Domain Adaptation in remote sensing: increasing the portability of land-cover classifiers

Among the types of remote sensing acquisitions, optical images are certainly one of the most widely relied upon data sources for Earth observation. They provide detailed measurements of the electromagnetic radiation reflected or emitted by each pixel in the scene. Through a process termed supervised land-cover classification, this allows to automatically yet accurately distinguish objects at the surface of our planet. In this respect, when producing a land-cover map of the surveyed area, the availability of training examples representative of each thematic class is crucial for the success of the classification procedure. However, in real applications, due to several constraints on the sample collection process, labeled pixels are usually scarce. When analyzing an image for which those key samples are unavailable, a viable solution consists in resorting to the ground truth data of other previously acquired images. This option is attractive but several factors such as atmospheric, ground and acquisition conditions can cause radiometric differences between the images, hindering therefore the transfer of knowledge from one image to another. The goal of this Thesis is to supply remote sensing image analysts with suitable processing techniques to ensure a robust portability of the classification models across different images. The ultimate purpose is to map the land-cover classes over large spatial and temporal extents with minimal ground information. To overcome, or simply quantify, the observed shifts in the statistical distribution of the spectra of the materials, we study four approaches issued from the field of machine learning. First, we propose a strategy to intelligently sample the image of interest to collect the labels only in correspondence of the most useful pixels. This iterative routine is based on a constant evaluation of the pertinence to the new image of the initial training data actually belonging to a different image. Second, an approach to reduce the radiometric differences among the images by projecting the respective pixels in a common new data space is presented. We analyze a kernel-based feature extraction framework suited for such problems, showing that, after this relative normalization, the cross-image generalization abilities of a classifier are highly increased. Third, we test a new data-driven measure of distance between probability distributions to assess the distortions caused by differences in the acquisition geometry affecting series of multi-angle images. Also, we gauge the portability of classification models through the sequences. In both exercises, the efficacy of classic physically- and statistically-based normalization methods is discussed. Finally, we explore a new family of approaches based on sparse representations of the samples to reciprocally convert the data space of two images. The projection function bridging the images allows a synthesis of new pixels with more similar characteristics ultimately facilitating the land-cover mapping across images.

Physical quality of an Oxisol cultivated with maize submited to cover crops in the pre-cropping period

The intensive use of land alters the distribution of the pore size which imparts consequences on the soil physical quality. The Least Limiting Water Range (LLWR) allows for the visualization of the effects of management systems upon either the improvement or the degradation of the soil physical quality. The objective of this study was to evaluate the physical quality of a Red Latosol (Oxisol) submited to cover crops in the period prior to the maize crop in a no-tillage and conventional tillage system, using porosity, soil bulk density and the LLWR as attributes. The treatments were: conventional tillage (CT) and a no-tillage system with the following cover crops: sunn hemp (Crotalaria juncea L.) (NS), pearl millet (Pennisetum americanum (L.) Leeke) (NP) and lablab (Dolichos lablab L.) (NL). The experimental design was randomized blocks in subdivided plots with six replications, with the plots being constituted by the treatments and the subplots by the layers analyzed. The no-tillage systems showed higher total porosity and soil organic matter at the 0-0.5 m layer for the CT. The CT did not differ from the NL or NS in relation to macroporosity. The NP showed the greater porosity, while CT and NS presented lower soil bulk density. No < 10 % airing porosity was found for the treatments evaluated, and value for water content where soil aeration is critical (θPA) was found above estimated water content at field capacity (θFC) for all densities. Critical soil bulk density was of 1.36 and 1.43 Mg m-3 for NP and CT, respectively. The LLWR in the no-tillage systems was limited in the upper part by the θFC, and in the bottom part, by the water content from which soil resistance to penetration is limiting (θPR). By means of LLWR it was observed that the soil presented good physical quality.

