# The dynamics of land use/land cover change modeling and their implication for the flood damage assessment in the Tondano watershed, North Sulawesi, Indonesia

## Abstract

The Markov Chain and Cellular Automata (Markov-CA) approach have been applied to create the dynamics of land use/land cover (LULC) change modeling in the Tondano watershed, North Sulawesi, Indonesia. The multi temporal of remotely sensed data, Landsat 5 TM in 1997, Landsat 7 TM in 2002 and Landsat 8 LDCM in 2015 were used to produce the LULC maps. Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Map (GDEM) data were used as input for the flood modeling created by the Monte Carlo algorithm. The LULC maps in 1997 and 2002 were used to create predictions and modeling LULC map with the Markov-CA approach in the next few years (for the year 2015, 2025, 2035 and 2050). Meanwhile, the LULC map in 2015 with an accuracy of 80.11 % based on the calculation of the Kappa index has been used as a reference map to determine the accuracy of the Markov-CA approach to produce a model of the LULC map in 2015. The result of the accuracy by using cross-correlation matrix between the LULC model in 2015 with the LULC reference in 2015 is 75.88 %. The dynamics of LULC changes showed that area-class forest, dry land, paddy fields and shrubbery would be expected to experience an area decreases in the extent from the year 2015 to 2050, with the rate of change in average: 10.52, 13.22, 14.49 and 1.15 ha/year, respectively. Meanwhile, the area-class bare soil, plantation, settlement and water body would be expected to experience an area increases, with the rate of change in average: 6.79, 11.14, 11.49 and 9.7 ha/year, respectively. Furthermore, flood damage assessment can be calculated by estimating LULC area affected by the flood, which is determined based on the overlay between LULC maps from the result of Markov-CA with flood maps from the result of Monte Carlo algorithm. Under current conditions, estimated flood damage exposure to extreme flood events with return periods of 100 years for the water level scenario Hc = 3 m and Hc = 5 m is more than €520 and €958 million, respectively.

## Keywords

Remote sensing Flood damage assessment Tondano watershed North Sulawesi Indonesia## Introduction

Land use/land cover (LULC) can be defined as part of the physical composition and characteristics of the land elements on the earth’s surface. LULC systems have complex dynamics, which consists of natural, social and economic spatially. The availability and distribution of LULC change have significant impacts on climate, environmental issue and natural ecosystem conditions (Cihlar 2000; Wang et al. 2012; Yan et al. 2015). The management and good planning to LULC can provide an important role in the moderation of extreme events, climate change mitigation and land degradation. For example, its role can mitigate flood events, landslide and erosion trough water retention in soils and water uptake by vegetation (Stürck et al. 2015). The LULC changes can be caused by several factors, such as human activities, e.g., urbanization, deforestation, agriculture intensification and others. In addition, natural factors also contributed to the change of LULC (Lambin 1997; Halmy et al. 2015).

The dynamics of LULC change can be monitored by using remotely sensed data. The availability of remotely sensed data have been provided a variety of temporal, spectral and spatial resolution, which can be used to detect changes on the earth’s surface (Rogan and Chen 2004; Wu et al. 2006; Halmy et al. 2015). Furthermore, the utilizations of remotely sensed data were used as input in generating a LULC map. The monitoring of the LULC change dynamics can be done by comparing a LULC map at currently with the conditions in previous years as a multi-temporal (Behera et al. 2012). The use of remotely sensed data for mapping of the dynamics of LULC changes has been used extensively. Example include the use of Landsat MSS, Landsat TM, Landsat ETM+, Indian Remote sensing Satellite LISS-III (IRS), MODIS, NOAA/AVHRR, CBERS, among others (e.g., Sun et al. 2007; Gong et al. 2015; Wehmann and Liu 2015; Shooshtari and Gholamalifard 2015; Nejadi et al. 2012; Pouliot et al. 2014; Chen et al. 2013; Yang et al. 2011).

To determine the condition of the LULC in the next several years can be done with modeling of LULC predictions. The modeling can be done either by the model of Markov-CA (Guan et al. 2011; Yang et al. 2014; Gong et al. 2015). The use of Markov-CA is an interesting approach and it has advantages in the modeling of LULC change, spatially and temporally (Sylvertown et al. 1992; Wang and Zhang 2001; Behera et al. 2012). One of the implications of the use of monitoring the dynamics of LULC change and predictions can be used for the flood damage assessment (e.g., Marfai and King 2007; Ward et al. 2010; 2011; Joling 2013; Yulianto et al. 2015a). The flood damage assessment is required to estimate the impact of LULC in the next several years from the occurrence of floods. The existence of such information can then be used as input for flood risk mapping in a region (Su et al. 2005; Metzler 2011). Several other studies related to the implications of LULC for the flood damage assessment (e.g., Jonge et al. 1996; Herath 2003; Yen and Anh 2010; Jongman et al. 2012; Beckers et al. 2013).

The availability of information related to the impact of flood events in the research area is important to be done with the aim to quickly estimate of losses caused by flooding. One of the immediate measures that can be used to calculate the impact of flooding on LULC. In addition, the calculation of the impact of flooding from LULC change needs to be done in order to determine the potential loss to the effects of flooding in the future. That information can be used as input to government and several stakeholders to anticipate and minimize the impact of flood risk and it can also be used as a reference in environmental management policies in the research area. The objectives of this research are: (a) to analyze, modeling and predictions the dynamics of LULC change from multi-temporal remotely sensed data, (b) to implement the results of modeling the dynamics of LULC change and predictions for the flood assessment in the research area.

