Land-use and land-cover change (LULCC) is not the only component of global environmental change, but it has major impacts on climate change, ecosystem services, and sustainability. Tropical environments are among the most important locations of global LULCC. Tropical forests in particular are often being reported rapidly diminishing. A common approach to studying tropical deforestation is to consider it as binary process (i.e. forest and non-forest). In fact, LULCC dynamics includes processes of regeneration as well as vegetation cover loss; the net balance is the result of subtraction and addition derived from both trends. A deep understanding about spatiotemporal patterns as well as processes and causes of LULCC is definitely required as an approach to examine this balance. The term of 'spatiotemporal pattern' was emerged to emphasize the LULCC regular structures, in particular repeating structures that change in space and time as a result of interaction between biophysical and human dimensions.

The main objective of this study is to comprehend spatiotemporal patterns and determining the processes and causes of LULCC towards a better management of tropical environment. The specific objectives of this study are (1) to monitor spatiotemporal patterns of LULCC using time-series satellite remote-sensing data; (2) to analyze the spatiotemporal pattern of LULCC with regard to biophysical, local social characteristic, and land-use allocation policy; (3) to predict future land-cover change using scenario based probabilistic cellular automata model; and (4) to estimate the risk of forest and land fires as an important proximate cause of LULCC.

Seven chapters of discussion were included in this study. Chapter 1 described about the general introduction and rationale of this research. Chapter 2 described the general condition of study areas. To provide a comprehensive analysis of LULCC, case studies were conducted in two areas in Indonesia with different social and physical characteristics. The first case study was conducted in the Rawa Danau Nature Reserve which is located in Java Island (Banten Province, Indonesia), and the second case study was conducted in the Middle Mahakam area which is located in Borneo Island (East Kalimantan Province, Indonesia).

The objectives of Chapter 3 were to examine the spatiotemporal pattern of agricultural expansion in Rawa Danau Nature Reserve by analyzing satellite remote-sensing data and to characterize the situation of local people causing that agricultural expansion using information sourced from a socioeconomic household survey. The analysis was performed against the background of the Indonesian economic crisis that occurred in 1997. By remote sensing analysis, it was found that agricultural expansion rapidly increased during the period after the economic crisis compared with the period prior to the crisis. By household survey, it was found that aside from proximity of access, the range of options for land-use practices for local people influenced the characteristics of agricultural expansion in this area.

The objectives of Chapter 4 were to observe LULCC in the tropical lowland of Middle Mahakam area using satellite remote sensing data by applying an improved hybrid land-cover change analysis and to evaluate such changes with regard to existing forest land-use allocation, namely TGHK (Tata Guna Hutan Kesepakatan) prescribed by the Indonesian Government in 1980's. The main result of Chapter 4 was proposal of an improved method of hybrid land-cover change analysis that is more suitable to apply in the tropical environment, which emerged as the originality of this study. By this method, it was found that forest degradation and deforestation occurred beyond the forest regeneration. Land-use allocation policy had greatly influenced the spatiotemporal pattern of LULCC in the study area. Deforestation pattern largely occurred in Non-forest area, whereas forest degradation largely took place in Production forests. The lowest LULCC was found in Protection forests since this area was either located in very remote areas or relatively inaccessible. A critical point was noted for the Conservation forest, in which this area should had the lowest percentage in LULCC but actually, deforestation rates had become one of the highest in the study area.

In Chapter 5, the predictive power and reliability of several probabilistic cellular automata (CA) based LULCC models were compared and the most reliable model was applied to predict future land-cover states. Scenarios were built on the basis of transition probabilities that fall on each different forest land-use allocation policy as discussed in Chapter 4. Scenarios were conducted to visualize how if different land-allocation which implies different level of human activities were implemented into the whole study area. The first scenario was if LULCC in the study area was dictated by the transition probability of the whole study area which implied to a 'usual' LULCC. The second scenario was if LULCC in the study area was determined by the transition probability of the Non-forest area which implied a condition of 'rapid deforestation'. The third scenario was if LULCC in the study area was established by the transition probability of the Production forest area which implied a condition of 'rapid forest degradation'. Finally, the fourth scenario was if LULCC in the study area was determined by the transition probability of the Protection forest area which implied a condition of 'rapid forest regrowth'. Model comparison selected the generated probabilistic CA model with land allocation using 21×21 neighborhood size and incorporating local transition probability at 11×11 window size. The main result of Chapter 5 was the predicted future land cover maps based on four scenarios were created.