Morphostructural characterization of soil conventionally tilled with mechanized and animal traction with and without cover crop

The structural stability and restructuring ability of a soil are related to the methods of crop management and soil preparation. A recommended strategy to reduce the effects of soil preparation is to use crop rotation and cover crops that help conserve and restore the soil structure. The aim of this study was to evaluate and quantify the homogeneous morphological units in soil under conventional mechanized tillage and animal traction, as well as to assess the effect on the soil structure of intercropping with jack bean (Canavalia ensiformis L.). Profiles were analyzed in April of 2006, in five counties in the Southern-Central region of Paraná State (Brazil), on family farms producing maize (Zea mays L.), sometimes intercropped with jack bean. The current structures in the crop profile were analyzed using Geographic Information Systems (GIS) and subsequently principal component analysis (PCA) to generate statistics. Morphostructural soil analysis showed a predominance of compact units in areas of high-intensity cultivation under mechanized traction. The cover crop did not improve the structure of the soil with low porosity and compact units that hamper the root system growth. In areas exposed to animal traction, a predominance of cracked units was observed, where roots grew around the clods and along the gaps between them.

Nitrogen washing from C3 and C4 cover grasses residues by rain

Crop species with the C4 photosynthetic pathway are more efficient in assimilating N than C3 plants, which results in different N amounts prone to be washed from its straw by rain water. Such differences may affect N recycling in agricultural systems where these species are grown as cover crops. In this experiment, phytomass production and N leaching from the straw of grasses with different photosynthetic pathways were studied in response to N application. Pearl millet (Pennisetum glaucum) and congo grass (Brachiaria ruziziensis) with the C4 photosynthetic pathway, and black oat (Avena Strigosa) and triticale (X Triticosecale), with the C3 photosynthetic pathway, were grown for 47 days. After determining dry matter yields and N and C contents, a 30 mm rainfall was simulated over 8 t ha-1 of dry matter of each plant residue and the leached amounts of ammonium and nitrate were determined. C4 grasses responded to higher fertilizer rates, whereas N contents in plant tissue were lower. The amount of N leached from C4 grass residues was lower, probably because the C/N ratio is higher and N is more tightly bound to organic compounds. When planning a crop rotation system it is important to take into account the difference in N release of different plant residues which may affect N nutrition of the subsequent crop.

Tillage systems and nutrient sources affecting soil cover, temperature and moisture in a clayey oxisol under corn

Tillage affects soil physical properties, e.g., porosity, and leads to different amounts of mulch on the soil surface. Consequently, tillage is related to the soil temperature and moisture regime. Soil cover, temperature and moisture were measured under corn (Zea mays) in the tenth year of five tillage systems (NT = no-tillage; CP = chisel plow and single secondary disking; CT = primary and double secondary disking; CTb = CT with crop residues burned; and CTr = CT with crop residues removed). The tillage systems were combined with five nutrient sources (C = control; MF = mineral fertilizer; PL = poultry litter; CS = cattle slurry; and SS = swine slurry). Soil cover after sowing was greatest in NT (88 %), medium in CP (38 %) and lowest in CT treatments (< 10 %), but differences decreased after corn emergence. Soil temperature was related with soil cover, and significant differences among tillage were observed at the beginning of the growing season and at corn maturity. Differences in soil temperature and moisture in the surface layer of the tilled treatments were greater during the corn cycle than in untilled treatments, due to differences in intensity of soil mobilization and mulch remaining after soil management. Nutrient sources affected soil temperature and moisture in the most intense part of the corn growth period, and were related to the variation of the corn leaf area index among treatments

Phosphatase activity in sandy soil influenced by mycorrhizal and non-mycorrhizal cover crops