## Study area

The number of the population living at the Tondano watershed in the year 2014

Districts | Number of population in 2014 |
---|---|

Manado | |

Mapanget | 51,259 |

Paal Dua | 42,798 |

Singkil | 47,963 |

Tikala | 29,136 |

Tuminting | 51,039 |

Wanea | 22,768 |

Wenang | 33,048 |

Minahasa | |

Langowan Timur | 12,475 |

Langowan Barat | 15,308 |

Langowan Selatan | 7539 |

Langowan Utara | 8158 |

Tompaso | 16,117 |

Kawangkoan | 26,578 |

Sonder | 18,064 |

Tombulu | 16,196 |

Tondano Barat | 19,513 |

Tondano Selatan | 21,815 |

Remboken | 11,175 |

Kakas | 21,473 |

Eris | 9819 |

Kombi | 9874 |

Tondano Timur | 14,130 |

Tondano Utara | 12,525 |

Minahasa Utara | |

Kauditan | 24,263 |

Airmadidi | 28,153 |

Kalawat | 29,743 |

Dimembe | 23,568 |

Talawaan | 19,932 |

Tomohon | |

Tomohon Selatan | 21,682 |

Tomohon Tengah | 21,329 |

Tomohon Timur | 10,453 |

Total | 697,893 |

## Method

### Data availability

The availability of satellite image data were used in this research

Date type | Spatial resolution (m) | Path and row | Acquisition date | Source |
---|---|---|---|---|

Landsat 5 TM | 30 | 112, 059 | June 04, 1997 | USGS |

Landsat 7 TM | 30 | 112, 059 | May 25, 2002 | USGS |

Landsat 7 TM | 30 | 112, 059 | July 28, 2002 | USGS |

Landsat 8 LDCM | 30 | 112, 059 | January 29, 2015 | LAPAN |

Landsat 8 LDCM | 30 | 112, 059 | April 03, 2015 | LAPAN |

Landsat 8 LDCM | 30 | 112, 059 | May 21, 2015 | LAPAN |

ASTER GDEM | 30 | N01, E124 | October 17, 2011 | Japan space systems |

### Data processing and analysis

#### Pre-processing satellite images

The pre-processing stages in this research were conducted by converting the value of Digital Number (DNs) into the reflectance values of Top of Atmosphere. In the standard product Landsat 5 TM and Landsat 7 TM imagery have DNs format 8-bit unsigned integer. Meanwhile, the standard product of Landsat 8 LDCM imagery has DNs format 16-bit unsigned integer. The pre-processing stage is done with the purpose to obtain the standard reflectance values, because there are differences in the DNs format of the image data. Thus, the results of the pre-processing data can be used for processing and analysis in the later stages. Landsat data processing in this research was done using the Semi-Automatic Classification Plugin contained on Open Source software Quantum GIS (QGIS) 2.12.2 Lyon and it was released on December 18, 2015 (http://www.qgis.org/en/site/forusers/download.html). According to Luca et al. (2013), this plugin is a free plugin for QGIS and it can be used for the semi-automatic supervised classification designed to expedite the processing of multispectral or hyperspectral remote sensing images, which provides a set of tools for pre-processing and post-processing.

#### The classification process for Land use/land cover mapping

At this stage, the multi-temporal of Landsat imagery were used as input for processing and classification of LULC mapping. The processing is done by the individual training areas. The maximum likelihood (ML) was used as a classification algorithm on the semi-automatic classification plugin, which is part of the supervised classifier categories.

*i*= 1,…, M and M is the total number of classes. \(x\) is the vector pixel in

*n*-dimensional, where

*n*is the number of bands. \({\text{p}}(\omega_{i} )\) is the probability that the correct class, in \(\omega_{i}\) for a pixel at position

*x*. \(|\sum_{i} |\) is the determinant of the covariance matrix of the data in class \(\omega_{i}\). \(\mathop \sum \nolimits i\) is the inverse of the covariance matrix. \(m_{i}\) is the mean vector.

#### Modeling and prediction of the dynamic land use/land cover change

In this research, modeling and prediction of the dynamic LULC change has been carried out by using a Markov-CA. LULC maps in 1997 and 2002, which is the results of multi-temporal classification of Landsat imagery, were used as input data for the modeling and prediction LULC in the next years. Meanwhile, LULC map in 2015 that is the classification result of Landsat imagery in 2015 was used as a reference map to describe the condition LULC today. The reference of LULC map was used to perform the calculation of the accuracy assessment LULC in 2015 with Markov-CA approach.

Markov-CA approach is a combination model of Markov Chain and Cellular Automata approach, which can be used to predict and simulate LULC change in several next year (Behera et al. 2012; Yang et al. 2014). The Markov Chain model is a stochastic process model that can describe the probability of a change from one object to another. The model is one of the recommended methods for modeling LULC based on the use of the time evolutionary trend. It means that the time (t − 1) to time (t) can be used as input probability LULC change, which is further used for LULC prediction in the next year (t + 1) (Thomas and Laurence 2006; Behera et al. 2012). Cellular Automata is a part of the essential elements of geo-spatial, which focus on the variations in the dynamics of change and it is able to simulate the characteristics of the spatial–temporal complex system and it cannot be represented by a specific equation (Mousivand et al. 2007; Arsanjani et al. 2013; Yang et al. 2014). In this research, the modeling LULC have been processed by the tools of Markov-CA module contained in the IDRISI Andes software, which is developed by Clark Labs at Clark University.