In Chapter 6, an attempt to employ GIS based Multi-criteria decision analysis (GIS-MCDA) was conducted for estimating forest and land fires risk as the main proximate cause of LULCC in the Middle Mahakam area. The main purpose of this chapter was to identify areas that are most susceptible to fire breaks in a spatially explicit method, thus to mitigates further unfavorable land-cover change in this area. Several criteria were included within the analysis, i.e. fuel index, climate index, topographic index, and human index. Each criterion was defined by one or several alternatives. Fire risk estimation as a multi-criteria decision making problem was formulated using the Analytical Hierarchy Process (AHP) technique. The AHP technique allows preferences of decision group to be incorporated into fire risk estimation. Decision group was six government officers (forestry service, agricultural service and environmental service) and six experts from universities (forest planning, forest fire and biological conservation). From twelve decision makers, only four decision makers (i.e. two experts and two government service officers) passed inconsistency measure. Fire risk map was then produced by combining decision-makers consideration to a single model using compromise programming and interactive thresholding method. To notice future fire risk in the study area, the predicted land cover change derived from Chapter 5 was incorporated into the analysis. The main result of Chapter 6 was production of the forest and land fire risks map based on actual land cover and scenario based predicted land cover.

General discussions and conclusions are remarked in Chapter 7. The importance of this study is situated in two case studies conducted in two areas which enclosed invaluable differences both physical and social conditions that gave rise to LULCC. The results of these case studies had brought to a deeper understanding about spatiotemporal patterns, processes, and causes of LULCC in the tropical environment, particularly, in Indonesia.

Case studies suggested that there were different processes and causes of LULCC occurred in the densely populated and sparsely populated areas. In a densely populated area, a number of mediating factors (e.g. farming as the main livelihood, low income, and the range of land-use option) were involved that encourage the local people to do agricultural expansion. In a sparsely populated area, LULCC is mostly underlined by institutional factor (e.g. land-use allocation policy) which was initially intended for natural resources exploitation. Other important proximate cause that mainly underlined by both institutional factor and biophysical factor is a short-term change such as the great fire of Borneo which occurred in 1997-1998.

Scenario based LULCC prediction highlighted interesting results. A 'rapid forest degradation' scenario implied that as long as good practice logging is implemented, it will maintain the forest cover in a continuous form. In addition, 'rapid forest regrowth' scenario seemed unrealistic; however, this provides a different outlook that tropical forest may possibly regain as long as human pressure to the forest is minimized. These results again emphasized that land-use allocation policy had a major impact that underlined various proximate causes of LULCC in the study area.

By this study, land and forest fire risk estimation maps were developed as a mitigation strategy for further unfavorable land-cover changes. These maps are also important as an initial reference to seek a good management practice for determining trade-offs between the benefits of fires, and the risks to human life, capital, and production caused by fires. The results call for a more cautious land use should be implemented on the area with higher risk of fire.