Cover crops may difffer in the way they affect rhizosphere microbiota nutrient dynamics. The purpose of this study was to evaluate the effect of mycorrhizal and non-mycorrhizal cover crops on soil phosphatase activity and its persistence in subsequent crops. A three-year experiment was carried out with a Typic Quartzipsamment. Treatments were winter species, either mycorrhizal black oat (Avena strigosa Schreb) or the non-mycorrhizal species oilseed radish (Raphanus sativus L. var. oleiferus Metzg) and corn spurry (Spergula arvensis L.). The control treatment consisted of resident vegetation (fallow in the winter season). In the summer, a mixture of pearl millet (Pennisetum americanum L.) with sunnhemp (Crotalaria juncea L.) or with soybean (Glycine max L.) was sown in all plots. Soil cores (0-10 cm) and root samples were collected in six growing seasons (winter and summer of each year). Microbial biomass P was determined by the fumigation-extraction method and phosphatase activity using p-nitrophenyl-phosphate as enzyme substrate. During the flowering stage of the winter cover crops, acid phosphatase activity was 30-35 % higher in soils with the non-mycorrhizal species oilseed radish, than in the control plots, regardless of the amount of P immobilized in microbial biomass. The values of enzyme activity were intermediate in the plots with corn spurry and black oat. Alkaline phosphatase activity was 10-fold lower and less sensitive to the treatments, despite the significant relationship between the two phosphatase activities. The effect of plant species on the soil enzyme profile continued in the subsequent periods, during the growth of mycorrhizal summer crops, after completion of the life cycle of the cover crops.

Microbial biomass and soil fauna during the decomposition of cover crops in no-tillage system

The decomposition of plant residues is a biological process mediated by soil fauna, but few studies have been done evaluating its dynamics in time during the process of disappearance of straw. This study was carried out in Chapecó, in southern Brazil, with the objective of monitoring modifications in soil fauna populations and the C content in the soil microbial biomass (C SMB) during the decomposition of winter cover crop residues in a no-till system. The following treatments were tested: 1) Black oat straw (Avena strigosa Schreb.); 2) Rye straw (Secale cereale L.); 3) Common vetch straw (Vicia sativa L.). The cover crops were grown until full flowering and then cut mechanically with a rolling stalk chopper. The soil fauna and C content in soil microbial biomass (C SMB) were assessed during the period of straw decomposition, from October 2006 to February 2007. To evaluate C SMB by the irradiation-extraction method, soil samples from the 0-10 cm layer were used, collected on eight dates, from before until 100 days after residue chopping. The soil fauna was collected with pitfall traps on seven dates up to 85 days after residue chopping. The phytomass decomposition of common vetch was faster than of black oat and rye residues. The C SMB decreased during the process of straw decomposition, fastest in the treatment with common vetch. In the common vetch treatment, the diversity of the soil fauna was reduced at the end of the decomposition process.

Soil microbial biomass under different management and tillage systems of permanent intercropped cover species in an orange orchard

To mitigate soil erosion and enhance soil fertility in orange plantations, the permanent protection of the inter-rows by cover species has been suggested. The objective of this study was to evaluate alterations in the microbial biomass, due to different soil tillage systems and intercropped cover species between rows of orange trees. The soil of the experimental area previously used as pasture (Brachiaria humidicola) was an Ultisol (Typic Paleudult) originating from Caiuá sandstone in the northwestern part of the State of Paraná, Brazil. Two soil tillage systems were evaluated: conventional tillage (CT) in the entire area and strip tillage (ST) (strip width 2 m), in combination with different ground cover management systems. The citrus cultivar 'Pera' orange (Citrus sinensis) grafted onto 'Rangpur' lime rootstock was used. Soil samples were collected after five years of treatment from a depth of 0-15 cm, under the tree canopy and in the inter-row, in the following treatments: (1) CT and an annual cover crop with the leguminous species Calopogonium mucunoides; (2) CT and a perennial cover crop with the leguminous peanut Arachis pintoi; (3) CT and an evergreen cover crop with Bahiagrass Paspalum notatum; (4) CT and a cover crop with spontaneous Brachiaria humidicola grass vegetation; and (5) ST and maintenance of the remaining grass (pasture) of Brachiaria humidicola. Soil tillage and the different cover species influenced the microbial biomass, both under the tree canopy and in the inter-row. The cultivation of brachiaria increased C and N in the microbial biomass, while bahiagrass increased P in the microbial biomass. The soil microbial biomass was enriched in N and P by the presence of ground cover species and according to the soil P content. The grass species increased C, N and P in the soil microbial biomass from the inter-row more than leguminous species.