#### Implication for the flood damage assessment

Their implication for flood damage assessment in the research area has been carried out by calculating the impact of flooding inundation of the LULC, which can scripted and predictable for the next few years. In this research, flood inundation scenarios have been created by the Monte Carlo algorithm approach. Flood inundation modeling with the Monte Carlo algorithm approach does not require complex parameters and does not require long processing times, as it has been applied to some previous studies (e.g., Smemoe et al. 2007; Kalyanapu 2011; Yulianto et al. 2015b). The use of Monte Carlo algorithms previously were conducted by Felpeto et al. (2007) and Felpeto (2009) to simulate an inundation on the lava flows. The algorithm model is a probabilistic model, with the assumption that the topographic conditions have a dominant factor for determining the material flows.

The parameter data were used for simulation of the probability of flood inundation using a Monte Carlo algorithm in the upper boundary Tondano river

Return periods (years) | Volume (m | Maximum flow length (m) | Start iteration position (X; Y) | Height correction for water level (Hc) | The number of iterations | |
---|---|---|---|---|---|---|

3 m | 5 m | |||||

2 | 91,411,390 | 7435 | 711593; 163768 | 3 m, 5 m | 33,800 | 20,300 |

5 | 107,900,660 | 7435 | 711593; 163768 | 3 m, 5 m | 40,000 | 24,000 |

10 | 120,547,770 | 7435 | 711593; 163768 | 3 m, 5 m | 44,800 | 26,800 |

20 | 133,995,330 | 7435 | 711593; 163768 | 3 m, 5 m | 49,600 | 33,700 |

50 | 153,526,310 | 7435 | 711593; 163768 | 3 m, 5 m | 57,000 | 38,200 |

100 | 169,695,400 | 7435 | 711593; 163768 | 3 m, 5 m | 62,900 | 42,500 |

The flood damage assessment can be created using overlaying the maps of flood inundation model under scenario return periods: 2, 5, 10, 20, 50 and 100 years with the maps of LULC modeling under scenario for the 2015, 2025, 2035 and 2050 years. Based on the results of the overlay can be estimated LULC affected by floods for several next year. The value of assets has been calculated with reference to the research conducted by Marfai and King (2007) and Ward et al. (2011) and we have the same assumptions for asset values in the research area with the Semarang and Jakarta, Indonesia.

The value of assets for settlement was estimated to be ca. €1.2 million per hectare, for agriculture (e.g., paddy field, dry land and plantation) was estimated to be ca. €80,000 per hectare, and also for open area (e.g., bare soil and shrubbery) was estimated to be ca. €1700 per hectare. The absolute value of exposure damage estimated in this research are considered as an indication. Exposure damage was estimated based on the actual market value per hectare in each class LULC. However, these estimates give a good impression of exposure damage on LULC changes due to flooding.

## Results

### Land use/land cover mapping in the research area

The estimate area of LULC changes for the year 1997–2015 in the research area

LULC Class | Year | ||
---|---|---|---|

1997 Area (ha) | 2002 Area (ha) | 2015 Area (ha) | |

Forest | 16,498.44 (29.20) | 15,617.37 (27.64) | 14,305.65 (25.32) |

Dry land | 9390.15 (16.62) | 10,181.44 (18.02) | 10,596.88 (18.76) |

Bare soil | 587.61 (1.04) | 810.01 (1.43) | 946.36 (1.67) |

Plantation | 12,487.05 (22.10) | 13,309.27 (23.56) | 13,591.90 (24.06) |

Settlement | 2811.42 (4.98) | 3270.47 (5.79) | 3957.39 (7.00) |

Paddy fields | 6660.54 (11.79) | 5095.80 (9.02) | 4714.21 (8.34) |

Shrubbery | 3651.84 (6.46) | 3721.96 (6.59) | 3836.89 (6.69) |

Water body | 4413.96 (7.81) | 4494.69 (7.96) | 4551.74 (8.06) |

Total | 56,501.01 (100) | 56,501.01 (100) | 56,501.01 (100) |

The estimate of the average rate area of LULC changes for the year 1997–2015 in the research area

LULC class | Change area (ha) | Average area (ha/year) | |
---|---|---|---|

1997–2002 | 2002–2015 | 1991–2015 | |

Forest | −881.07 | −1311.73 | −121.82 |

Dry land | +791.29 | +415.44 | +67.04 |

Bare soil | +222.40 | +136.35 | +19.93 |

Plantation | +822.22 | +282.63 | +61.38 |

Settlement | +459.05 | +686.93 | +63.67 |

Paddy fields | −1564.74 | −381.60 | −108.13 |

Shrubbery | +70.12 | +114.93 | +10.28 |

Water body | +80.73 | +57.05 | +7.65 |

### Modeling and prediction of the dynamic land use and land cover change

The accuracy of the Kappa index for the LULC map in 2015, which is the result of the maximum likelihood supervised classification and corrected by using high-resolution imagery from Google Earth

LULC Classes | Reference | Total | |||||||
---|---|---|---|---|---|---|---|---|---|

Forest | Dry land | Bare soil | Plantation | Settlement | Paddy fields | Shrubbery | Water body | ||

Forest | 151 | 3 | – | 16 | – | – | – | – | 170 |

Dry land | – | 70 | – | 2 | – | 11 | 1 | – | 84 |

Bare soil | 1 | 0 | 6 | 1 | 2 | 1 | 2 | 1 | 14 |

Plantation | 3 | 1 | 1 | 55 | 1 | 4 | 1 | – | 66 |

Settlement | – | – | – | 1 | 49 | – | – | – | 50 |

Paddy fields | – | 8 | 2 | 7 | 1 | 62 | 4 | 1 | 85 |

Shrubbery | 4 | 1 | 3 | 2 | – | 4 | 26 | – | 40 |

Water body | – | – | – | – | – | – | – | 39 | 39 |

Total | 159 | 83 | 12 | 84 | 53 | 82 | 34 | 41 | 548 |

The estimate area of LULC changes for the year 2015–2050 in the research area, which is the result of Markov-CA approach

