Spatial analysis of maize crop distribution in Telangana’s deccan plateau using multi-temporal Landsat 8 data

Introduction

Maize (Zea mays L.) is a cornerstone of global agriculture, serving as a critical source of food, animal feed, and industrial products. In India, it ranks as the third most important cereal crop after rice and wheat, with significant cultivation in Telangana under rainfed (kharif) and irrigated (rabi) conditions. Precise mapping and delineation of maize areas are essential for optimizing agricultural planning, yield estimation, and sustainable resource management, particularly in regions with varied agro-climatic conditions (Thenkabail et al., 2022) [12] . Traditional field-based surveys are often labor-intensive and limited in spatial scope, making them inadequate for large-scale crop monitoring. Remote sensing, particularly through platforms like Landsat 8 with its Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS), provides moderate-resolution imagery (15-100 m) and a 16-day revisit cycle, enabling comprehensive crop monitoring. Spectral indices such as the Normalized Difference Vegetation Index (NDVI) and Land Surface Water Index (LSWI) facilitate the differentiation of crop types and water regimes, crucial for mapping rainfed and irrigated areas to enhance agricultural productivity and water use efficiency. Recent advancements in remote sensing underscore its efficacy for crop mapping. Thenkabail et al. (2021) [11] emphasize the role of multi-temporal satellite data in global cropland monitoring, while Kumar et al. (2023) [8] demonstrate the integration of Sentinel-2 and Landsat data for crop-type classification in India. Zhang et al. (2024) [15] highlight the use of NDVI and LSWI for distinguishing irrigated and rainfed agriculture. Additionally, Gumma et al. (2022) [5] showcase machine learning with Landsat data for maize and rice mapping in South Asia, and Li et al. (2023) [9] explore multi-sensor data fusion for enhanced crop delineation. Recent studies by Wang et al. (2024) [14] and Patel et al. (2025) [10] further advance the application of deep learning and high-resolution imagery for precise crop monitoring, while Chen et al. (2023) [2] illustrate the integration of remote sensing with ground-based data for improved accuracy in agricultural mapping. This study focuses on mapping maize crops across eight Telangana districts (Karimnagar, Jagtial, Peddapalli, Rajanna Siricilla, Jayashanker Bhupalpally, Medak, Sangareddy, and Siddipet), spanning 19,263 km² in the Deccan Plateau Agro-climatic region. Using Landsat 8 data from 2015-2017, a hybrid approach integrating NDVI, LSWI, isodata clustering, and decision rule classification was employed to delineate maize areas during kharif (rainfed) and rabi (irrigated) seasons of 2016-17. These methodologies align with the current study's approach, which incorporates ground truth GPS data and weather parameters to ensure robust analysis of maize distribution and irrigation patterns, providing valuable insights for agricultural policy and resource management.

Materials and Methods

Description of Study Area

The study area selected was erstwhile Karimnagar and Medak districts of Telangana, currently covering eight districts viz., Karimnagar, Jagtial, Peddapalli, Rajanna Siricilla, Jayashanker Bhupalpally, Medak, Sangareddy and Siddipet falling in the Deccan Plateau Agro-climatic region of India (Figure 1). Geographically it is located between 17 0 .76'N to 18 0 .83'N latitudes and 77 0 .87'E to 79 0 .46'E longitudes. It covers an area of approximately 19,263 km 2 . 1). The GARMIN, GPS 72H GPS receiver in stand-alone mode in Fasal FDC application was used to collect the information regarding the geographical co-ordinates, soil type, crop type, stage of the crop and crop photo of the field points sites for the present investigation (Figure 2 and Table 2). ERADAS Imagine 2016 software was used for importing, image geo-referencing and rectification for studying crop changes. It also played a key role in preparing models and for classifying the satellite image. ARC GIS 10.8.1 was used for map composition and MS office was used for database preparation. , hence generated a subset for the area of interest as primary data analysis using district shape file. The information in the false colour composite was not related to the real world. Hence geometric correction was performed through georeferencing, which established the relation between row/column numbers and real world coordinates using UTM projection with WGS-84 spheroid datum (Table 3). Thus, all the data sets were brought to a common projection. This ensured the satellite imagery to have a spatial reference with the real world.

