Land usage alteration is a major issue of planetary environment alteration. The mold and projecting of land usage alteration is indispensable to the appraisal of attendant environmental impacts. Recent development of theoretical accounts such as cellular zombi ( CA ) provides a powerful tool for the dynamic mold of land usage alteration. This research aims at look intoing urban kineticss of Nakuru Municipality between 1986 and 2030 utilizing three different land alteration theoretical accounts Markov Chain, Cellular Automata and Geomod. Classified informations of two Landsat clip series orbiter images ( 1986, and 2000 ) obtained after image cleavage and a texturing process will be used as input together with suitableness maps to foretell urban growing scenarios with two categories ( urban ; non urban ) . Models ‘ proof and public presentation will be assessed comparing predicted twelvemonth of 2000 against twelvemonth 2000 existent categorization utilizing several statistical indices and CPU processing clip. Two different declarations will be tested: pels and the objects obtained by image cleavage for the categorization of image of 2000. Consequences will bespeak that one must utilize same degree of abstraction employed in image categorization to measure truth of theoretical accounts ‘ end products.
Keywords: Cellular Automata, Geomod, Markov Chain, Nakuru Municipality, urban growing, urban simulation
ACRONYMS AND ABBREVIATIONS
CA Cellular Automata
ETM Enhanced Thematic Mapper
FCC False colour complex
GIS Geographic Information System
GPS Global placement systems
ISODATA Iterative ego forming informations algorithm
MSS Multi-spectral scanner
NIR Near Infra Red
NDVI Normalized Difference Vegetation Index
PCA Principal Component Analysis
RS Remote feeling
TM Thematic Mapper
UTM Universal Transverse Mercator
WGS World Geodetic co-ordinate System
Urban surveies are going of import tools for contrivers cognizing that in 2015 more than half universe ‘s population will be populating in metropoliss, ( UNECE, 2003 ) . Models are, possibly, the best manner of understanding the land alteration phenomenon and anticipate correct planning activities for sustainable metropoliss. This is an of import subject in current research docket and a important figure of scientists are giving attempts in the survey of this phenomenon, ( Batty et al. , 1999, Cheng, 2003, Herold et Al, 2003 ) .
The determination about which theoretical account to utilize may non be an easy undertaking. Complexity of the phenomenon to pattern, information demands and type of end products vary significantly and one must leverage all these factors before taking the determination of taking a theoretical account. In this survey, we compare three theoretical accounts to analyze urban growing in Nakuru municipality. These include Markov Chain Models, Cellular Automata ( CA ) and GEOMOD implemented in Idrisi Kilimanjaro. Markov type theoretical accounts are based on the Markov ironss proposed by the Russian mathematician Andrei A. Markov in 1907. These theoretical accounts merely became spatially explicit in early 1990s when they started incorporating a cellular zombi constituent leting passage chances of one pel to be a map of the neighbouring pel. Several urban surveies have been based in the usage of CA and Markov ironss. Geomod is a more recent and simpler theoretical account and was originally designed to imitate the loss of tropical woods and to gauge the ensuing C dioxide emanations. This theoretical account uses the measures specified by the user ( alternatively of a passage matrix ) and a suitableness map to imitate alteration of a individual class utilizing a additive relationship between get downing and stoping clip sums.
The most employed standard for formalizing land-cover alteration theoretical accounts is based on the per centum of right classified pels obtained from comparing of a existent map against the end product of the theoretical account. However, a high figure of right classified pels does non intend that the theoretical account has a good prognostic power between two clip minutes due to temporal autocorrelation.
In this survey, the public presentation of theoretical accounts will be assessed comparing truth of estimated map of 2000 against the void theoretical account ( i.e. , no alteration ) utilizing different statistical indices viz. Kappa statistic, Percent correct. In this research, two different declarations are employed in theoretical account proof and truth appraisal: pels and objects obtained by twelvemonth 2000 image cleavage. Image objects proof was sensible because image categorization was performed over objects obtained by image cleavage.
In this research, the modellers will be used to analyze land usage alteration and anticipation of future tendencies in Nakuru Municipality.
Statement of the job
Natural resource development undertakings are more and more making usage of distant detection and Geographic Information Systems ( GIS ) . But for many undertakings the entree to informations is limited and non up to day of the month. Sustainable direction of natural resources requires a functional system that provides efficient and accurate information. ( Andre Nagelhout, 2001 )
How will the three theoretical accounts execute?