LULC class | Year | |||
---|---|---|---|---|

2015 Area (ha) | 2025 Area (ha) | 2035 Area (ha) | 2050 Area (ha) | |

Forest | 14150.82 (25.05) | 14012.89 (24.80) | 13884.33 (24.57) | 13782.46 (24.39) |

Dry land | 9567.41 (16.93) | 9527.44 (16.86) | 9432.80 (16.69) | 9104.78 (16.11) |

Bare soil | 763.82 (1.35) | 787.82 (1.39) | 930.16 (1.65) | 1025.41 (1.81) |

Plantation | 11461.22 (20.28) | 11561.19 (20.46) | 11667.34 (20.65) | 11851.05 (20.97) |

Settlement | 5476.19 (9.69) | 5546.91 (9.82) | 5558.26 (9.84) | 5878.28 (10.40) |

Paddy fields | 7138.69 (12.63) | 6968.71 (12.33) | 6842.73 (12.11) | 6631.38 (11.74) |

Shrubbery | 3643.86 (6.45) | 3615.60 (6.40) | 3603.49 (6.38) | 3603.28 (6.38) |

Water body | 4275.00 (7.57) | 4504.47 (7.97) | 4582.12 (8.11) | 4624.16 (8.18) |

Total | 56,501.01 (100) | 56,501.01 (100) | 56,501.01 (100) | 56,501.01 (100) |

The estimate of the average rate area of LULC changes for the year 2015–2050 in the research area, which is the result of Markov-CA approach

LULC class | Change area (ha) | Average area (ha/year) | ||
---|---|---|---|---|

2015–2025 | 2025–2035 | 2035–2050 | 2015–2050 | |

Forest | −137.93 | −128.56 | −101.86 | −10.52 |

Dry land | −39.97 | −94.64 | −328.02 | −13.22 |

Bare soil | +24.00 | +166.34 | +95.25 | +6.79 |

Plantation | +99.96 | +106.15 | +183.71 | +11.14 |

Settlement | +70.71 | +11.35 | +320.02 | +11.49 |

Paddy fields | −169.98 | −125.98 | −211.34 | −14.49 |

Shrubbery | −28.26 | −12.32 | −0.21 | −1.15 |

Water body | +229.47 | +77.65 | +42.04 | +9.97 |

### Implication for flood damage analysis

The damage exposure estimation of the flood scenario return period 2 years and LULC model prediction in 2015, 2025, 2035 and 2050 (value in 1000 EUR)

LULC | Return period 2 years, Hc = 3 m | Return period 2 years, Hc = 5 m | ||||||
---|---|---|---|---|---|---|---|---|

2015 | 2025 | 2035 | 2050 | 2015 | 2025 | 2035 | 2050 | |

Dry land | 232 | 224 | 264 | 192 | 640 | 600 | 808 | 400 |

Bare soil | 61 | 54 | 65 | 37 | 137 | 120 | 138 | 98 |

Plantation | 2992 | 3472 | 3464 | 2784 | 6648 | 7528 | 7680 | 5896 |

Settlement | 412,680 | 414,600 | 417,960 | 449,760 | 883,320 | 886,080 | 891,360 | 956,880 |

Paddy fields | 8784 | 8520 | 7656 | 7720 | 24,240 | 23,896 | 22,000 | 22,080 |

Shrubbery | 31 | 33 | 34 | 30 | 106 | 109 | 117 | 110 |

Total | 424,780 | 426,903 | 429,443 | 460,523 | 915,091 | 918,334 | 922,103 | 985,464 |

The damage exposure estimation of the flood scenario return period 5 years and LULC model prediction in 2015, 2025, 2035 and 2050 (value in 1000 EUR)

LULC | Return period 5 years, Hc = 3 m | Return period 5 years, Hc = 5 m | ||||||
---|---|---|---|---|---|---|---|---|

2015 | 2025 | 2035 | 2050 | 2015 | 2025 | 2035 | 2050 | |

Dry land | 240 | 224 | 256 | 200 | 768 | 696 | 952 | 440 |

Bare soil | 58 | 51 | 62 | 35 | 139 | 118 | 139 | 114 |

Plantation | 3040 | 3520 | 3528 | 2816 | 6904 | 7936 | 8112 | 6136 |

Settlement | 426,360 | 427,680 | 431,760 | 462,960 | 889,560 | 891,000 | 893,400 | 955,440 |

Paddy fields | 8704 | 8432 | 7632 | 7640 | 25,384 | 25,096 | 23,280 | 23,312 |

Shrubbery | 26 | 29 | 28 | 26 | 112 | 117 | 125 | 116 |

Total | 438,428 | 439,936 | 443,266 | 473,677 | 922,867 | 924,963 | 926,008 | 985,558 |

The damage exposure estimation of the flood scenario return period 10 years and LULC model prediction in 2015, 2025, 2035 and 2050 (value in 1000 EUR)

LULC | Return period 10 years, Hc = 3 m | Return period 10 years, Hc = 5 m | ||||||
---|---|---|---|---|---|---|---|---|

