Alokasi Lahan Tanaman Pangan pada Ruang Antar Muka Kehutanan dan Pertanian

Abstract

Indonesia merupakan salah satu negara dengan luas tutupan hutan tropis terluas di dunia yang berperan penting sebagai sumber keanekaragaman hayati dan fungsi ekologi secara global. Namun demikian, saat ini deforestasi atau beralihnya tutupan hutan menjadi penggunaan lainnya masih menjadi ancaman. Tahun 2021 KLHK melaporkan rata-rata deforestasi tahun 2010-2020 sebesar 541 ribu hektar per tahun, sementara FAO tahun 2020 mengestimasi rata-rata deforestasi Indonesia sekitar 753 ribu hektar per tahun pada kurun waktu yang sama. Berdasarkan hasil penelitian terdahulu, fenomena deforestasi didorong berbagai faktor antara lain illegal logging, kebakaran hutan dan lahan, industri hutan alam, hutan tanaman, perkebunan, pertambangan, serta ekspansi lahan pertanian. Fenomena yang terakhir, diperkirakan akan semakin meningkat dengan peningkatan kebutuhan pangan yang merupakan konsekuensi terus bertambahnya populasi. Disisi lain, pemerintah Indonesia kembali berupaya melakukan ekspansi lahan pertanian terutama di luar Pulau Jawa melalui program food estate. Upaya ini memiliki potensi tinggi bersinggungan dengan hutan dan kawasan hutan, mengingat saat ini tutupan hutan di Indonesia seluas 95,6 juta hektar, sementara luas kawasan hutan mencapai 125 juta hektar atau masing-masing mencapai 50,9% dan 66,5% dari total luas daratan. Dengan demikian, ekstensifikasi lahan untuk pengembangan pertanian perlu memperhatikan eksistensi hutan dan kawasan hutan guna menghindari berbagai dampak ekologis negatif yang ditimbulkan. Tujuan utama penelitian ini adalah memberikan arahan alokasi lahan tanaman pangan pada ruang antar muka kehutanan dan pertanian melalui beberapa tahapan tujuan antara yaitu: 1) menganalisis tutupan lahan dan kecenderungan deforestasi serta faktor pendorongnya, 2) mengidentifikasi kondisi kesesuaian lahan tanaman pangan, 3) menyusun skenario alokasi lahan dalam meminimalkan deforestasi serta mendorong kecukupan pangan. Padi (Oryza sativa) dan Sorgum (Sorgum bicolor) dipilih sebagai jenis tanaman yang dianalisis. Machine learning digunakan untuk menganalisis tutupan lahan multi temporal tahun 2004, 2013, dan 2022 dengan menggunakan citra Landsat. Random Forest (RF) dan Support Vector Machines (SVM) diperbandingkan kinerjanya berdasarkan overall accuracy (OA) yang dihasilkan dalam mengklasifikasikan citra satelit tahun 2022. Selanjutnya, dinamika tutupan lahan dianalisis untuk mengetahui kecenderungan perubahan, khususnya deforestasi. Analisis regresi logistik dilakukan untuk mengidentifikasi faktor pendorong dengan mengintegrasikan peta biner terhadap berbagai variabel prediktor deforestasi. Sementara, analisis kesesuaian lahan dilakukan menggunakan metode multi criteria evaluation (MCE) dimana masing-masing parameter diboboti melalui teknik analytical hierarchy process (AHP). Opini pakar digunakan untuk menentukan bobot masing-masing parameter untuk jenis tanaman Padi dan Sorgum. Terakhir, peta kawasan hutan diintegrasikan untuk mengidentifikasi potensi