shukla, Y., Joshi, P.K. (2004), Land cover mapping using

SPOTVEGETATION for South Central Asia. TREES 2- GLC 2000/ IIRS-JRC. India. Indian

Institute of Remote Sensing

Contact this candidate