2015 | 2025 | 2035 | 2050 | 2015 | 2025 | 2035 | 2050 | |

Dry land | 296 | 264 | 320 | 200 | 648 | 624 | 856 | 344 |

Bare soil | 70 | 62 | 73 | 43 | 133 | 116 | 134 | 103 |

Plantation | 3112 | 3616 | 3584 | 2912 | 6400 | 7328 | 7544 | 5688 |

Settlement | 485,880 | 488,520 | 492,360 | 525,000 | 892,920 | 893,400 | 897,120 | 960,720 |

Paddy fields | 10,008 | 9688 | 8816 | 8936 | 24,672 | 24,392 | 22,536 | 22,624 |

Shrubbery | 32 | 34 | 34 | 32 | 100 | 104 | 113 | 104 |

Total | 499,398 | 502,184 | 505,187 | 537,123 | 924,873 | 925,964 | 928,303 | 989,583 |

The damage exposure estimation of the flood scenario return period 20 years and LULC model prediction in 2015, 2025, 2035 and 2050 (value in 1,000 EUR)

LULC | Return period 20 years, Hc = 3 m | Return period 20 years, Hc = 5 m | ||||||
---|---|---|---|---|---|---|---|---|

2015 | 2025 | 2035 | 2050 | 2015 | 2025 | 2035 | 2050 | |

Dry land | 296 | 264 | 336 | 208 | 704 | 648 | 880 | 416 |

Bare soil | 71 | 62 | 74 | 45 | 146 | 128 | 146 | 111 |

Plantation | 3216 | 3712 | 3704 | 3000 | 6480 | 7400 | 7640 | 5712 |

Settlement | 496,320 | 498,120 | 501,600 | 538,680 | 914,400 | 916,560 | 921,480 | 987,240 |

Paddy fields | 9616 | 9408 | 8456 | 8336 | 25,424 | 25,024 | 23,080 | 23,192 |

Shrubbery | 33 | 35 | 36 | 32 | 106 | 112 | 120 | 111 |

Total | 509,551 | 511,601 | 514,206 | 550,301 | 947,260 | 949,872 | 953,345 | 1,016,783 |

The damage exposure estimation of the flood scenario return period 50 years and LULC model prediction in 2015, 2025, 2035 and 2050 (value in 1000 EUR)

LULC | Return period 50 years, Hc = 3 m | Return period 50 years, Hc = 5 m | ||||||
---|---|---|---|---|---|---|---|---|

2015 | 2025 | 2035 | 2050 | 2015 | 2025 | 2035 | 2050 | |

Dry land | 328 | 304 | 360 | 208 | 712 | 696 | 928 | 416 |

Bare soil | 69 | 61 | 73 | 43 | 138 | 120 | 140 | 110 |

Plantation | 2992 | 3480 | 3480 | 2792 | 7024 | 8032 | 8176 | 6200 |

Settlement | 498,600 | 499,560 | 504,120 | 538,680 | 923,520 | 925,680 | 927,720 | 992,760 |

Paddy fields | 9880 | 9680 | 8728 | 8768 | 26,368 | 26,016 | 24,080 | 24,272 |

Shrubbery | 28 | 30 | 30 | 27 | 116 | 118 | 130 | 120 |

Total | 511,897 | 513,115 | 516,790 | 550,518 | 957,878 | 960,662 | 961,174 | 1,023,878 |

The damage exposure estimation of the flood scenario return period 100 years and LULC model prediction in 2015, 2025, 2035 and 2050 (value in 1000 EUR)

LULC | Return period 100 years, Hc = 3 m | Return period 100 years, Hc = 5 m | ||||||
---|---|---|---|---|---|---|---|---|

2015 | 2025 | 2035 | 2050 | 2015 | 2025 | 2035 | 2050 | |

Dry land | 296 | 288 | 376 | 208 | 720 | 672 | 912 | 424 |

Bare soil | 70 | 61 | 71 | 45 | 147 | 129 | 147 | 111 |

Plantation | 3144 | 3608 | 3592 | 2896 | 6880 | 7872 | 7984 | 6032 |

Settlement | 506,040 | 506,880 | 512,760 | 547,080 | 923,880 | 925,920 | 929,280 | 997,440 |

Paddy fields | 10,472 | 10,320 | 9240 | 9184 | 26,512 | 26,160 | 24,192 | 24,416 |

Shrubbery | 33 | 35 | 37 | 32 | 117 | 120 | 132 | 121 |

Total | 520,054 | 521,191 | 526,075 | 559,446 | 958,256 | 960,873 | 962,647 | 1,028,544 |

## Discussion

This research has produced information related to the dynamics of LULC change modeling and their implication for the flood damage assessment. In this section, we will discuss related limitations, uncertainty and future direction for the use of methods that have been applied in the research area.

The results of LULC information that has been generated in this research, there are limitations. These limitations are related to the use of Markov-CA approach which does not include the parameters and factors that could affect or inhibit the occurrence of the LULC change in the research area. Thus, the resulting of LULC map updates in the next few years are assumed to be linear and the results are a scenario with a normally indication, without any triggering factor and inhibit the occurrence of the LULC change. This can be shown on difference in the average area change between the results of LULC classification in 1997–2015 with the results of LULC modeling in 2015–2050. Where, the average area change of the LULC classification in 1997–2015 is tending to relatively large. Meanwhile, the average area change of the LULC modeling in 2015–2050 is tending to relatively small. As an example for class of forest and paddy fields, the LULC classification in 1997–2015 had a rate of change: 121.82 and 108.13 ha/year, respectively. Meanwhile, the LULC modeling in 2015–2050 had a rate of change: 10.52, 14.49 ha/year, respectively. It can be used as input for future research in the research area to add the factors and parameters that can trigger factor and inhibit the occurrence of the LULC change. Thus, the information on the prediction and modeling LULC maps that will be able to provide good accuracy and precision (note: the result of the accuracy by using cross-correlation matrix between the LULC models in 2015 with the LULC reference in 2015 is 75.88 %), as has been done by several previous studies (e.g., Guan et al. 2011; Halmy et al. 2015)