Mapping and delineation of maize area

Delineation of maize crop in the study area, a hybrid approach was followed using 15 days OLI optical data (Landsat 8) for the year 2016-17. For mapping kharif and rabi maize temporal datasets from October to November and November to February period respectively were processed and NDVI was computed using stacked NDVI images for the defined period. NDVI measures the distinction between near-infrared (which vegetation strongly reflects) and red light (which vegetation absorbs). The stacked NDVI was masked with potential agricultural area and it is followed by isodata clustering generating 100 classes and 25 iterations. These clusters are labeled as maize / other classes based on field data points using decision rule approach. Decision rule was based on the maize crop profile based on top profile of maize crop in that area. The maize area delineated with remote sensing was compared with the observed data obtained from Area and Production Statistics, 2016-17.

Mapping and delineation of maize Irrigated and rainfed area

Three years data was used to delineate the potential irrigated zone in the study area. For this purpose we have emphasized on rabi season where the effect of rainfall on the crop was nullified. Irrigated zones were delineated based on assumptions of healthy maize crop (higher NDVI) represents irrigated area. Hence for delineating potential irrigated area the temporal datasets OLI (15 days composite) was processed during the period (October to February) for three years (2015-18). NDVI computed for respective image and stacked seasonal NDVI and LSWI were created. NDVI indicates measure of the amount and vigor of vegetation while, LSWI is sensitive to the total amount of liquid water in the crop thus widely used to monitor moisture stress. These stacked NDVI and LSWI layer was masked with potential agricultural area. For mapping irrigated maize again hybrid approach was followed. Isodata clustering was carried out by giving 100 classes and 25 iterations then followed by a decision rule classification. Decision rule was developed based on the ground observation where irrigated maize crop was assured to be associated tagged with higher NDVI and LSWI throughout rabi season. The same step methodology was repeated for three years to generate three irrigated zones separately. To generate potential irrigated map union operation of three years datasets was carried out. The irrigated area delineated with remote sensing was compared with the observed data obtained from Area and Production Statistics, 2016-17. False color composite (2, 3 and 4 bands) was generated for the kharif 2016 and rabi 2016-17 from to demarcate the agricultural area from non-agricultural area by following steps viz., layerstack, mosaicking and extraction of the area of interest.

Results and Discussion Weather Information of the Study Area

The study area comprised Karimnagar, Jagitial, Peddapalle, Rajanna Siricilla, Jayashanker Bhupalpally, Medak, Siddipet, Sangareddy districts. The weather parameters viz., rainfall, maximum and minimum temperature were observed during the crop growing period. The detailed weather prevailed during kharif 2016 at eight districts are illustrated in Figures 3 and4. Maize Area during Kharif 2016 Among the eight districts monitored for maize area during kharif 2016 (Table 4), Siddipet recorded the large, homogeneous and landscape dominating maize area of 1,01,381 ha as against the mapped area (81,136 ha) with -20% deviation followed by Medak and Jagitial registering an area of and 24,948 ha respectively. Sangareddy and Karimnagar were the next best districts recording maize areas of 20,462 ha and 18,643 ha. Jayashanker Bhupalpally followed by Rajanna Siricilla, Peddapalli recorded the lowest maize area of 2,890, 5,592 and 8,065 ha, respectively. The mapped area also followed the same trend as the observed area. Maize area maps during kharif 2016 covering the study area are depicted in Figure 7. Data showed that a good agreement between observed and mapped maize area and the lowest deviations ranged from 3.8 and to 9.4% was observed in Peddapalli and Karimnagar districts among all eight districts mapped. However, the observed maize area (2,13,326 ha) was lower by 4.2% compared to the total area mapped (2,04,278 ha).