serta menyusun skenario alokasi lahan pada ruang antar muka kehutanan dan pertanian. Hasil analisis menunjukkan OA yang diperoleh RF di semua penyetelan eksperimental lebih unggul dari SVM pada berbagai jenis kernel, dimana hasil optimal diperoleh pada penyetelan ntree=50. Hasil analisis tutupan lahan multi temporal menunjukkan, tutupan hutan mendominasi area studi > 74% selama periode pengamatan dengan kecenderungan luasan yang terus menurun. Hasil analisis menunjukkan luasan kehilangan hutan bersih (net forest loss) meningkat antara periode pertama ke periode kedua, dengan luas total selama waktu pengamatan sebesar 6,93%, atau (~141.317 ha). Namun demikian, total deforestasi selama waktu pengamatan mencapai 247.108 ha atau rata-rata hampir 14 ribu ha per tahun. Laju deforestasi cenderung mengalami peningkatan selama periode pengamatan, dimana berdasarkan pengamatan citra satelit, ekspansi perkebunan kelapa sawit ditemukan menjadi penyebab utama deforestasi. Sebagai tambahan, hasil analisis faktor pendorong deforestasi menunjukkan bahwa kawasan hutan memiliki peluang deforestasi lebih kecil dibandingkan APL. Selain itu, kawasan lindung terbukti berperan penting dalam meminimalisir serta menghambat laju deforestasi. Kami juga menemukan peluang deforestasi di dalam area izin pemanfaatan hutan lebih kecil dibandingkan di luar perizinan. Selanjutnya, hasil analisis menggunakan AHP menunjukkan parameter jarak ke sumber air, pH, dan jenis tanah mendapatkan bobot tertinggi untuk tanaman Padi. Sementara untuk tanaman Sorgum yaitu pH, jenis tanah dan jarak ke jalan. Berdasarkan hasil analisis MCE, sebaran kelas kesesuaian untuk tanaman Padi terdiri dari: sangat sesuai (S1) sebesar 4% (74.254 ha), sesuai (S2) sebesar 6% (130.634 ha), sesuai marginal (S3) sebesar 38% (769.078 ha), dan tidak sesuai (N) sebesar 52% (1.036.082 ha). Sementara, untuk Sorgum: S1 sebesar 5% (108.956 ha), S2 sebesar 19% (377.493 ha), S3 sebesar 48% (976.820 ha), dan N sebesar 28% (574.635 ha). Integrasi metode MCE dan AHP pada lingkungan GIS terbukti dapat menghasilkan klasifikasi kelas kesesuaian lahan secara efektif dan dapat diandalkan. Penerapan dua skenario yaitu Bussines as Usual (BAU) dan penggunaan kawasan hutan menunjukkan hasil sebagai berikut. Pertama, pada kondisi BAU terjadi penurunan luas yang signifikan pada potensi alokasi lahan baik untuk Padi maupun Sorgum. Hal ini disebabkan karena alokasi hanya dilakukan di luar kawasan hutan /APL. Kedua, pada penerapan skenario penggunaan kawasan hutan, potensi alokasi tanaman Padi tanpa menyebabkan deforestasi diidentifikasi seluas hampir 333 ribu ha, sedangkan hampir 67 ribu ha lainnya masih berhutan. Potensi alokasi tanaman Sorgum tanpa menyebabkan deforestasi seluas hampir 384 ribu ha, sedangkan hampir 205 ribu ha lainnya masih berhutan. Hasil analisis menunjukkan besarnya potensi alokasi pada hutan produksi khususnya Hutan Produksi yang dapat Dikonversi (HPK). Sehingga, secara umum dapat disimpulkan bahwa kawasan hutan memiliki potensi yang signifikan di dalam pengembangan lahan pertanian tanaman pangan.