Which theoretical account will give most accurate consequences?
Is Landsat imagination sufficient for the research?
Can GIS be used to reliably predict alteration?
The application of a stochastic mold techniques in GIS, Markov Chain Analysis, cellular zombi, and Geomod with categorised land usage informations derived from satellite imaginations in Nakuru municipality
To quantify land usage land screen alterations
The survey country
Nakuru municipality lies about between latitudes 0A° 15 ‘ and 0A° 31 ‘ South, and longitude 36A° 00 ‘ and 36A° 12 ‘ East, with an mean height of 1,859 metres above sea degree, covering an country of 230kmA? ( Figure 1 ) . Within Nakuru municipality lies Nakuru town, Lake Nakuru National Park, and Lanet town. Nakuru town is located 160 km North West of Nairobi and is the 4th largest urban Centre in Kenya after Nairobi, Mombassa and Kisumu.
Justification of the survey
Why Markov simulation, celullar zombi and Geomod
Clark Labs has developed some new spacial statistical faculties for Idrisi GIS.
The spirit of wonder that the current survey will be undertaken and besides to see what these new mold faculties in Idrisi GIS could make.
The range of this research is to supply a comprehensive reappraisal of different techniques and algorithms that are used to deduce utile information from remotely sensed digital images ( obtained chiefly through the Thematic Mapper and Enhance Thematic Mapper ) to bring forth informations merchandises that are both appropriate to, and instantly useable within different scientific applications. Modeling the alterations in Nakuru municipality will uncover land usage land screen alterations.
Failing of the survey
Uncertainties in the modellers
Handiness of informations e.g. informations sets required have different beginnings of which some organisations possibly loath to provide such informations.
Incompatibility of informations sets e.g. different co-ordinate systems
Technical complications e.g. deficient clip for land truthing, hardware and package complications
GIS database for the survey country
Thematic maps with statistical computations
Final undertaking study depicting all research elements and merchandises
Definition of cardinal footings
Remote Sensing is the scientific discipline and art of geting information ( spectral, spacial, and temporal ) about stuff objects, country, or phenomenon, without coming into physical contact with the objects, or country, or phenomenon under probe.
Geographic Information System
Geographic Information System ( GIS ) is a computing machine based information system used to digitally stand for and analyze the geographic characteristics present on the Earth ‘ surface and the events ( non-spatial properties linked to the geographics under survey ) that taking topographic point on it.
Land screen and land usage
The term land screen relates to the type of characteristic nowadays on the surface of the Earth. Corn Fieldss, lakes, maple trees, and concrete main roads are all illustrations of land screen types. The term land usage relates to the human activity or economic map associated with a specific piece of land. As an illustration, a piece of land of land on the periphery of an urban country may be used for single-family lodging. Depending on the degree of mapping item, its land usage could be described as urban usage, residential usage, or single-family residential usage. The same piece of land of land would hold a land screen consisting of roofs, paving, grass, and trees. ( Lillesand, 1999 )
Chapter Two: LITERATURE REVIEW
The preparation of regional and urban policies requires up-to-date socio-economic and land usage informations. In add-on a model is required to hive away disparate datasets and to dynamically prove scenarios and schemes. Spatial modeling are analytical processs that are applied to spacial datasets and such techniques can include spacial interaction, location-allocation and web modeling. These processs are undertaken within a Geographic Information System ( GIS ) , which are used to capture, shop, show and analyse spacial informations stand foring characteristics on the Earth ‘s surface.
These integrated analytical environments are normally referred to as Spatial Decision Support Systems ( SDSS ) . They have emerged as a consequence of the progressively complex inquiries that urban contrivers face in trying to do reciprocally consistent, long-run programs ( Wadell, 2001, Bailey and Gatrell, 1995 ) . SDSS developed in line with the progresss in related engineerings, such as GIS, remote feeling and multi-criteria determination support systems, all of which are cardinal to successful sustainable land-use and transit planning.