Maize Area during Rabi 2016-17

Among the eight districts monitored for maize area during rabi 2016-17 (Table 5), Karimnagar recorded the maize area of 16,234 ha as against the mapped area 11,072 ha with deviation of -16.0%, followed by Siddipet and Jagitial registered an area of 13,340 ha and 11,270 ha respectively. Medak and Peddapalli have recorded maize areas of 9,931 ha and 4,490 ha. Rajanna Siricilla, Sangareddy and Jayashanker Bhupalpally districts recorded the lowest area of maize 1,013; 2,051 and 2,792 ha, respectively. The mapped area also followed the same trend as the observed area. Maize area maps during rabi 2016-17 covering the study area depicted in Figure 8. Data showed a good agreement between observed and mapped maize area with the lowest deviations ranged from -10.0 and -10.6% was observed in Jagtial and Peddapalli districts among districts mapped. However, the observed maize area (61,121 ha) was lower by 8.5% compared to the total area mapped (55,915 ha). The short-term variations of sown areas show a high temporal variation over recent decades in the Hyderabad, with 41 to 58% of sown areas in kharif and 8 to 22% in rabi between 2005 and 2016 (Ferrant et al., 2017) [3] .

Irrigated Area Mapping

Under irrigation during rabi (Table 6) Jagtial district showed the normal maize area of 93,859 ha as against the mapped area 1,02,082 ha with a deviation of 8.8% followed by Karimnagar and Siddipet registered an area of 81,188 against 77,569 ha and 74,257 against 83,146 ha with deviations of -4.5 and 12%, respectively. Peddapalli and Medak were the next best districts recorded a normal area 67,188 as against 75,655 ha and 59,275 ha as against 62,459 ha with the deviation of 12.6% and 5.4%, respectively. Rajanna Siricilla (38,342 ha) followed by Sangareddy (38,138 ha) and Jayashanker Bhupalpally (34,344 ha) recorded the lowest normal maize area, respectively. Irrigated area maps covering the study area are depicted in Figure 9. Data showed that a good agreement between normal and irrigated area mapped. The lowest deviations ranged from 2.3 -4.5% was in Rajanna Siricilla and Karimnagar districts. However, the total irrigated area (4,86,591 ha) was lower by 6.2% compared to the total area mapped (5,16,926 ha). Census of India, the irrigated area varies between 15-17%. Large changes in land cover over a small time interval of 10 years was seen in the statistics of different studies are not possible and thus are attributed to other causes. Irrigated areas in two remote sensing applications were 35 and 16%, respectively at crop mapping of Malaprabha, Karnataka (Heller et al., 2012) [6] . Kumar et al. (2015) [7] found areas under paddy and cotton were 39 and 29% respectively in mapped crop areas in the Fatehabad district of Haryana. Recent studies, Velpuri et al. (2021) [13] , Thenkabail et al. (2022) [12] and Zhang et al. (2024) [15] , highlighted the efficacy of NDVI and LSWI in distinguishing irrigated from rainfed crops, while Abuzar et al. (2023) [1] and Gao et al. (2024) [4] emphasize multi-sensor data fusion and machine learning for precise irrigation mapping. These methodologies align with the current study's approach, reinforcing the reliability of remote sensing for irrigated area delineation and its potential for informing water management strategies.

Conclusion

The study successfully mapped maize crop areas across eight Telangana districts using Landsat 8 data, demonstrating the effectiveness of remote sensing techniques in agricultural monitoring. The hybrid approach, integrating NDVI and LSWI with isodata clustering and decision rule classification, accurately delineated maize and irrigated areas, with deviations ranging from 3.8-20% for maize and 2.3-12.6% for irrigated zones compared to observed data. Siddipet and Karimnagar emerged as key maize-producing districts, with significant areas under both rainfed (kharif) and irrigated (rabi) conditions. Weather data underscored the influence of climatic factors on crop growth, with distinct temperature and rainfall patterns across seasons. The methodology ensured robust spatial alignment through geo-referencing and UTM projection, enhancing the reliability of the maps. These findings provide valuable data for agricultural planning and policy-making in Telangana, with potential applications in other agro-climatic regions for crop monitoring and resource management.

