Assessing the suitability of using Sentinel-1A SAR multi-temporal imagery to detect fallow periods between rice crops

Abstract

Fallow, understood as a period where no crop is cultivated during a growing season, is an important crop management practice. Fallow plays an important role in pest and disease management, and monitoring the presence and duration of fallow can have implications for understanding the extent and effectiveness of this agricultural practice. Therefore, fallow period mapping using remotely sensed data is a widely researched topic. However, its identification using remotely sensed data is often problematic due to the fallow backscatter values being similar to other land cover classes. SAR data tends to be applied to crop identification, paying little to no attention to fallow. This thesis investigated fallow backscatter characteristics and detection methods that could be used for identification of fallow periods between rice crops. Timeseries of backscatter from Sentinel-1A data were plotted for 83 fields in two areas in the Philippines. Rice crop and fallow temporal characteristics were derived from field data and correlated with the plotted backscatter.

The start and end of fallow showed significant differences that could be used for detecting fallow periods. However, the detection gave acceptable results only using VH polarisation and the VV/VH ratio, while VV polarisation resulted in the lowest accuracy. A backward fallow detection method was also employed, where the end of the rice season indicated the fallow start, and the start of the rice season marked the fallow end. This method proved more successful in determining fallow duration than using the fallow itself. Fallow duration was also categorised according to its length into short and long fallow. In short fallow periods, backscatter was mainly decreasing during the whole fallow period, while in long fallow periods, some variation was observed approximately halfway through the fallow duration, which were attributed to short-term drought tolerant crops. Fallow backscatter values for irrigated and rainfed fields were also compared. Ecosystem type was found to have a significant effect on fallow backscatter only in one of the two study areas, indicating that water availability may affect fallow duration. The study found that SAR backscatter can be used to map fallow by detecting rice seasons, although the detected fallow duration was slightly overestimated. The study lacked information about the conditions of the fields during the fallow periods; therefore, future studies could benefit from more information about weed prevalence, planting of non-rice crops, water or rainfall deficit, and flooding/tillage activities during or after fallow.

Popular Abstract

Fallow is a type of agricultural practice that takes place after crop harvest. It refers to a time when no crop is being cultivated for a given amount of time before planting the next crop. During fallow, the soil has time to regenerate itself. In addition, the lack of crops prevents pests and disease to develop. It is therefore important to understand fallow practice. This research focuses on fallow between rice crops in the Philippines. It uses SAR (radar) satellite images to assess whether fallow can be detected. This is a popular method of studying agricultural practices, but a lot of research to date has focused on identifying rice types and rice growth stages, or predicting how much rice a given filed is likely to produce. Limited research on fallow has been undertaken, and this dissertation fills this gap.

A series of images, representing different stages of cropping in a number of fields in two different areas in the Philippines, were analysed using graphs and statistical tests. The dates of selected rice cropping stages, from planting to harvest and the start and the end of fallow, were estimated on the basis of questionnaires filled out by farmers. This allowed to compare the dates reported by the farmers with the satellite images obtained at a similar time.

The results showed that the signal from the satellite, which is sent to the ground and reflects back, shows a clear continuous pattern of rice cropping and fallow. At the beginning, when rice is planted, a lot of the signal sent by the satellite bounces away from the exposed flooded soil. With the rice plant growth, more signal is picked up by the satellite, with the peak observed near harvest. At the start of fallow, the signal starts to decrease, because the soil exposure is greater, and reaches its lowest point at the end of fallow period. However, just focusing on the points of the fallow start and end, fallow period is overestimated. This is because fallow start is detected too early and fallow end too late. A better method to detect fallow was found by detecting the start and end of rice crop season and classifying anything else as fallow. This is referred to as a backward detection approach. In addition, this study considered whether there is a difference in fallow length based on the water source. Some fields rely on rain, while others have access to irrigation. The study showed that the water source affects fallow duration only in one area under study, where fallow tended to start later and last shorted in irrigated fields, compared to those which were rainfed.

• author: Langowski, Lukasz
• supervisor: Andy Nelson
• organization: Dept of Physical Geography and Ecosystem Science
• course: GISM01 20202
• year: 2021
• type: H2 - Master's Degree (Two Years)
• subject: Earth and Environmental Sciences
• keywords: Geography, Geographical Information System, GIS, Remote Sensing, Fallow, Detection, Sentinel-1A, Synthetic-aperture radar, SAR
• publication/series: Master Thesis in Geographical Information Science
• report number: 138
• language: English
• additional info: External supervisor: Prof. Andy Nelson, University of Twente
• id: 9069071
• date added to LUP: 2021-12-14 14:03:17
• date last changed: 2021-12-14 14:03:17