The limitations of the flood inundation models were used in this research, namely: (a) modeling was performed on the segment of the Tondano river with a range point of downstream boundary to the point of the upper boundary Tondano river, (b) the Tikala river has been modeled as lateral inflow into Tondano river, (c) the existence of drainage, tributary and bridges in the segment of the Tondano river were ignored. These limitations have the same with the research created by Nanlohy et al. (2008). Map of flood inundation models that have been created by using the Monte Carlo approach presents a static and not dynamic modeling. In addition, the limitations of the resulting model can only present the distribution of flood inundation that is shown with a probability value at each scenario in the running. Where, at a value of 0 indicates a low probability and value of 1 indicates a high probability. Thus, the depth of flood information is not available in this model. However, the modeling can provide relatively quick benefits of generating flood inundation models. It is suitable for rapid mapping required during emergency response. The results of the flood inundation modeling have not been validated in this research. Thus, the field of measurement and observation for model validation is required and recommended in the future research.

Implementation of the uses of LULC and flood inundation modeling was estimated flood damage assessment in the research area. The process is done in a simple, namely: overlay between the flood inundation model with the LULC model that has been calculated for the next few years. Thus, the information of LULC affected by flooding can be determined in the research area. Estimates of flood damage exposure for the next few years has been calculated based on the asset value per hectare under current conditions in each LULC class. The results obtained by the flood scenario in extreme conditions for the return period of 100 year, the total of flood damage exposure can be estimated more than €520 million for the water level scenario Hc = 3 m and €958 million for the water level scenario Hc = 5 m. The market value of each class of LULC has not been done in detail. Thus, the limitations can be shown by the incorporation of market value in agricultural class (e.g., paddy fields, plantations and dry land) and the open area class (e.g., bare soil and shrubbery). It is necessary and recommended in the future research to focus in calculating the asset value for each LULC class in detail.

## Conclusion

Markov-CA and Monte Carlo algorithm approaches have been successfully used in this research to create LULC and flood modelling, which can then be implemented to estimate the flood damage assessment for the next few years. The utilization of multi-temporal remotely sensed Landsat and ASTER GDEM have supported the acquisition of input data for modeling created by Markov-CA and Monte Carlo algorithm approaches. The results of LULC and flood modeling in this research could be used to support the mapping at a scale of 1:25,000–1:50,000. In the future research, it is advisable and recommended to use multi-temporal remotely sensed data with high resolution, such as Ikonos, Quickbird, Worldview, Pleiades and others. Thus, the need modeling for detailed mapping or less than scale 1:10,000 can be fulfilled. In addition, the use of stereo imagery from SPOT 6 and 7 with a spatial resolution of 6 m can be used for detailing topography information or DEM in the research area. The requirement of DEM data with a spatial resolution of less than 10 m is needed in flood modeling. Thus, the accuracy of the results of flood modeling can be improved and increased. The results of LULC modeling and their implication not only used for estimating the flood damage assessment, but also the results can be used to predict the flow discharge of flooding due to LULC changes for the next few years. Thus, in the future research can be suggested and recommended to predict the flow discharge of flooding due to LULC changes, and the results can be used as one of the stages in the flood mitigation effort in the research area.

## Notes

### Acknowledgments

This paper is a part of the research activities entitled “The utilization of remotely sensed data to support analysis of floods in Indonesia”. This research was funded by the budget of DIPA LAPAN activities in 2015, Remote Sensing Application Center, Indonesian National Institute of Aeronautics and Space (LAPAN). Thanks go to Dr. M. Rokhis Khomarudin (Director of Remote Sensing Application Center, LAPAN) who has supported the implementation of this research. Drs. Taufik Maulana, MBA., Dr. Wikanti Asriningrum and colleagues at the Remote Sensing Application Center, LAPAN for support, discuss and give suggestions in this research. Landsat 5 MSS and Landsat 7 TM images were provided by the U.S. Geological Survey (USGS). Landsat 8 LDCM images were provided by LAPAN. ASTER GDEM data were provided by the Japan Space Systems (Japan-US ASTER Science Team).