Acknowledgement

AcknowledgementThe authors recognize the assistance and financial, infrastructure and technical support provided by the Department of Agronomy, Professor Jayashankar Telangana State Agricultural University, Rajendranagar, Hyderabad.

References

1. Abuzar M, Sheffield K, McAllister A. Mapping irrigated agriculture using Sentinel-1 and Sentinel-2 data: A case study in northern Victoria, Australia. Agricultural Water Management. 2023;280(1):108198.
2. Chen Y, Lu D, Moran E. Integrating remote sensing and ground-based data for improved crop mapping in heterogeneous landscapes. Computers and Electronics in Agriculture. 2023;205(1):107623.
3. Ferrant S, Selles A, Le Page M, Herrault PA, Pelletier C, Al-Bitar A, et al. Detection of irrigated crops from Sentinel1 and Sentinel-2 data to estimate seasonal groundwater use in South India. Remote Sensing. 2017;9(11):1119-1138.
4. Gao Q, Zribi M, Escorihuela MJ. Multi-sensor data fusion for irrigated area mapping using Sentinel-1 SAR and Landsat 8/9 optical data. Remote Sensing of Environment. 2014;302(1):113987.
5. Gumma MK, Thenkabail PS, Teluguntla P. Mapping rice and maize areas in South Asia using multi-temporal Landsat and machine learning approaches. International Journal of Applied Earth Observation and Geoinformation. 2022;104(1):102567.
6. Heller E, Rhemtulla JM, Lele S, Kalacska M, Badiger S, Sengupta R, et al. Mapping crop types, irrigated areas, and cropping intensities in heterogeneous landscapes of southern India using multi-temporal medium-resolution imagery: Implications for assessing water use in agriculture. Photogrammetric Engineering and Remote Sensing. 2012;78(8):815-827.
7. Kumar A, Sheoran HS, Yadav M, Hooda RS. Geospatial approach for block-wise area estimation of kharif season crops in Fatehabad district, Haryana. International Journal of Science Engineering and Technology Research. 2015;4(1):148-153.
8. Kumar S, Jain M, Singh R. Mapping crop types in India using multi-temporal Sentinel-2 and Landsat data: A comparative analysis. Remote Sensing of Environment. 2023;287(1):113456.
9. Li J, Roy DP, Zhang H. Multi-sensor data fusion for crop mapping using Landsat and Sentinel-2: A review and applications. Remote Sensing. 2023;15(4):987-1003.
10. Patel R, Sharma V, Kumar A. Deep learning approaches for high-resolution crop mapping in tropical agro-ecosystems using Sentinel-2 data. Precision Agriculture. 2025;26(2):345-362.
11. Thenkabail PS, Gumma MK, Teluguntla P. Remote sensing of global croplands for food security. Remote Sensing. 2021;13(8):1543-1561.
12. Thenkabail PS, Velpuri NM, Gumma MK. Remote sensing of irrigated agriculture: Advances and challenges in global cropland mapping. ISPRS Journal of Photogrammetry and Remote Sensing. 2022;184(1):234-250.
13. Velpuri NM, Thenkabail PS, Gumma MK. Mapping irrigated areas in South Asia using MODIS and Landsat data with machine learning algorithms. Remote Sensing. 2021;13(16):3245-3262.
14. Wang Y, Zhang Q, Yu F, Zhang N, Zhang X, Li Y, et al. Progress in research on deep learning-based crop yield prediction. Agronomy. 2024;14(10):2264.
15. Zhang C, Di L, Lin L. Multi-temporal analysis of NDVI and LSWI for irrigated and rainfed crop mapping using Landsat and Sentinel-2 data. Agricultural Water Management. 2024;275(1):108032.