Predictions of future land screen are of import for a figure of preservation and Restoration ends, including aiming countries for Restoration, measuring the impacts of possible Restoration and extenuation scenarios, and finding the exposures of assorted resource lands to future land conversion.A
Models are defined diversely. They can be considered as the formal representation of some theory of a system of involvement ( Wilson 1974, 4 ) . More loosely, theoretical accounts can be considered as abstractions, estimates of world which is achieved though simplification of complex existent universe dealingss to the point that they are apprehensible and analytically manageable. The representation of world is expressed through the usage of symbols.
Models of land usage alteration – Categorization
Eight interconnected beginnings of fluctuation, in a approximately diminishing order of importance, can be discerned in extant theoretical accounts: the intent for which the theoretical account is built, the theory ( or, the deficiency of it ) underlying the theoretical account ( reflecting, in portion, the types of the determiners of land usage alteration taken into history ) , the spacial graduated table and degree of spacial collection adopted every bit good as the grade of “ spacial explicitness ” of the theoretical account, the types of land usage considered as chief objects of analysis, the types of land usage alteration processes considered, the intervention of the temporal dimension ( which in the instance of analysis of alteration, in general, should be built-in in any undertaking ) , and the solution techniques used. Hence, there exist:
descriptive, explanatory, normative, prognostic and impact appraisal theoretical accounts
micro-economic and macro-economic theoretic theoretical accounts, gravitation or spacial interaction theory-based theoretical accounts, incorporate theoretical accounts every bit good as a-theoretic theoretical accounts
local, regional, interregional, national and planetary degree theoretical accounts
geo-referenced ( to the full spatially expressed ) and non-geo-referenced ( incompletely spatially expressed ) theoretical accounts
urban ( largely residential ) , agricultural ( harvest ) , forest sector theoretical accounts
deforestation, urbanisation, etc. theoretical accounts
inactive, quasi-static ( or, quasi-dynamic ) and dynamic theoretical accounts ( nevertheless counterintuitive inactive theoretical accounts of alteration may sound )
statistical, programming, gravity-type, simulation and incorporate theoretical accounts.
Chapter Three: Methodology
For the intent of defining land utilizations in the survey country which includes flora, agribusiness, urban and H2O categories, Bands 1, 2, 3, 4 and 5 of Landsat images will be used.
Assorted false colour compositing options exist and assorted objects appear otherwise in different set combinations. Band combination increases the ocular visual aspect of characteristics. For case 432 false colour set combinations will be used flora and H2O extraction.
The individually classified images for the two old ages will so be analyzed in the three modellers viz. Markov Chain, Cellular Automata and Geomod. Statistics will be generated indicating alterations and truth of each of the methods.
This research will use two types of datasets. The first type will be remotely sensed informations as a major information for processing and thematic map production. The 2nd type will be the topographic maps of scale 1:50000 of the survey country.
Remotely sensed informations
Landsat TM & A ; ETM digital informations of 1986 and 2000 will be employed in this survey. This information set will be supplied by the Regional Centre for Mapping of Resources for Development ( RCMD ) .
Topographic maps of scale 1:50000 of the survey country will be obtained from Surveying and Photogrammetry lab, Jomo Kenyatta University of Agriculture and Technology ( JKUAT ) .
Land truth informations will include:
Land truth points indiscriminately obtained in the survey country utilizing GPS
The suitableness maps will be obtained from FAO Afri-cover datasets which will include incline, population, land screens etc.
Data and analysis
Landsat imaginations of survey country.
1986, 2000 information sets will be used
Land cover categorization with Idrisi
Land usage categories will be re-classed to 6 classs for each image for the modellers:
Drumhead statistics, Markov probabilites, cellular zombi simulation and Geomod simulation done with Idrisi Kilimanjaro GIS.
Description of the Models and Calibration
Markov Chain Analysis
It ‘s an sum, macroscopic, stochastic, patterning procedure. Predictions of future alteration are based on alterations that have occurred in the yesteryear. Markov analysis can be used in three different ways: for ex-post impact appraisal of land usage ( and associated environmental ) alterations of undertakings or policies, for projecting the equilibrium land usage vector every bit good as for come closing the clip skyline at which it may be obtained and projecting land usage alterations at any clip in the hereafter given an initial passage chance matrix. Imagine an country subdivided into a figure of cells each of which can be occupied by a given type of land usage at a given clip. On the footing of observed informations between clip periods MCA computes the chance that a cell will alter from one land usage type ( province ) to another within a specified period of clip. The chance of traveling from one province to another province is called a passage chance ( Turner, 1987 ) .