## References

- Arsanjani JJ, Helbich M, Kainz W, Boloorani AD (2013) Integration of logistic regression, Markov chain and cellular automata models to simulate urban expansion. Int J Appl Earth Obs Geoinf 21:265–275CrossRefGoogle Scholar
- Beckers A, Dewals B, Erpicum S, Dujardin S, Detrembleur S, Teller J, Pirotton M, Archambeau P (2013) Contribution of land use changes to future flood damage along the river Meuse in the Walloon region. Nat Hazards Earth Syst Sci 13:2301–2318CrossRefGoogle Scholar
- Behera MD, Borate SN, Panda SN, Behera PR, Roy PS (2012) Modelling and analyzing the watershed dynamics using Cellular Automata (CA)-Markov model—a geo-information based approach. J Earth Syst Sci 121:1011–1024CrossRefGoogle Scholar
- Chavez PS Jr (1988) An improved dark-object subtraction technique for atmospheric scattering correction of multispectral data. Remote Sens Environ 24:459–479CrossRefGoogle Scholar
- Chen X, Yu SX, Zhang YP (2013) Evaluation of spatio temporal dynamics of simulated land use/cover in China using a probabilistic Cellular Automata-Markov model. Pedosphere 23(2):243–255CrossRefGoogle Scholar
- Cihlar J (2000) Land cover mapping of large areas from satellites: status and research priorities. Int J Remote Sens 21:1093–1114CrossRefGoogle Scholar
- Dayantolis W, Fitri HT (2014) Overview climatological flooding on January 15, 2014 in Manado. Category Archives: Manado. https://bencanasulut.wordpress.com/category/manado. Accessed 08 Jan 2015
**(in Indonesian)** - Felpeto A (2009) VORIS a GIS-based tool for volcanic hazard assessment. User’s Guide Version 2.0.1Google Scholar
- Felpeto A, Matri J, Ortiz R (2007) Automatic GIS-based system for volcanic hazard assessment. J Volcanol Geotherm Res 166:106–116CrossRefGoogle Scholar
- Gong W, Yuan L, Fanc W, Stott P (2015) Analysis and simulation of land use spatial pattern in Harbin prefecture based on trajectories and cellular automata—Markov modelling. Int J Appl Earth Obs Geoinf 34:207–216CrossRefGoogle Scholar
- Guan D, Li HF, Inohae T, Su W, Nagaie T, Hakao K (2011) Modeling urban land use change by the integration of cellular automaton and Markov model. Ecol Model 222:3761–3772CrossRefGoogle Scholar
- Hall R, Wilson MEJ (2000) Neogene sutures in eastern Indonesia. J Asian Earth Sci 18(6):781–808 CrossRefGoogle Scholar
- Halmy MWA, Gessler PE, Hicke JA, Salem BB (2015) Land use/land cover change detection and prediction in the northwestern coastal desert of Egypt using Markov-CA. Appl Geogr 63:101–112CrossRefGoogle Scholar
- Herath (2003) Flood damage estimation of an urban catchment using remote sensing and GIS. International Training Program on Total Disaster Risk ManagementGoogle Scholar
- Huang SL, Wang SH, Budd WW (2009) Sprawl in Taipei’s peri-urban zone: responses to spatial planning and implications for adapting global environmental change. Landsc Urban Plan 90(1–2):20–32CrossRefGoogle Scholar
- Indonesian of Central Agency Statistics (BPS) (2014) Statistics of Minahasa DistrictGoogle Scholar
- Indonesian of Central Agency Statistics (BPS) (2014) Statistics of Minahasa Utara DistrictGoogle Scholar
- Indonesian of Central Agency Statistics (BPS) (2014) Statistics of Tomohon DistrictGoogle Scholar
- Indonesian of Central Agency Statistics (BPS) (2014) Statistics of Manado DistrictGoogle Scholar
- Joling RJ (2013) Adding more detail to potential flood damage assessment: An object based approach. Thesis BSc Aarde and Economie. Vrije Universiteit Amsterdam Faculteit Aard-en LevenswetenschappenGoogle Scholar
- Jonge T, Kok M, Hogeweg M (1996) Modelling floods and damage assessment using GIS. HydroGIS 96: application of geographic information systems in hydrology and water resources management. In: Proceedings of the Vienna Conference. IAHS Publ 235Google Scholar
- Jongman B, Kreibich H, Apel H, Barredo JI, Bates PD, Feyen L, Gericke A, Neal J, Aerts JCJH, Ward PJ (2012) Comparative flood damage model assessment: towards a European approach. Nat Hazards Earth Syst Sci 12:3733–3752CrossRefGoogle Scholar
- Kalyanapu AJ (2011) Monte Carlo based flood risk analysis using a graphics processing unit-enhanced two-dimensional flood model. Dissertation. Department of Civil and Environmental Engineering. The University of UtahGoogle Scholar
- Lambin EF (1997) Modelling and monitoring land-cover change processes in tropical regions. Prog Phys Geogr 21:375–393CrossRefGoogle Scholar
- Luca C, Michele M, Silvia M (2013) Investigating the relationship between land cover and vulnerability to climate change in Dares Salaam. Working Paper, Rome: Sapienza UniversityGoogle Scholar
- Marfai MA, King L (2007) Tidal inundation mapping under enhanced land subsidence in Semarang, Central Java Indonesia. Nat Hazards 44:93–109CrossRefGoogle Scholar
- Metzler SM (2011) Land use interpretation in flood damage estimation. Master’s Theses and Graduate Research, San Jose State University. Paper 4103Google Scholar
- Mousivand AJ, Sarab AA, Shayan S (2007) A new approach of predicting land use and land cover changes by satellite imagery and Markov chain model (case study: Tehran). MSc Thesis. Tarbiat Modares University, Tehran, IranGoogle Scholar
- Nanlohy BJB, Jayadi R, Istiarto (2008) The study of flood control alternatives Tondano river in Manado city. Forum civil engineering XVIII:756–767