Indonesia is one of the countries with the largest tropical forest cover in the world, which serves as a vital source of biodiversity and ecological functions globally. However, deforestation or the conversion of forest cover to other uses is still a threat. In 2021, the Ministry of Environment and Forestry reported that the average deforestation for the period 2010–2020 was 541 thousand hectares per year, while the FAO estimated that the average deforestation for Indonesia during the same period was approximately 753 thousand hectares per year. According to prior studies, deforestation can be driven by a multitude of variables, such as illegal harvesting, land and forest fires, the natural forest industry, plantation forests, plantations, mining, and the expansion of agricultural land. The latter phenomenon is expected to be widespread with increasing food needs, which is a consequence of the continuous rise in population. On the other hand, the Indonesian government is attempting to expand agricultural land, particularly outside of Java, through the food estate program. However, this endeavor has the potential to intersect with forests and forest area designation (FAD), given that the current forest coverage remains 95,6 million hectares, and the area of FAD is 125 million acres, or 50.9% and 66.5% of the total land mass, respectively. Consequently, the expansion of land for agricultural development must consider the presence of forests and forest areas to prevent various kinds of adverse environmental impacts. This study aims to allocate the land for food crops at the interfaces of forestry and agriculture through several stages of objectives: 1) analyze multi-temporal land cover change, deforestation trend, and their driving factors, 2) identify land suitability for certain crops, 3) develop land allocation scenarios to minimize deforestation and promote food sufficiency. Rice (Oryza sativa) and sorghum (Sorghum bicolor) were selected as the plant species analyzed. Machine learning is used to analyze multi-temporal land cover in 2004, 2013, and 2022 using Landsat imagery. Random Forest (RF) and Support Vector Machines (SVM) are compared for their performance in classifying satellite imagery in 2022 based on overall accuracy (OA). The dynamics of land cover are then analyzed to figure out trends of change, particularly deforestation. Utilizing a binary map of deforestation along with a number of predictor variables, a logistic regression analysis was conducted to estimate the drivers of deforestation. In the meantime, land suitability analysis was conducted using the multi-criteria evaluation (MCE) method, with each parameter weighted using the analytical hierarchy process (AHP) method. Expert opinion is used to determine the weight of each parameter for rice and sorghum, respectively. Moreover, the FAD map is incorporated to identify potentials and develop land allocation scenarios at the forestry-agriculture interface. The results of the analysis reveal that the OA obtained by RF at all experimental setups to be superior to SVM at various kernel types, where optimal results were obtained at ntree=50 setup. The multi-temporal analysis of land cover reveals that forest cover comprises > 74% of the study area during the observation period, with a decreasing area trend. The results of the analysis show that the area of net forest loss increased between the first period and the second period, with a total area during the observation period of 6.93%, or (~141,317 ha). However, total deforestation during the observation period reached 247,108 ha, or an annual average of nearly 14 thousand hectares. The rate of deforestation tends to increase during the observation period. The rate of deforestation tends to increase during the observation period, whereas based on satellite imagery observations, the expansion of oil palm plantations was the primary cause of deforestation. In addition, the results of the analysis of drivers of deforestation indicate that FAD is associated with a lower likelihood of deforestation than APL. Likewise, protected areas have been shown to play an essential part in reducing and preventing deforestation rates. We also discovered that the likelihood of deforestation within the permit area is lower than outside the permit. In addition, the AHP analysis revealed that for rice plants, the parameters of distance to water sources, pH, and soil type were given the most weight. In the meantime, for sorghum plants, pH, soil type, and road proximity are crucial. Based on the results of the MCE analysis, the distribution of suitability classes for rice plants was as follows: high suitable (S1) for 4% (74,254 ha), moderately suitable (S2) for 6% (130,634 ha), marginally suitable (S3) for 38% (769,078 ha), and unsuitable (N) for 52% (1,036,560 ha). Meanwhile, for sorghum: S1 was 5% (108,956 ha), S2 was 19% (377,493 ha), S3 was 48% (976,820 ha), and N was 28% (574,635 ha). The integration of the MCE and AHP methods within a GIS environment has been demonstrated to produce a reliable and effective classification of land suitability classes. The implementation of two scenarios, Business as Usual (BAU) and forest area utilization, shows the following results. First, under BAU conditions, the potential land allocated for rice and sorghum decreased significantly. Since allocations are only conducted outside the FAD. Second, in implementing the scenario of utilizing forest areas, the potential allocation of rice plants without causing deforestation has been identified on nearly 333 thousand hectares of land, while nearly 67 thousand hectares of land remain forested. Almost 384 thousand hectares are available for sorghum cultivation without causing degradation, whereas nearly 205 thousand hectares are still forested. The results found a substantial allocation potential in production forests, particularly in convertible production forests (HPK). Thus, it is possible to conclude, in general, that FAD offer substantial potential for the development of agricultural land for food commodities.