Soil physical properties and grape yield influenced by cover crops and management systems

The use of cover crops in vineyards is a conservation practice with the purpose of reducing soil erosion and improving the soil physical quality. The objective of this study was to evaluate cover crop species and management systems on soil physical properties and grape yield. The experiment was carried out in Bento Gonçalves, RS, Southern Brazil, on a Haplic Cambisol, in a vineyard established in 1989, using White and Rose Niagara grape (Vitis labrusca L.) in a horizontal, overhead trellis system. The treatments were established in 2002, consisting of three cover crops: spontaneous species (SS), black oat (Avena strigosa Schreb) (BO), and a mixture of white clover (Trifolium repens L.), red clover (Trifolium pratense L.) and annual rye-grass (Lolium multiflorum L.) (MC). Two management systems were applied: desiccation with herbicide (D) and mechanical mowing (M). Soil under a native forest (NF) area was collected as a reference. The experimental design consisted of completely randomized blocks, with three replications. The soil physical properties in the vine rows were not influenced by cover crops and were similar to the native forest, with good quality of the soil structure. In the inter-rows, however, there was a reduction in biopores, macroporosity, total porosity and an increase in soil density, related to the compaction of the surface soil layer. The M system increased soil aggregate stability compared to the D system. The treatments affected grapevine yield only in years with excess or irregular rainfall.

Chemical and biological properties of phosphorus-fertilized soil under legume and grass cover (Cerrado region, Brazil)

The use of cover crops has been suggested as an effective method to maintain and/or increase the organic matter content, while maintaining and/or enhancing the soil physical, chemical and biological properties. The fertility of Cerrado soils is low and, consequently, phosphorus levels as well. Phosphorus is required at every metabolic stage of the plant, as it plays a role in the processes of protein and energy synthesis and influences the photosynthetic process. This study evaluated the influence of cover crops and phosphorus rates on soil chemical and biological properties after two consecutive years of common bean. The study analyzed an Oxisol in Selvíria (Mato Grosso do Sul, Brazil), in a randomized block, split plot design, in a total of 24 treatments with three replications. The plot treatments consisted of cover crops (millet, pigeon pea, crotalaria, velvet bean, millet + pigeon pea, millet + crotalaria, and millet + velvet bean) and one plot was left fallow. The subplots were represented by phosphorus rates applied as monoammonium phosphate (0, 60 and 90 kg ha-1 P2O5). In August 2011, the soil chemical properties were evaluated (pH, organic matter, phosphorus, potential acidity, cation exchange capacity, and base saturation) as well as biological variables (carbon of released CO2, microbial carbon, metabolic quotient and microbial quotient). After two years of cover crops in rotation with common bean, the cover crop biomass had not altered the soil chemical properties and barely influenced the microbial activity. The biomass production of millet and crotalaria (monoculture or intercropped) was highest. The biological variables were sensitive and responded to increasing phosphorus rates with increases in microbial carbon and reduction of the metabolic quotient.

Upland rice under no-tillage preceded by crops for soil cover and nitrogen fertilization

The grain yield of upland rice under no-tillage has been unsatisfactory and one reason could be the nitrate/ammonium balance in the soil. Cover crops and nitrogen fertilization can be used to change the nitrate/ammonium relation in the soil and improve conditions for the development of upland rice in the no-tillage (NT) system. The aim was to study the effect of cover crops and nitrogen sources on grain yield of upland rice under no tillage. The study was carried out on the Fazenda Experimental Lageado, in Botucatu, State of São Paulo, Brazil, in an Oxisol area under no-tillage for six years. The experiment was arranged in a randomized block split-plot design with four replications. The plots consisted of six cover crop species (Brachiaria brizantha, B. decumbens, B. humidicola, B. ruziziensis, Pennisetum americanum, and Crotalaria spectabilis) and the split-plots of seven forms of N fertilizer management. Millet is the best cover crop to precede upland rice under NT. The best form of N application, as nitrate, is in split rates or total rate at topdressing or an ammonium source with or without a nitrification inhibitor, in split doses. When the cover crops C. spectabilis, B. brizantha, B. decumbens, B. humidicola, and B. ruziziensis preceded rice, they induced the highest grain yield when rice was fertilized with N as ammonium sulfate source + nitrification inhibitor in split rates or total dose at topdressing.