Markov Chain Analysis in Idrisi Kilimanjaro
MARKOV takes two qualitative land screen images from different day of the months and generates the undermentioned files:
A passage matrix. Contains the chance that each land screen class will alter to every other class
A passage countries matrix. Contains the figure of pels that are expected to alter from each land cover type to each other land screen type over the specified figure of clip units.
A set of conditional chance images. Reports the chance that each land screen type would be found at each pel after the specified figure of clip units.
Restrictions to Markov
Markov analysis does non account the causes of land usage alteration.
It ignores the forces and procedures that produced the ascertained forms.
It assumes that the forces that produced the alterations will go on to make so in the hereafter.
An even more serious job of Markov analysis is that it is insensitive to infinite: it provides no sense of geographics.
Although the passage chances may be accurate for a peculiar category as a whole, there is no spacial component to the mold procedure.
Using cellular zombis adds a spacial dimension to the theoretical account.
A simple illustration
The lattice is 1-dimensional row of 20 cells.
Each row represents a individual clip measure of the zombi ‘s development.
Each cell ‘s development is affected by its ain province and the province of its immediate neighbours to the left and right.
Cells with an uneven figure of black neighbours ( numbering themselves ) will be black at the following clip measure.
Otherwise, they are white.
Figure: A more complicated illustration: John Conway ‘s Game of Life
Two cell provinces: black and white.
Each cell is affected by the province of its 8 neighbours in the grid.
A white cell becomes black if it has 3 black neighbours.
A black cell stays black if it has 2 or 3 black neighbours.
Cellular automata-MCA in Idrisi
It combines cellular zombis and the Markov alteration land screen anticipation. It besides adds spacial adjacency every bit good as cognition of the likely spacial distribution of passages to Markov alteration analysis ( Clarke, 1997 ) . The CA procedure creates a “ spatially-explicit weighting factor which is applied to each of the suitablenesss, weighing more to a great extent countries that are in propinquity to bing land utilizations and guaranting that land usage alteration occurs in propinquity to bing like land usage categories, and non in a entirely random mode ” ( Eastman, 2003 ) .
In each loop of the simulation each category will usually derive land from one or more of the other categories or it may lose some to one or more of the other categories. Claimant categories take land from the host based on the suitableness map for the claimant category.
CA_MARKOV uses the passages country file from MCA and a land usage suitableness file and a 5 Ten 5 cell adjacency filter to “ turn ” land usage from clip two to some specified ulterior clip period.
By filtrating a Boolean mask of the category being considered, the average filter = 1 when it is wholly within the bing category and 0 when it is wholly outside it.
When it crosses a boundary, the filter produces values that rapidly passage from 1 to 0. This consequence is multiplied by the suitableness image for that category, increasingly downweighting the suitablenesss with distance from bing cases of that category.
At each loop, new category masks are created that reflect the altering geographics of each category.
In Geomod, we can imitate the passage from one class to another. This theoretical account merely needs one clip minute to do an appraisal based on expected clip measures, a suitableness map and, optionally, a adjacency regulation ( Pontius, 2001 ) . One can stipulate predicted measures randomly but, for the interest of comparing, we will utilize in our survey the same predicted measures obtained with Markov passages for the three theoretical accounts. Geomod assumes continuity of the phenomenon to be modelled, i.e. , if a pel was classified as urban in 1986 it will stay urban in 2000. This state of affairs is different from Markov theoretical accounts and world but, it is acceptable for this phenomenon as it really unlikely that an urban country becomes non urban for the clip period in analysis. The suitableness map and the optional
adjacency regulation work the same manner as in Markov theoretical accounts.
Geomod, unlike the two other tested theoretical accounts in this survey, enables the coevals of suitableness maps utilizing driving factors and bing urban countries in 1986. Information of the urban countries of 1986 ensured that a pel that was urban in 1986 remained urban in 2000 during urban growing procedure. Both constrained ( utilizing different sizes ) and unconstrained options were tested. In the forced manner, a user-defined filter assigns a larger suitableness to the non-built pels that are near bing built pels. In the unconstrained manner, choice of pels that will alter to urban is based unambiguously on the suitableness map.