土地被覆および土地利用変化(LULCC)は地球環境変化の唯一の要因ではないが、気候変動や生態系サービス、持続性に対して大きな影響を与えるものである。地球規模のLULCCの中で熱帯環境は大きな位置を占めており、特に熱帯林は急速に減少していることが知られている。熱帯林減少研究では、森林/非森林という二値的プロセスを用いたアプローチが一般的であるが、LULCCの動態には植生被覆喪失と共にその再生プロセスも含まれ、正味の収支はその両者の差引である。この差引を吟味するためには、LULCCのプロセスや原因だけでなく、その時空間パターンを深く理解することが求められる。

本研究の主目的は、熱帯環境のよりよい管理のために、LULCCのプロセスや原因を明らかにし、時空間パターンを理解することである。本研究は7章からなり、第1章および2章では研究の背景と目的、研究対象地について述べた。

第3章では、インドネシア西ジャワ州ダナウ湖自然保護区における農地拡大の時空間パターンを衛星リモートセンシング(RS)データを用いて明らかにし、農地拡大を引き起こしている地域住民の状況を社会経済的な世帯調査により特徴づけた。RSデータ解析により1997年の経済危機後に急速な農地拡大が発生したことが、世帯調査から自然保護区への近接性以外に土地利用手段の多寡が農地拡大に影響を与えていることが明らかになった。

第4章では、本研究で開発した改良型ハイブリッド土地被覆変化解析手法を用いて、インドネシア東カリマンタン州マハカム河中流域のLULCCを明らかにし、この変化傾向とインドネシア政府が1980年代に策定した森林の土地利用政策(TGHK)との関連を評価した。その結果、この地域では森林減少および劣化が森林再生より遙かに多く発生していること、TGHKが土地被覆変化に大きな影響を与えていることがわかった。森林減少は非森林区域で、森林劣化は生産林区域で発生している。LULCCは保安林区域で最も少なかったが、これはこの区域が遠隔地で到達困難であることによる。重要な点は、LULCCが最小であるべき保護林区域で森林減少割合が最も高かったことである。

第5章では、複数の確率セルオートマン(CA)に基づくLULCCモデルの予測能力と信頼性の比較を行い、21×21近傍で11×11ウィンドウサイズの遷移確率を適用し面積配分を行った確率CAモデルが最適であることを明らかにした。そしてTGHK区分毎の遷移確率を利用した予測シナリオに基づき、異なった4段階の強度の人間活動が地域全体に行われた場合を予測し、将来の土地被覆予測図を作成した。

第6章では、GISによる多基準意思決定分析を用いて、マハカム河中流域のLULCC重要原因である森林火災リスクの推定を行った。最も被災しやすい地域の地理的位置を明らかにすることで、対象地域の好ましくないこれ以上の土地被覆変化を軽減することが目的である。12名からなる意思決定者グループに対し、階層化分析法を用いて森林火災リスク推定の4基準(燃料、気候、地形、人間)を構成する複数の代替案についてウェイトを決定し、妥協的計画法により森林火災リスクマップを作成した。さらに、前章の土地被覆変化予測結果を用いて将来のリスクマップも作成した。

総合考察と結論を第7章で述べた。人口の低密度地域と高密度地域では異なったLULCCのプロセスと原因があったことが、二つのケーススタディにより示唆された。高密度地域では数々の間接的な要因が地域住民による農地拡大を助長しているのに対し、低密度地域では当初は自然資源開発を目的とした制度的要因がLULCCに大きな影響を与えている。シナリオに基づいたLULCC予測では、適切な伐採行為が行われれば森林被覆は持続的に保たれること、人間による森林への圧力が最小化されれば熱帯林は回復することが示されたことから、この地域におけるTGHKの影響の大きさが明らかになった。また本研究で作成した森林火災リスクマップは、火災による便益と人間生活・財産・生産に対するリスクとのトレードオフを見極め正しい管理活動を模索する際に、まず最初に参照されるべきものとして重要である。

以上、本論文において、衛星RSデータを用いた土地被覆変化解析手法および確率CAモデルに関する新たな手法が開発され、LULCCの時空間パターンを考える上で非常に有効な手段であることが明らかにされただけではなく、複数の将来シナリオに基づく森林火災リスクマップ作成が行われ、今後の当該地域の土地利用計画策定に資する有用な情報が示された。その成果は学術面だけでなく、地方政府の政策立案といった応用面でも貢献するところが少なくない。よって審査委員一同は、本論文が博士(農学)の学位論文として価値あるものと認めた。