Effect of management systems and cover crops on organic matter dynamics of soil under vegetables

Vegetable production in conservation tillage has increased in Brazil, with positive effects on the soil quality. Since management systems alter the quantity and quality of organic matter, this study evaluated the influence of different management systems and cover crops on the organic matter dynamics of a dystrophic Red Latosol under vegetables. The treatments consisted of the combination of three soil tillage systems: no-tillage (NT), reduced tillage (RT) and conventional tillage (CT) and of two cover crops: maize monoculture and maize-mucuna intercrop. Vegetables were grown in the winter and the cover crops in the summer for straw production. The experiment was arranged in a randomized block design with four replications. Soil samples were collected between the crop rows in three layers (0.0-0.05, 0.05-0.10, and 0.10-0.30 m) twice: in October, before planting cover crops for straw, and in July, during vegetable cultivation. The total organic carbon (TOC), microbial biomass carbon (MBC), oxidizable fractions, and the carbon fractions fulvic acid (C FA), humic acid (C HA) and humin (C HUM) were determined. The main changes in these properties occurred in the upper layers (0.0-0.05 and 0.05-0.10 m) where, in general, TOC levels were highest in NT with maize straw. The MBC levels were lowest in CT systems, indicating sensitivity to soil disturbance. Under mucuna, the levels of C HA were lower in RT than NT systems, while the C FA levels were lower in RT than CT. For vegetable production, the C HUM values were lowest in the 0.05-0.10 m layer under CT. With regard to the oxidizable fractions, the tillage systems differed only in the most labile C fractions, with higher levels in NT than CT in the 0.0-0.05 m layer in both summer and winter, with no differences between these systems in the other layers. The cabbage yield was not influenced by the soil management system, but benefited from the mulch production of the preceding maize-mucuna intercrop as cover plant.

Forms of phosphorus in an oxisol under different soil tillage systems and cover plants in rotation with maize

Phosphorus fractions play a key role in sustaining the productivity of acid-savanna Oxisols and are influenced by tillage practices. The aim of this study was to quantify different P forms in an Oxisol (Latossolo Vermelho-Amarelo) from the central savanna region of Brazil under management systems with cover crops in maize rotation. Three cover crops (Canavalia brasiliensis, Cajanus cajan (L.), and Raphanus sativus L.) were investigated in maize rotation systems. These cover crops were compared to spontaneous vegetation. The inorganic forms NaHCO3-iP and NaOH-iP represented more than half of the total P in the samples collected at the depth of 5-10 cm during the rainy season when the maize was grown. The concentration of inorganic P of greater availability (NaHCO3-iP and NaOH-iP) was higher in the soil under no-tillage at the depth of 5-10 cm during the rainy season. Concentrations of organic P were higher during the dry season, when the cover crops were grown. At the dry season, organic P constituted 70 % of the labile P in the soil planted to C. cajan under no-tillage. The cover crops were able to maintain larger fractions of P available to the maize, resulting in reduced P losses to the unavailable pools, mainly in no-tillage systems.

Cover plants and mineral nitrogen: effects on organic matter fractions in an oxisol under no-tillage in the cerrado

Cover plants are essential for the sustainability of no-tillage systems in tropical regions. However, information on the effects of these plants and N fertilization on soil organic matter fractions is still scarce. This study evaluated the effect of cover crops with different chemical composition and of N topdressing on the labile and humified organic matter fractions of an Oxisol of the Cerrado (savanna-like vegetation). The study in a randomized complete block design was arranged in split-plots with three replications. Four cover species were tested in the plots and the presence or absence of N topdressing in the subplot. The following cover species were planted in succession to corn for eight years: Urochloa ruziziensis; Canavalia brasiliensis M. ex Benth; Cajanus cajan (L.) Millsp; and Sorghum bicolor (L.) Moench. In general, the cultivation of U. ruziziensis increased soil C levels, particularly of C in the humic acid and particulate organic C fractions, which are quality indicators of soil organic matter. The C in humic substances and mineral organic C accounted for the highest proportions of total organic C, demonstrating the strong interaction between organic matter, Fe and Al oxides and kaolinite, which are predominant in these weathered soils of the Cerrado.