**(in Indonesian)**Google Scholar - NASA (2011) Landsat 7 science data users handbook. Maryland: Landsat Project Science Office at NASA’s Goddard Space Flight Center in GreenbeltGoogle Scholar
- Nejadi A, Jafari HR, Makhdoum MF, Mahmoudi M (2012) Modeling plausible impacts of land use change on wildlife habitats, application and validation: lisar protected area, Iran. Int J Environ Res 6(4):883–892Google Scholar
- Pouliot D, Latifovic R, Zabcic N, Guindon L, Olthof I (2014) Development and assessment of a 250 m spatial resolution MODIS annual land cover time series (2000–2011) for the forest region of Canada derived from change-based updating. Remote Sens Environ 140:731–743CrossRefGoogle Scholar
- Richards JA, Jia X (2006) Remote sensing digital image analysis: an introduction. Springer, BerlinGoogle Scholar
- Rogan J, Chen D (2004) Remote sensing technology for mapping and monitoring land-cover and land-use change. Prog Plan 61:301–325CrossRefGoogle Scholar
- Seniarwan Baskoro DPT, Gandasasmita K (2013) Spatial modelling of flood inundation: case study Mangottong river area, Sinjay Regency, South Sulawesi Province. Sci J Globe 15(1):62–67
**(in Indonesian)**Google Scholar - Shooshtari SJ, Gholamalifard M (2015) Scenario-based land cover change modeling and its implications for landscape pattern analysis in the Neka Watershed, Iran. Remote Sens Appl Soc Environ 1:1–19Google Scholar
- Smemoe CM, Nelson EJ, Zundel AK, Miller W (2007) Demonstrating floodplain uncertainty using flood probability maps. J Am Water Resour Assoc 43(2):359–371CrossRefGoogle Scholar
- Sompotan AF (2012) Gological stucture of Sulawesi Struktur Geologi Sulawesi. Earth Sciences Library. Bandung Institute of Technology (ITB)
**(in Indonesian)**Google Scholar - Stürck J, Schulp CJE, Verburg PH (2015) Spatio-temporal dynamics of regulating ecosystem services in Europe e The role of past and future land use change. Appl Geogr 63:121–135CrossRefGoogle Scholar
- Su MD, Kang JL, Chang LF, Chen AS (2005) A grid-based gis approach to regional flood damage assessment. J Mar Sci Technol 13(3):184–192Google Scholar
- Sun H, Forsythe W, Waters N (2007) Modeling urban land use change and urban sprawl: Calgary, Alberta, Canada. Netw Spat Econ 7(4):353–376CrossRefGoogle Scholar
- Sylvertown J, Hotlier S, Johnson J, Dale P (1992) Cellular automaton models of inter specific competition for space the effect of pattern on process. J Ecol 80:527–534CrossRefGoogle Scholar
- The Management Center of Tondano Watershed—Ministry of Forestry (2014) Characteristics of Tondano watershed. Final report
**(in Indonesian)**Google Scholar - Thomas H, Laurence HM (2006) Modeling and projecting land-use and land-cover changes with a cellular automaton in considering landscape trajectories: An improvement for simulation of plausible future states. EARSeLeProc 5:63–76Google Scholar
- USGS (2013) Landsat missions: frequently asked questions about the Landsat missions US Geological Survey. Last modified: 5/30/123. Landsat.usgs.govGoogle Scholar
- USGS (2013) Using the US Geological survey Landsat 8 product. Last modified: 5/30/123. Landsat7.usgs.govGoogle Scholar
- Wang X, Zhang C (2001) A dynamic modelling approach to simulating socioeconomic effects on landscape changes. Ecol Model 140:141–162CrossRefGoogle Scholar
- Wang SQ, Zheng XQ, Zang XB (2012) Accuracy assessments of land use change simulation based on Markov-cellular automata model. Proc Environ Sci 13:1238–1245CrossRefGoogle Scholar
- Ward PJ, Marfai MA, Yulianto F, Hizbaron DR, Aerts JCJH (2010) Coastal inundation and damage exposure estimation: a case study for Jakarta. Nat Hazards 56:899–916CrossRefGoogle Scholar
- Ward PJ, Moel H, Aerts JCJH (2011) How are flood risk estimates affected by the choice of return-periods? Nat Hazards Earth Syst Sci 11:3181–3195CrossRefGoogle Scholar
- Wehmann A, Liu D (2015) A spatial–temporal contextual Markovian kernel method for multi-temporal land cover mapping. ISPRS J Photogramm Remote Sens 107:77–89CrossRefGoogle Scholar
- Wu Q, Li H, Wang R, Paulussen J, He Y, Wang M et al (2006) Monitoring and predicting land use change in Beijing using remote sensing and GIS. Landsc Urban Plan 78:322–333CrossRefGoogle Scholar
- Yan WY, Shaker A, El-Ashmawy N (2015) Urban land cover classification using airborne LiDAR data: a review. Remote Sens Environ 158:295–310CrossRefGoogle Scholar
- Yang H, Du L, Guo H, Zhang J (2011) Tai’an land use analysis and prediction based on RS and Markov model. Proc Environ Sci 10:2625–2630CrossRefGoogle Scholar
- Yang X, Zheng XC, Chen R (2014) A land use change model: integrating landscape pattern indexes and Markov-CA. Ecol Model 283:1–7CrossRefGoogle Scholar
- Yen NTM, Anh TV (2010) Application of remote sensing data for mapping of damage assessment of flood to the land cover, an experiment in Phu Yen province, Vietnam. In: International Symposium on Geoinformatics for Spatial Infrastructure Development in Earth and Allied SciencesGoogle Scholar
- Yulianto F, Sofan P, Zubaidah A, Sukowati KAD, Pasaribu JM, Khomarudin MR (2015a) Detecting areas affected by flood using multi-temporal ALOS PALSAR remotely sensed data in Karawang, West Java, Indonesia. Nat Hazards 77:959–985CrossRefGoogle Scholar
- Yulianto F, Tjahjono B, Anwar S (2015b) The applications of Monte Carlo algorithm and energy con model to produce the probability of block-and-ash flows of the 2010 eruption of Merapi volcano in Central Java, Indonesia. Arab J Geosci 8:4717–4739CrossRefGoogle Scholar
- Zhu J (2010) GIS based urban flood inundation modeling. Second WRI Glob Congr Intell Syst 2:140–143Google Scholar